Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Essays on Human Capital,Innovation, and Growth withHeterogeneous Abilities
A thesis submitted to the University of Manchester for the degree of
Doctor of Philosophy
in the Faculty of Humanities
2017
King Yoong Lim
School of Social SciencesEconomics
Contents
List of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Declaration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Copyright Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Dedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Introduction 14
1 Industrial Transformation with Heterogeneous Labour and ForeignExperts 211.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.2 FDI Heterogeneity in Developing Host Economies . . . . . . . . . . . 24
1.3 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.3.1 Domestic Sectors in Host Economy . . . . . . . . . . . . . . . 29
1.3.2 Foreign Sector . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
1.3.3 Government and Market-clearing Conditions . . . . . . . . . . 51
1.3.4 Dynamic System and Steady State . . . . . . . . . . . . . . . 53
1.4 Model Parameterisation . . . . . . . . . . . . . . . . . . . . . . . . . 57
1.5 Policy Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
1.5.1 Individual Policies . . . . . . . . . . . . . . . . . . . . . . . . 65
1.5.2 Composite Policy Reform Programmes . . . . . . . . . . . . . 76
1.5.3 Endogenous Technological Change and Policy Complementar-
ities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
1.6 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
1.7 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
1.7.1 Estimation of FDI composition data . . . . . . . . . . . . . . 84
1.7.2 Technical Notes . . . . . . . . . . . . . . . . . . . . . . . . . . 87
2
1.8 Tables and Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
2 Unemployment, Growth and Welfare Effects of Labour Market Re-forms 1142.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
2.2 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
2.2.1 Individuals . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
2.2.2 Final Good . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
2.2.3 Intermediate Goods . . . . . . . . . . . . . . . . . . . . . . . . 125
2.2.4 Innovation Sector . . . . . . . . . . . . . . . . . . . . . . . . . 126
2.2.5 Government . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
2.2.6 The Labour Market . . . . . . . . . . . . . . . . . . . . . . . . 130
2.2.7 Savings-Investment Balance . . . . . . . . . . . . . . . . . . . 133
2.3 Balanced Growth Equilibrium . . . . . . . . . . . . . . . . . . . . . . 133
2.4 Properties of the Equilibrium . . . . . . . . . . . . . . . . . . . . . . 134
2.5 Model Parameterisation . . . . . . . . . . . . . . . . . . . . . . . . . 137
2.6 Policy Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
2.6.1 Reduction in Minimum Wage . . . . . . . . . . . . . . . . . . 147
2.6.2 Reduction in Unemployment Benefit Rates . . . . . . . . . . . 148
2.6.3 Reduction in the Union’s Wage Mark-Up . . . . . . . . . . . . 150
2.6.4 Reduction in Training Cost . . . . . . . . . . . . . . . . . . . 151
2.7 Composite Reform Programmes . . . . . . . . . . . . . . . . . . . . . 153
2.7.1 Core Programmes . . . . . . . . . . . . . . . . . . . . . . . . . 153
2.7.2 Infrastructure Investment . . . . . . . . . . . . . . . . . . . . 155
2.7.3 Policy Externalities . . . . . . . . . . . . . . . . . . . . . . . . 158
2.8 Sensitivity Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
2.8.1 Reduction in Minimum Wage . . . . . . . . . . . . . . . . . . 159
2.8.2 Reduction in Unemployment Benefit Rates . . . . . . . . . . . 161
2.8.3 Reduction in the Union’s Wage Mark-Up . . . . . . . . . . . . 163
2.8.4 Reduction in Training Cost . . . . . . . . . . . . . . . . . . . 165
2.8.5 Composite Reform Programmes . . . . . . . . . . . . . . . . . 166
2.8.6 ModelWithout Unemployment Benefit Consideration for Train-
ing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
2.9 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
2.10 Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
2.10.1 Dynamic Form . . . . . . . . . . . . . . . . . . . . . . . . . . 171
2.10.2 Welfare Evaluation . . . . . . . . . . . . . . . . . . . . . . . . 182
3
2.11 Tables and Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
Summary and Conclusion 209
Bibliography 211
Final Word Count: 38,978 words
4
List of Tables
1.1 Calibrated Parameter Values: Benchmark for Host Economy . . . . . 98
1.2 Calibrated Parameter Values: Benchmark for Foreign sector . . . . . 99
1.3 Calibrated Parameter Values for Generalised Logistic Curve . . . . . 99
1.4 Individual Policies: Steady-state Effects . . . . . . . . . . . . . . . . . 100
1.5 Composite Reform Programmes: Steady-state Effects . . . . . . . . . 101
1.6 Composite Reform Programmes: Steady-state Effects (continue) . . . 102
1.7 Sensitivity Analysis: Endogenous Technological Change with Com-
posite Reform Programmes: Steady-state Effects . . . . . . . . . . . . 103
1.8 Policy Complementarities: Comparison across Composite Programme
A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
2.1 Calibrated Parameter Values: Benchmark Case . . . . . . . . . . . . 185
2.2 Initial Steady-State Values of Key Variables . . . . . . . . . . . . . . 186
2.3 High-Income Economy: Summary of Benchmark Policy Results . . . 187
2.4 Middle-Income Economy: Summary of Benchmark Policy Results . . 188
2.5 Summary of Benchmark Composite Reform Programmes . . . . . . . 189
2.6 Sensitivity Analysis: Policy Experiments for (i) Reduction in Base
Minimum Wage, and (ii) Reduction in Untrained Workers’UB In-
dexation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
2.7 Sensitivity Analysis: Policy Experiments for (i) Reduction in Spe-
cialised Workers’UB Indexation, and (ii) Reduction in Both UB In-
dexation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
2.8 Sensitivity Analysis: Policy Experiments for (i) Reduction in Un-
trained UnionMark-up, and (ii) Reduction in Specialised UnionMark-
up over Target Wage . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
2.9 Sensitivity Analysis: Policy Experiments for (i) Advanced Education
Cost Cut, and (ii) Composite Reform Programme A . . . . . . . . . . 193
2.10 Sensitivity Analysis: Policy Experiments for (i) Composite Reform
Programme B, and (ii) Composite Reform Programme C . . . . . . . 194
4
2.11 High-Income Economy: Sensitivity Analysis, comparison between Bench-
mark Model and Model without UB consideration . . . . . . . . . . . 195
2.12 Middle-Income Economy: Sensitivity Analysis, comparison between
Benchmark Model and Model without UB consideration . . . . . . . 196
5
List of Figures
1.1 Estimated FDI Composition fromU.S. to selected East Asian Economies,
1999-2008 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
1.2 Production and Labour Allocation in Host Economy . . . . . . . . . 106
1.3 Foreign Sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
1.4 Policy Experiment for Skills Acquisition Cost Cut . . . . . . . . . . . 107
1.5 Policy Experiment for Labour Hiring Cost-mark up Reduction in the
Innovation Sector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
1.6 Policy Experiment for Investment Incentive targeted only at Vertical
Multinationals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
1.7 Policy Experiment for Investment Incentive targeted only at Horizon-
tal Multinationals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
1.8 Policy Experiment for economy-wide Investment Liberalisation for
All Foreign Multinationals . . . . . . . . . . . . . . . . . . . . . . . . 111
1.9 Policy Experiments for Composite Policy Reform Programmes . . . . 112
1.10 Industrial Composition Ratio - Composite Policy Reform Programme
A (Absolute deviation from baseline) . . . . . . . . . . . . . . . . . . 113
2.1 Overview of Production Structure and Labour Market . . . . . . . . . 197
2.2 Individual and Composite Experiments: Steady-state effects . . . . . 198
2.3 Transitional Dynamics of Composite Reform Programme A . . . . . . 199
2.4 Transitional Dynamics of Composite Reform Programme B . . . . . . 200
2.5 Transitional Dynamics of Composite Reform Programme C . . . . . . 201
2.6 Sensitivity Analysis: Transitional Dynamics of A Permanent Decrease
in Base Minimum Wage . . . . . . . . . . . . . . . . . . . . . . . . . 202
2.7 Sensitivity Analysis: Transitional Dynamics of A Permanent Decrease
in Untrained Workers’UB Indexation . . . . . . . . . . . . . . . . . . 203
2.8 Sensitivity Analysis: Transitional Dynamics of A Permanent Decrease
in Specialised Workers’UB Indexation . . . . . . . . . . . . . . . . . 204
2.9 Sensitivity Analysis: Transitional Dynamics of A Permanent Decrease
in Both UB Indexation . . . . . . . . . . . . . . . . . . . . . . . . . . 205
6
2.10 Sensitivity Analysis: Transitional Dynamics of A Permanent Decrease
in Untrained Union’s Mark-Up over Target Wage . . . . . . . . . . . 206
2.11 Sensitivity Analysis: Transitional Dynamics of A Permanent Decrease
in Specialised Union’s Mark-Up over Target Wage . . . . . . . . . . . 207
2.12 Sensitivity Analysis: Transitional Dynamics of A Permanent Decrease
in Advanced Education Cost . . . . . . . . . . . . . . . . . . . . . . . 208
7
AbstractThe critical role of human capital in promoting innovation-driven growth has
long been recognised, though when in the presence of heterogeneous abilities among
individuals, its complex interactions with other cross-cutting factors in an economy
are less understood. A rigorous examination of these links is important towards
gaining better understanding of externalities among policies, notably in the context
of real-world policymaking where reforms are often implemented in packages. This
thesis examines the links of human capital (with heterogeneous abilities), growth,
and two such policy areas, foreign direct investment (FDI) and labour market re-
forms, using multisectorial endogenous growth models.
Chapter 1 develops an imitation-innovation (continuous time) growth model
with heterogeneous labour and foreign multinationals (MNCs) to examine industrial
transformation for a developing host economy. With FDI modelled at the disaggre-
gated level of foreign experts, we formalise a MNC composition-determination frame-
work that explains Dunning’s ‘internalisation advantage’(1977) as being driven by
the presence of asymmetric views on productivity of domestic workers. Specifically,
foreign experts perceive heterogeneity among the productivity of domestic workers.
As productivity is a transformation of ability, this allows us to link the skills ac-
quisition decision and foreign subsidiaries’operational mode choice along the same
ability distribution in the host economy. In addition, asymmetry is also introduced
specifically for Vertical MNCs to capture the increasingly costly nature for foreign
experts to identify the best among the most productive workers in a host econ-
omy. Calibrated for Malaysia, these novel features enable the model to generate
simulation results that are consistent with some stylised observations documented
in the FDI literature, and uncover complementarities between human capital and
FDI-promoting policies. These complementarities are stronger with endogenous
technological change.
In Chapter 2, the effects of labour market reforms are studied in an innovation-
driven, overlapping generations (OLG) model of endogenous growth with a hetero-
geneous labour force, labour market rigidities, and structural unemployment. The
model is parameterised for stylised high- and middle-income economies and used to
perform a range of experiments, including both individual labour market reforms
(cuts in the minimum wage and unemployment benefit rates) and composite reform
programmes involving additional measures. The results show that individual re-
forms may generate conflicting effects on growth and welfare in the long run, even
in the presence of positive policy externalities. A reduction in training costs may
also create an oversupply of qualified labour and higher unemployment in the long
8
run. Public investment in infrastructure, partly through its productivity effects on
innovation, can help to mitigate this oversupply problem.
In short, the studies in these two chapters show that, when the supply side
of the labour market is explicitly modelled by introducing heterogeneous abilities,
promoting innovation-driven growth is no longer a straightforward reform provision
of “throwing everything at the wall to see if it sticks”, as there are much more
complex interactions in terms of policy externalities to be understood.
9
DeclarationI declare that no portion of the work referred to in the thesis has been submitted
in support of an application for another degree or qualification of this or any other
university or other institute of learning.
10
Copyright Statementi. The author of this thesis (including any appendices and/or schedules to this
thesis) owns certain copyright or related rights in it (the “Copyright”) and he
has given The University of Manchester certain rights to use such Copyright,
including for administrative purposes.
ii. Copies of this thesis, either in full or in extracts and whether in hard or
electronic copy, may be made only in accordance with the Copyright, Designsand Patents Act 1988 (as amended) and regulations issued under it or, where
appropriate, in accordance with licensing agreements which the University has
from time to time. This page must form part of any such copies made.
iii. The ownership of certain Copyright, patents, designs, trade marks and other
intellectual property (the “Intellectual Property”) and any reproductions of
copyright works in the thesis, for example graphs and tables (“Reproduc-
tions”), which may be described in this thesis, may not be owned by the
author and may be owned by third parties. Such Intellectual Property and
Reproductions cannot and must not be made available for use without the
prior written permission of the owner(s) of the relevant Intellectual Property
and/or Reproductions.
iv. Further information on the conditions under which disclosure, publication and
commercialisation of this thesis, the Copyright and any Intellectual Property
and/or Reproductions described in it may take place is available in the Uni-
versity IP Policy
see http://www.campus.manchester.ac.uk/medialibrary/policies/intellectualproperty.pdf,
in any relevant Thesis restriction declarations deposited in the University Li-
brary, The University Library’s regulations
see http://www.manchester.ac.uk/library/aboutus/regulations
and in The University’s policy on presentation of Theses.
11
Dedication
To my beloved parents,
and my sister, Misel, for their sacrifice and unwavering support
12
Acknowledgement
First and foremost, I would like to thank Professor Pierre-Richard Agénor, mymain supervisor, for his dedicated and invaluable guidance. His supervision over thisperiod is instrumental in my overall development, and I believe this is well-reflectedin the thesis. I would also like to thank Christoph Himmels, Kyriakos Neanidis, andTimo Trimborn for their helpful comments and feedback on a previous version ofChapter 1 of this thesis.Besides, I have also received helpful feedback provided by the many participants
in various conferences over this period. The comments received contributed towardsthe many revisions and improvements of the two chapters in this thesis. These areduly recognised as follows. Previous versions of both chapters have been presentedin contributed sessions of the 66th Annual Congress of the French Economic Asso-ciation (AFSE) and the Macroeconomic Seminar hosted by the Centre for Growthand Business Cycle Research, University of Manchester. In addition, for Chapter1, valuable feedback were also received when it was presented at the first Interna-tional Development Economics (GDRI-IDE) Conference hosted by CERDI, at theninth FIW Research Conference on International Economics hosted by the Univer-sity of Vienna in 2016, as well as during the revise and resubmission process of itscorresponding working paper with the journal, Macroeconomic Dynamics.On the other hand, in addition to AFSE, for Chapter 2, I would like to ac-
knowledge the participants’comments received during the presentation in the RoyalEconomic Society 2017 Annual Conference and the Scottish Economic Society 2017Annual Conference. I am also grateful to Baris Alpaslan for kindly reviewing aprevious version of Chapter 2.Lastly, I would also like to thank my immediate family members, fellow PhD col-
leagues in Manchester, and former colleagues and friends from Khazanah Nasionaland Securities Commission Malaysia (notably the three mentors who supported myPhD endeavour back then: Roselee Shah Shaharudin, Nungsari Ahmad Radhi, andAlbert Gaspard Zeufack) for their support and encouragement. Words cannot ex-press how grateful I am to my parents and sister for all of their sacrifices throughoutthe duration of the doctoral programme.
13
Introduction
Human capital is at the heart and soul of a successful high-income strategy. As
middle-income economies seek to improve their competitiveness and to climb the
income ladder, policymakers often formulate economic transformation strategies in
an attempt to move from imitation- to innovation-driven. The latter is especially
knowledge intensive, which makes having the foundation of a deep and quality stock
of human capital essential because knowledge is embodied in human capital. Indeed,
it can be argued that the role of quality human capital extends to the context of
structural reforms in a developed economy, which is increasingly a significant policy
agenda as major developed economies attempt to drag their economies out from
sluggish growth since the 2008-09 global financial crisis. As such, human capital
is a core component in the growth agenda for both middle-income and developed
economies.
In the literature of economic growth, the conventional approach to the modelling
of human capital is based on the Uzawa-Lucas tradition. The Uzawa-Lucas model is
a human capital accumulation-based endogenous growth model in which human cap-
ital accumulation is specified as akin to another type of capital. This class of models
explain long-run economic growth as a consequence of human capital accumulation.
The problem of this approach is that, it is based on a disembodied interpretation of
human capital, which makes subsequent reconciliation to the actual data of labour
shares– —which by definition is embodied in nature– —inappropriate. By modelling
human capital in the same way of modelling physical capital, the Uzawa-Lucas
framework also essentially allows human capital to grow without bounds, which is
slightly unrealistic. Furthermore, if the different abilities of different individuals were
to matter (which is often the case in determining the proportion of individuals that
have high enough cognitive abilities to work in innovation), a modelling approach
that accounts for this additional quality dimension will be needed. I therefore adopt
14
an embodied human capital modelling approach similar to those in Agénor and Dinh
(2013), where the supply of human capital in an economy is modelled using a con-
tinuous distribution of abilities, from which the probability distribution moments
will allow a direction determination of the proportion of skilled workers within an
economy. In both chapters of this thesis, the embodied human capital modelling
approach is a common theme.
In addition, the presence of heterogeneous abilities often affects an economy
beyond the labour supply when the role of human capital is examined in a high-
dimensional, multisectorial growth model with a taxonomy of policy factors in play.
In such models, the modelling of the knowledge sector usually assumes a horizon-
tal innovation structure (Romer 1990), which then allows for the examination of
the complex interactions of human capital and other cross-cutting factors in an
economy. Two relatively interesting yet underexplored areas in the endogenous
growth literature concern the interactions of heterogeneous human capital and FDI
and unemployment. Indeed, the former remains one of the most important policy
agenda for developing economies, while the latter for developed (and selected middle-
income) economies. Both play crucial roles in determining the potential success of
any reform strategy, and therefore represent the main policy themes examined in
this thesis (FDI in the first chapter, followed by unemployment in the second). The
dissertation therefore consists of two substantive chapters, with the content of each
chapter briefly discussed below.
Chapter 1 is a relatively ambitious exercise due to its attempt to incorporate
many development policy elements and to model them in a formal manner. The
main motivations of Chapter 1 can be understood in the overarching context of the
middle-income trap. According to studies such as Agénor and Canuto (2015b) and
Agénor (2016), a middle-income economy previously relying heavily on imitation-
based industrial strategies to escape poverty quickly can get ensnared in the middle-
income trap when she experiences a prolonged growth slowdown due to an inability
15
to switch successfully to innovation-driven growth.1
Many interesting and contrasting examples can be seen during the East Asian
experience in the 1990s, where the presence of FDI is significant yet does not always
lead to successful industrial transformation. Studies such as Agénor (2016) posit that
the failure of an economy in escaping middle-income trap can often be attributed
to the economy not having the appropriate policy combination to improve the three
elements of technology transfer, absorption capacity, and diffusion. In a FDI litera-
ture that is scarce in formal theoretical studies explaining the relative importance of
the different types of FDI from the perspective of the developing host economy2, this
chapter develops an industrial transformation model with heterogeneous FDI and
labour that would account for these elements. Further, against the backdrop of the
ever-changing nature of modern foreign multinational activities3, FDI is modelled
in the disaggregated form of foreign experts, whose compositions are determined
using a stylised framework designed to explain the least modelled ‘internalisation
advantage’posited in the famous Eclectic Paradigm of Dunning (1977). Dunning
(1977) introduced the OLI (Ownership-Location-Internalisation advantages) frame-
work to explain the international activities of multinationals as being driven by
ownership-specific, location-specific, and internalisation advantages. In essence, the
1Many middle-income economies appear to have experienced this same fate: Based on WorldBank data in 2012, there were only 13 out of the 101 middle-income economies in 1960 thatsuccessfully moved up to become high-income by 2008. As much as a developing economy can useimitation-based strategy to escape from poverty trap quickly, the same strategy often becomes themajor impediment that holds back the economy from switching successfully to innovation-driven.This results in the middle-income trap.
2Existing theoretical contributions on the role and determinants of FDI as a vehicle of interna-tional technology transfer have mostly concentrated on studying the determinants of internationalproduction choices by foreign MNCs in either international trade theory-motivated frameworkin the tradition of Helpman (1984) or models with underpinning industrial organisation theoriesheadlined by Markusen (1984).
3This phenomenon of an evolving characteristic of modern foreign enterprises in developingeconomies is documented in the literature on global talent management, such as Scullion andBrewster (2001), Scullion et al. (2007), and McDonnell et al. (2010). Instead of selling goods, theactivities of foreign enterprises are increasingly characterised by top quality advisory services andknow-how, with the identity of foreign subsidiaries often being assumed by the human capital of theforeign experts or assignees. Based on a sample of global multinationals, PricewaterhouseCoopers(2012) document that by 2020, it is expected that there will be 33 host locations for assignees fromthese companies. In addition, an average of 370 assignees per organisation is expected.
16
OLI framework links the strength of the firms, be it in physical or human capital
endowments, to location-specific factors determined by the institutional and policy
factors of a host economy, in influencing the internalisation decisions made. Of the
three main determinants posited, the ownership-specific and location-specific ad-
vantages have been well-incorporated in many theoretical contributions to model
and explain FDI. This chapter formalises a rigorous theoretical framework in an
attempt to explain the mechanism of the internalisation advantages. Specifically,
foreign firms’internalisation decisions are modelled as being driven by the presence
of asymmetric views on the productivity of domestic workers.
We solve for a dynamic system that allows for the determination of both the
skills acquisition choice of domestic workers and the foreign subsidiaries’operational
mode choice along the same ability distribution of the host economy. A parame-
terised version based on Malaysia– – a classic case of middle-income trap economy
despite having presence of leading foreign multinationals in innovation (Hill et al.
2012)– —is examined using various policy experiments. The model allows for the
generation of transitional dynamics (this in itself is another notable contribution)
that are consistent with some stylised facts surveyed from the various branches of the
FDI literature, as well as uncovering policy complementarities between human cap-
ital and FDI-promoting policies. Specifically, the main contributions of this chapter
are the novelty of some of the policy experiment results. Due to the introduction
of the laddered approach in modelling foreign experts, as well as the asymmetry
between Vertical FDI and other multinationals, the fundamentals– in this instance
the productivity of domestic workers– of a host economy becomes a much more im-
portant factor in attracting the best FDIs than direct investment incentives. This
can be referred to in a seemingly counterintuitive result associated with an individ-
ual FDI-promoting policy experiment in the chapter, which underlines the relevancy
of the adverse signalling effects documented in the ‘race-to-the-bottom’literature
and necessitates some degree of caution when come to designing foreign investment
17
incentives in a developing host economy. Overall, the implementation of foreign
investment liberalisation measures that are balanced and targeting all types of for-
eign firms is more innovation- and skills acquisition-promoting than disproportionate
ones biased towards selected types of foreign firms. Further, in the context of the
parameterised version of the model, a threshold doing-business cost value is also
identified for the FDI-growth nexus, below which standalone investment liberalisa-
tion measure is no longer enough to drive output growth. The results also underline
the importance of combining human capital and FDI-promoting policies to drive
industrial transformation, especially if the government of a host economy intends to
maximise the benefits of policy complementarities.
In Chapter 2, we switch focus to analyse another pressing structural issue that
is closer to the heart of policymakers in developed economies: long-term unemploy-
ment. Unlike in Chapter 1, unemployment is a topic frequently modelled in dynamic
macroeconomic models, notably search unemployment in the tradition of Mortensen-
Pissarides (Mortensen and Pissarides 1999). While search friction is popular among
the researchers whose interests lay primarily on explaining business cycle fluctua-
tions, it is inadequate in explaining long-term structural unemployment compared
to other labour market rigidities. Within the endogenous growth literature, there
are many studies focusing on explaining the effects of various labour market rigidi-
ties on long-term unemployment. Some of these studies, such as van Schaik and de
Groot (2000), Meckl (2004), and Zagler (2011) have dual labour market for wage
settings. In these specific studies, there is an innovation sector and some forms of
labour allocation features too. On the surface, the model developed in this chapter
will share many of the same features.
Nevertheless, the existing literature suffers from a few shortcomings. For in-
stance, almost all the models (with at least a dual labour market) reviewed do
not examine transitional dynamics, and therefore neglect the dynamic tradeoffs be-
tween growth and unemployment. In addition, the presence of individuals with
18
heterogeneous abilities is not accounted for when the supply of different workers are
examined. The main objective of Chapter 2 is therefore to develop a comprehensive
OLG model with various labour market institutions to address these issues. In com-
parison to Chapter 1, the model has a much richer feature to skills acquisition, with
multiple feedbacks from unemployment to labour supply. The aim is to develop a
prototype growth model with various labour market imperfections that will allow
for more realistic analysis of the impacts of labour market reforms on both growth
and welfare.
The key contributions of Chapter 2 include, in comparison to the existing labour
market reform literature, the richer features of the model allow me to consider com-
posite reform programmes, which are usually how labour market reforms are imple-
mented in real life. Labour market reforms are found to entail a two-way causality
between growth and unemployment: growth tends to lower unemployment, through
its impact on labour demand; but unemployment may lower growth because it re-
duces (through its wage signalling effects) incentives to acquire skills and constrains
the ability to expand innovation activities– a key engine of growth. In addition,
they may have conflicting effects on growth and welfare in the long run, a result
that is diffi cult to obtain when labour market reform is modelled in a piecemeal
approach. To some extent, this tradeoff can be tempered by exploiting policy exter-
nalities, though to avoid creating an oversupply of specialised workers, governments
must refrain from adopting policies that contribute to generating large numbers of
university graduates. The use of demand-side policies and the improvement of the
quality of education may prove to be more effective in supplementing conventional
labour market reforms.
Overall, the unifying themes of the two chapters are the use of the embodied
human capital with heterogeneous abilities approach, and the focus on composite
reforms. Both chapters share a few similar core production features: (i) there is a
final good sector, an intermediate input specification, and a knowledge production
19
sector; (ii) the S-U labour type and their respective allocation mechanisms where S-
type workers work in final good or innovation, while U-type workers in final good (or
imitation, in the context of Chapter 1). However, Chapter 1 focuses more on relating
human capital to FDI policies and the determination of the different MNCs in a
host economy. Chapter 2 differs in that the emphasis is on modelling labour market
rigidities more formally (hence introducing equilibrium unemployment), which is
merely introduced as a simple cost mark-up in Chapter 1. The policy implications
drawn from the two chapters would therefore differ along these lines too.
20
Chapter 1
Industrial Transformation withHeterogeneous Labour andForeign Experts
1.1 Introduction
Ever since Saggi (2002) documented the scarcity of studies modelling the relative
importance of the different types of FDI in the industrial transformation process
of developing economies, this remains an under-studied area in the growth litera-
ture. On industrial transformation, recent studies such as Agénor and Dinh (2013)
and Agénor and Alpaslan (2014) developed a growth framework with heterogeneous
labour to examine the non-linear transitional dynamics associated with industrial
transformation in a developing economy. However, they do not account for for-
eign MNCs, which play a significant role in the East Asian development experience
(Nelson and Pack 1999; Amsden 2001).
In terms of MNCs’role in developing economies, while literature surveys such as
Faeth (2009) indicate that the FDI phenomenon is largely a tale of heterogeneity,
the most prominent theory on MNCs’motives remains Dunning’s Eclectic Par-
adigm (1977). He introduces the Ownership-Location-Internalisation advantages
(OLI) framework to explain the international activities of MNCs as being driven
by ownership-specific, location-specific, and internalisation advantages. In essence,
the OLI framework links the strength of the firms, be it in physical or human cap-
21
ital endowments, to location-specific factors of a host economy in influencing the
internalisation decisions made. While the OLI framework is static, it suggests that
there appears to be sequential entry dynamics for foreign subsidiaries with regards
to the operational mode chosen for their activities in a host economy. Of the three
main determinants posited by Dunning, the ownership-specific and location-specific
advantages have been well-incorporated in many theoretical contributions (Faeth
2009), but there remains a vacuum for theoretical explanation of internalisation
advantages. Further, the internalisation decisions within MNCs with respect to
establishing foreign subsidiaries are often influenced by various micro-mechanisms
tied to the incentives of foreign experts. Indeed, it can be argued that, failure to
account for the micro-incentive (when FDI is modelled in its traditional form) may
have resulted in there often being mixed empirical evidence with respect to the
overall impact of FDI in promoting domestic innovation.1 Hence, to understand the
knowledge conduit role of FDI would necessarily involve modelling of the incentive
mechanism at a more disaggregated level.2
Given that a foreign expert-based, stylised ‘internalisation advantage’framework
for FDI is not an angle explored in the literature, this chapter examines industrial
transformation for a developing host economy by developing an imitation-innovation
model with heterogeneous labour and a stylised heterogeneous MNC composition-
determination framework, where MNC is modelled in the disaggregated form of
foreign experts, as suggested by Markusen and Trofimenko (2009). In the model,
the MNC composition-determination framework explains Dunning’s ‘internalisation
advantage’(1977) as being driven by the presence of asymmetric views on productiv-
ity of domestic workers. Specifically, foreign experts perceive heterogeneity among
1See Blomström and Sjöholm (1999) for examples of positive spillovers, while Haddad andHarrison (1993) and Djankov and Hoekman (2000) are examples with negative effects.
2Indeed, almost all of the literature on firm-level innovation capabilities building, such as thosereviewed in Bell and Figueiredo (2012), premise on an objective to overcome this problem byexamining the internal operations of foreign MNCs using mainly non-generalisable case studies.However, studies examining incentives at an even more disaggregated level– incentives of foreignexperts– remain scarce.
22
the productivity of domestic workers. As productivity is a transformation of ability,
the skills acquisition decision and foreign subsidiaries’operational mode choice are
linked along the same ability distribution in the host economy. This allows for the
examination of transitional dynamics of human capital and FDI-promoting policies,
so to uncover policy complementarities when a mixture of these policies are used.
Further, consistent with some well-documented stylised facts in the FDI literature,
an additional asymmetry between Vertical MNCs and other MNCs is also modelled.
This then enables us to lend some insights into the conventional debate on how best
to implement FDI-promoting policies in developing economies.
This chapter contributes to the literature by (i) developing an imitation-innovation
model with heterogeneous labour and foreign MNCs to examine industrial transfor-
mation for a developing host economy. With FDI modelled at the disaggregated
level of foreign experts, (ii) we formalise a MNC composition-determination frame-
work that explains Dunning’s ‘internalisation advantage’(1977) as being driven by
the presence of asymmetric views on productivity of domestic workers3. Further,
(iii) by linking domestic workers’skills acquisition choice and MNCs’operational
mode choice along the same ability distribution, the model allows for the examina-
tion of transitional dynamics of human capital and FDI-promoting policies, so to
uncover policy complementarities when a mixture of these policies are used. Sec-
tion 1.2 provides a brief discussion on the rationale of the modelling approach for
the FDI-composition framework, guided by the FDI literature on the various policy
issues that the model attempts to address. Section 1.3 presents the model. The
dynamic system derived is also presented in this section. Model parameterisation
is reported in Section 1.4. In Section 1.5, the various policy experiments analysed
are discussed. Section 1.6 draws together the policy implications and concludes the
chapter.
3Specifically, two sources of asymmetry are introduced: first between domestic and foreignfirms, and a second between Vertical MNCs and other MNCs. These are novel theoretical contri-butions in this study.
23
1.2 FDI Heterogeneity in Developing Host Economies
While macroeconomic studies examining FDI heterogeneity at the disaggregated
form of foreign experts are scarce, we can establish some contextual framework based
on the FDI literature. The most prominent early studies on the motive of foreign
MNCs as a driver of FDI flows can be referred to the Eclectic Paradigm of Dunning
(1977), who introduces the OLI framework to explain the international activities of
MNCs. Of the three main determinants posited by Dunning, the ownership-specific
and location-specific advantages have been well-incorporated in many theoretical
contributions to model and explain Vertical and Horizontal FDI (Faeth 2009).4 The
former tends to be explained as an equilibrium phenomenon due to factor endow-
ment differences across regions resulting in vertically-integrated firms, while the lat-
ter driven mainly by ownership-specific strength, such as Markusen’s (1984, 1995,
1998) knowledge capital models. Markusen’s contributions essentially explain FDI
as an outcome of MNCs capitalising on their unique proprietary knowledge, and are
therefore consistent with the ‘foreign expert’approach introduced in this chapter.
Based on the OLI framework (Dunning and Lundan 2008), typical entry and
operational decisions by foreign subsidiaries in a particular host economy reflect the
internalisation advantages of firms (first motivated by market-seeking objectives,
with subsequent investment being effi ciency-seeking), suggesting sequential entry
dynamics for foreign subsidiaries. To the extent that these subsidiaries are essen-
tially providing professional expertise, these also reflect sequential entry dynamics
for foreign experts. Consistent with Dunning’s explanation, firms are said to opt for
Horizontal over Vertical mode as the initial form of entry due to the know-how advan-
tage over rivals. Moreover, Vertical mode tends to be more costly too (Markusen
1995; Horstmann and Markusen 1996). Empirically, global FDI flows are docu-
mented by Brainard (1997) and Markusen and Maskus (2002) to be predominantly
4From this point onwards, the terminologies for FDI are used interchangeably with MNCdespite the difference in strict conceptual definition. For example, Horizontal/Vertical FDI modealso indicates Horizontal/Vertical MNCs.
24
driven by Horizontal MNCs. However, their definition of FDI composition is based
largely on the Horstmann-Markusen-Venables (HMV) interpretation (Horstmann
and Markusen 1987, 1992; Markusen and Venables 1999), which tends to ignore the
different aspects of factor endowment considerations that lead to a necessary further
distinction of vertically-integrated MNC activities.
As documented in international production fragmentation studies such as Athuko-
rala (2005), Athukorala and Hill (2010), the fragmented production process of
vertically-integrated MNCs often results in subsidiaries with vastly different resource
requirements, some being more skills- and technological-intensive than others. Be-
sides, the various FDI-targeting rules and ownership stipulations imposed in de-
veloping economies often inadvertently result in many non-mandated subsidiaries
or investment commitments made by vertically-integrated MNCs, in forms such as
technological licensing agreements (Saggi 2002). As MNCs often treat such commit-
ments as non-mandated subsidiaries internally or merely serving as manufacturing
platforms (Hanson et al. 2001), these result in MNCs that are neither imitation- nor
innovation-enhancing (see D’Costa (2002) and Hobday et al. (2004) for examples).
To account for these FDI modes, we group them as ‘Non-mandated MNC’.5
Indeed, once the non-mandatory FDIs is further differentiated from vertically-
integrated MNCs, we can define a hierarchy of internalisation decision-making with
regards to FDI mode, and the order of Non-mandated-Horizontal-Vertical matches
their respective importance in the host economy’s spillover. Due to factors such
as agency or information cost, MNCs tend to use basic Non-mandated mode as
default mode (Saggi 2002), which does not seem to play much of a role in driving
industrial development, save for in the poorest low-income economies deprived of
basic industrial structures. Both Horizontal and Vertical MNCs tend to invest in
knowledge-intensive industries and therefore prefer host economies with human cap-
5In the literature, common internationalisation modes explored also include direct exporting,offshoring, and more complicated vehicles of joint-ventures. These are related but peripheral issuesin the context of this chapter.
25
ital (Borensztein et al. 1998). However, given that the cost incurred by not getting
access to high quality human capital is much lower for horizontal operations, for-
eign firms would first opt for the Horizontal mode. Indeed, foreign subsidiaries are
only inclined to send in foreign experts with sophisticated innovation know-how if
the pool of human capital of a host economy is highly productive (Gersbach and
Schmutzler 2011). This implies that the top foreign experts coming in via Vertical
MNCs are likely to have an additional layer of preference to distinguish the brightest
of the most skilled workers.
To understand this in the context of developing economies, we conducted a pre-
liminary empirical exercise in estimating the different MNC composition for a num-
ber of East Asian countries.6 While the findings are largely consistent with the
studies reviewed, there are different MNC compositions across all the economies,
which necessitate the design of an elaborative framework for the determination of
MNC composition within a developing host economy undergoing industrial trans-
formation. Relating this to multi-sector growth models such as Funke and Strulik
(2000) and Agénor and Dinh (2013), this suggests different roles to be played by
the different types of MNCs across different production activities of a host economy.
Initial entrance by MNCs with non-mandated operations are likely significant only
to growth in low-income economies. For a middle-income economy with some hu-
man capital, a Horizontal MNC is likely to benefit the host economy in its imitation
activities, while a Vertical MNC would benefit the innovation activities.
Furthermore, a mixture of policies is often needed in the context of middle-
income economies since they do not have the appropriate policy combination to im-
prove technology transfer, absorption capacity, and diffusion (Agénor 2016). There
also appears to be indirect, nonlinear relationship between human capital and FDI-
promoting policies within a host economy (Javorcik 2004; Liu 2008; Kottaridi and
Stengos 2010), suggesting potential policy complementarities to be gained by using
6See Section 1.7.1 for the details, and Figure 1.1 for the estimated MNC compositions.
26
a mixture of these policies. For instance, empirical studies such as Blomström and
Kokko (2003) and Ciconne and Papaioannou (2009) documented complementarity
between FDI and education policies. Nonetheless, overly narrowed investment in-
centives have also been documented to result in adverse signalling effects in many
developing economies, in that many generous incentives targeted solely at top qual-
ity MNCs have often failed to achieve intended results. This is often the key finding
of the ‘race-to-the-bottom’literature on MNCs7, and is treated as a stylised fact for
MNCs in developing economies that we also seek to model.
1.3 The Model
The model host economy studied is populated with households consisting of in-
dividuals with different innate abilities. There are five production sectors, with
the modelling approach for the domestic production sectors adapted primarily from
Agénor and Dinh (2013), notably the production structures of final output and in-
termediate goods sectors. Knowledge growth in the host economy is assumed to
be driven by ‘horizontal’expansion of differentiated intermediate goods (IG) in the
tradition of Romer (1990).8 Skilled labour can work in either the final output sec-
tor or the innovation sector, earning skilled wages, whereas unskilled workers are
allocated between the final output and the imitation sector, earning unskilled wage.
The labour allocation mechanism in this economy is illustrated in Figure 1.2.
Productivity in both imitation and innovation sectors benefits from the presence
of foreign firms, though there is a largely separate foreign sector determining the
different types of foreign subsidiary mode operating in the host economy. The focus7For examples, see Oman (2000), Blomström (2002), Haaland and Wooton (2001), OECD
(2008), and Olney (2013).8In recent theoretical contributions, similar imitation-innovation trade-offs are modelled in
studies such as Benhabib et al. (2014) and Lucas and Moll (2014) using the Schumpeterian qualityladder framework. However, the matching of the overall empirical productivity distribution as anendogenous outcome of searching by heterogeneous agents is the primary emphasis in these studies.The primary interest here is to examine the interactions of human capital with other cross-cutting,horizontal factors in the economy, therefore the use of a Romerian framework would better suitthe purpose.
27
is on non-pecuniary externalities, which as pointed out in Saggi (2002), are the
critical (yet relatively unexplored) impacts of foreign MNCs on the development
of a developing host economy. It is assumed that there is only one foreign source
country that deploys subsidiary units in the form of experts to the host economy.
Dunning’s ‘internalisation advantage’seeks to understand how foreign MNCs shape
their ‘in-house’preference with respect to the involvement in different production
of a host economy. To construct a stylised framework that links this idea to the
human capital distribution of the host economy, we adopt a nested Dixit-Stiglitz
CES value function framework that is often used empirically to model heterogeneous
firms along a continuous distribution (see Brambilla et al. (2009) for example). It
is assumed that each subsidiary unit consists of one foreign expert with specific
process know-how that is only available in the foreign source country. Specifically,
standardisation know-how (used in imitation) for Horizontal MNC and sophisticated
know-how (used in innovation) for Vertical MNC. Consequently, the presence of
Vertical MNC is a necessary condition for innovation sector to exist. As our focus is
on middle-income economy with both imitation and innovation sectors, the role of
non-mandated subsidiaries in domestic production is largely abbreviated, modelled
only as a base entry mode.
As a result of foreign firms being effectively experts with specialised human cap-
ital, a dichotomous relationship exists between domestic and foreign firms. For
domestic firms, only the average productivity of workers matters. For foreign sub-
sidiaries, they perceive heterogeneity among the productivity of domestic workers.
As individual ability of domestic workers is not fully observable to foreign firms
(though they do know the overall distribution), for two different skilled workers
used to produce a same blueprint variety, foreign experts would have an additional
layer of preference to be ‘matched’to a worker with higher productivity– a trans-
formation of ability– hence resulting in a Melitz (2003) type of sorting process.
In deciding on operational mode, foreign experts are therefore sorted along the
28
ability distribution of the host economy, resulting in different threshold values for
different modes of operation. Consequently, these create an indirect link between
the foreign MNCs’operational mode choice and domestic workers’skills acquisition
decision. In other words, we explain ‘internalisation advantage’as resulting from
the implicit productivity requirement-induced information cost.9 Lastly, a demand
feedback channel from the industrial state of host economy to MNC composition-
determination is also introduced using an endogenous preference parameter in the
foreign experts’objective function, consistent with the international product market
dimension described by Felipe et al. (2012).
1.3.1 Domestic Sectors in Host Economy
Households
It is assumed that there is a continuum of dynastic representative households in the
economy, growing at an exogenous rate n > 0. Given initial number of members,
L0 in each household at time t = 0, the size of the representative family at time
t is Lt = exp(nt)L0. Each individual member within a household is assumed to
be infinitely lived, and possesses identical ability level, a, though different abilities
are assumed at the household level, as in Agénor and Alpaslan (2014). Ability
follows a Pareto distribution, indexed by a ∈ [am,∞), with probability density
function f(a) = χaχm/a1+χ and cumulative distribution function F (a) = 1−(am/a)χ.
χ is the Pareto index, where the larger the value, the smaller the proportion of
people with high cognitive ability. The mean ability of the population is given by
χam/(χ − 1), and χ > 2 and am > 1. A household with ability a and size L0
9Uncertainty of such nature may broadly be known as some sort of information cost, arisingfrom asymmetry in either demand or supply factors. An example of such cost is examined inHortsmann and Markusen (1996), though this chapter specifically attempts to link this choice ofMNCs to the ability distribution of workers in the host economy.
29
maximises intertemporal utility by solving the optimisation problem of
maxUat =
∫ ∞t
exp[−(ρ− n)(s− t)]L0u(cat )ds, (1.1)
where ρ > 0 is the subjective discount rate, u(cat ) is the utility function of individual
member of a household (depends on each individual member’s consumption, cat ),
given by the constant relative risk aversion form of
u(cat ) =1− (cat )
1/σ
1− 1σ
, (1.2)
subject to household budget constraint of
W at = rtW
at + (1− τ)Yt − Ltcat , (1.3)
where 1σ≥ 0 and σ denotes the constant elasticity of intertemporal substitution, rt
the riskfree market interest rate, Yt the economy’s output of final good, and τ ∈ (0, 1)
the tax rate on income. It is assumed that agents do not value leisure, and therefore
face no disutility from working or skills acquisition. Each representative household
is also assumed to make allocation of consumption equally among its members.
Household is not allowed to borrow. Standard transversality condition is assumed.
The solution to the family’s dynamic optimisation problem yields the standard
Euler equation,catcat
= σ(rt − ρ), (1.4)
which states that per capita consumption expenditure grows over time if and only
if the market interest rate exceeds the subjective discount rate. At the aggregate
level, the dynamics of household consumption, Ct = Ltcat , is then described by
CtCt
= σ(rt − ρ) + n. (1.5)
30
In terms of skills acquisition, individual members decide whether to acquire skills
or work straight away as unskilled workers, taking wages and interest rate as given.
Skill acquisition decisions are therefore made to maximise each member’s discounted
wage income, which is also equivalent to the representative household’s discounted
wage income. An individual with ability a ∈ [am,∞), fully observable by both
domestic firms and individuals, can either choose to enter the labour force at t as
an unskilled worker and earn from then on the wage wUt (which is independent of
the worker’s ability) or decide to spend first an exogenously given period of time
T to acquire skills. The education process occurs during the period of (t, t + T ),
and a direct cost of tct is incurred. Post-acquisition of skills, individual enters the
labour force at t + T as a skilled worker and earns a wage of aξwSt , where ξ > 0 is
a productivity parameter measuring the strength of ability’s effect on wages. This
would ensure that skilled workers with higher ability levels earn higher wages.10
Based on a generalised specification of Dinopoulos and Segerstrom (1999), an
individual with ability a ∈ [am,∞) would opt to become a skilled worker if and only
if ∫ ∞t+T
exp[−ρ(s− t)]aξwSs ds− tct ≥∫ ∞t
exp[−ρ(s− t)]wUs ds, (1.6)
where tct =∫∞t+T
exp[−ρ(s − t)]ΓaξwSs ds is the discounted value of the skills acqui-
sition cost that is assumed to be proportional to the skilled wages at Γ ∈ (0, 1).
The inequality (1.6) shows that the discounted value of the lifetime income of a
skilled worker, after accounting for skills acquisition cost during the period (t, T ),
must be higher or at least equal to the opportunity cost of education (discounted
wage income working as an unskilled worker). Hence, there exists a threshold level
of ability at such that (1.6) holds as an equality, expressed as
at = [exp(ρT ).(wUt /(1− Γ)wSt )]1/ξ. (1.7)
10Large ξ indicates strong effect, which implies that individuals with higher innate abilities wouldface lower cost in acquiring skills. This therefore also indicates the effi ciency of skills acquisition.
31
At any time t, (1− at)Lt individuals either work as skilled workers or are under-
going skills acquisition at any time t. If skills acquisition is assumed to take place
instantaneously, as in Eicher and García-Peñalosa (2001)11, equation (1.7) can be
simplified to
at = [wUt /(1− Γ)wSt ]1/ξ. (1.8)
Given Pareto distribution for abilities, and that productivity of unskilled workers
is assumed to equal unity, the effective supply of unskilled labour, LU,t at time t
equals
LU,t = Lt
∫ at
am
f(a)da = Ltaχm
[−a−χ
]atam
= Lt [1− (am/at)χ] . (1.9)
Given (1.9), the raw supply of skilled labour at time t is calculated as
Lt
∫ ∞at
f(a)da = (am/at)χLt,
though the average productivity of workers with ability a ∈ [at,∞) who have ac-
quired skills need to be accounted for. This gives the effective supply of skilled
labour at time t, LS,t, as
LS,t = Lt
∫ ∞at
af(a)da = χaχm
[a1−χ
1− χ
]∞at
Lt =χaχmχ− 1
(at)1−χLt.
Equivalently, in relative terms, the shares of unskilled and (effective) skilled
labour supply are given by
θU,t =LU,tLt
= [1− (am/at)χ] , and θS,t =
LS,tLt
=χaχmχ− 1
(at)1−χ. (1.10)
11Given the infinite horizon nature of the model, we follow Eicher and García-Peñalosa (2001)and Agénor and Dinh (2013) in imposing the assumption of T = 0. Knowing that individuals liveforever in the model, any training period specified within (0, T ) is small with respect to infinityand therefore can be treated as taking place instantaneously.
32
Imitation
In most existing contributions on imitation, innovation, and growth in the tradition
of Rustichini and Schmilz (1991), the imitation sector serves as a significant source of
growth, and the role of the imitation sector in driving growth has been documented
to be especially significant for the relatively ‘backward’economies playing catch-up
to developed peers. As agents in the economy learn from imitation, they would
develop the capacities to creatively imitate and subsequently, progress to engage in
indigenous innovation. This transition from imitation to innovation is known as the
stepping stone effect by Perez-Sebastian (2007) and Glass (2010).
Firms in the imitation sector produce imitative goods in the form of blueprints
that are purchased by firms producing basic intermediate inputs in the intermedi-
ate goods sector. Firms specialised in imitation employ only unskilled labour, in
quantity LU,I,t. There is no aggregate uncertainty in the research technology of
imitative blueprints production, though the production flow, M It at any time t is
determined by a productivity factor that depends on the economy-wide stock of
imitative blueprints at time t, M It , as well as an externality term associated with
the size of Vertical MNCs in innovation, nFV,tMRt . This productivity factor, ΦI
t is
expressed as:
ΦIt = (nFH,t)
ψI1(M It + ψI2nFV,tM
Rt ), (1.11)
where ψI1 ≥ 0 and ψI2 ∈ R, which feeds into the aggregate production technology of
imitative blueprints of
M It = ΦI
t (LU,I,tLt
), (1.12)
where it is assumed that, in consistent with the ‘dilution effect’discussed by Dinopou-
los and Segerstrom (1999), it is the ratio of unskilled workers to total population
that affects imitation activities.
The productivity component of imitative goods depends on: (i) a standard ini-
tial stock of blueprints (M It ), as in Jones’s (2005) ‘standing-on-shoulders’effect; (ii)
33
size of the presence of Horizontal MNCs, which given our definition of foreign firms,
refers to the total number of foreign experts that bring ‘know-how’ to imitation
production (expressed in proportion of total foreign firms, nFH,t); and (iii) an ex-
ternality term associated with the size of Vertical MNCs in the innovation sector.
As discussed earlier and implied in studies such as Markusen and Maskus (2002),
on aggregate, Horizontal FDIs are most likely to be imitation-enhancing (ψI1 > 0).
The externality term, ψI2nFV,tMRt , indicates a spillover channel from the innovation
sector. Consistent with the industrial transformation thesis, as the size of the in-
novation sector grows and more foreign subsidiaries opt to switch to operating as
Vertical MNCs, we would expect the sign of ψI2 to be negative. Nonetheless, given
that positive empirical evidence is also reported in some studies with regards to
leading foreign innovators’impacts on domestic firms’productivity, there is a pos-
sibility of a mildly positive ψI2 too12. As such, the parameter, ψI2, as well as the
stepping stone parameter, ψR2 (introduced in the innovation sector), are examined
across different values using sensitivity analysis.
The optimisation problem of firms in the imitation sector is to select the amount
of unskilled labour to employ so as to maximise profits of
ΠIt = RI
t MIt − (1 + ΛI)wUt LU,I,t,
subject to (1.12), taking the imitative blueprint price (RIt ) and unskilled wage rate
(wUt ) as given. The parameter ΛI is introduced as a proportionate cost factor in the
imitation sector that captures the impact of labour market distortions (for instance,
additional hiring and firing costs arising from non-competitive labour market prac-
tices). The same additional cost is faced by all firms in the sector. The interior
solution for unskilled labour employment in imitation (LU,I,t > 0) is given by the
12Empirical studies specifically in the area of international production networks, such as Athuko-rala (2005) and Kam (2013), find the presence of a positive productivity spillover from leadingforeign innovators to the productivity of domestic imitators, notably component part suppliers inthe host economy. On the contrary, studies such as Haddad and Harrison (1993) and Djankov andHoekman (2000) document negative effects of foreign firms on domestic firms’productivity.
34
following first-order condition:
wUt =1
1 + ΛI
RItΦ
It
Lt. (1.13)
Innovation
Firms in the innovation sector produce innovative blueprints using only skilled labour
(LS,R,t). In comparison to the employment specification made for imitation, inno-
vation sector is therefore skill-intensive. There is no aggregate uncertainty in in-
novation, though the research production flow at any time t is determined by a
productivity factor, ΦRt , defined as
ΦRt = (nFV,t)
ψR1 (MRt + ψR2 M
It ), (1.14)
where ψR1 ≥ 0 and ψR2 ≥ 0,13 which feeds into the aggregate production technology
of innovative blueprints:
MRt = ΦR
t (LS,R,tLt
). (1.15)
As in the imitation sector, the production technology of innovative goods cap-
tures the key knowledge spillover properties that are often documented in industrial
development literature. Following Agénor and Dinh (2013), the research process
of innovation depends on both the stock of innovative and imitative blueprints,
consistent with the stepping stone effect of imitation introduced by Glass (2010).14
The productivity gains associated with stepping stone effect of imitative goods may
be equal, stronger (ψR2 > 1), or weaker (ψR2 < 1) than that of innovative goods.
13There is a slight difference in the specification of the cross-sector externality term in theproductivity factor here compared to (1.11), where the activities of the Horizontal MNCs do notaffect the firms’productivity, unlike the role of Vertical MNCs in the imitation sector. As seenlater, this is largely due to Horizontal MNCs bringing in only standardisation know-how, which isuseless in the context of creating indigenous innovation.
14While the stepping stone effect is a key spillover mechanism for industrialising countries play-ing catch-up, it is worth noting that innovation may not necessarily depend on imitation at all,especially for developed economies at the frontier.
35
Consistent with studies such as Markusen (1998) and Braconier et al. (2005), Ver-
tical MNCs, nFV,t, are specified as the relatively skill-intensive type that engage in
leading-edge innovation and therefore beneficiary to innovation in the host economy.
As such, nFV,t refers to the total number of foreign experts that bring sophisticated
‘know-how’to innovation production in the host economy.15 Likewise, to eliminate
scale effects, innovation production is specified as depending on the ratio of employed
skilled workers to total population.
The optimisation problem of firms in the innovation sector is to select the amount
of skilled labour to employ so as to maximise profits of
ΠRt = QR
t MRt − (1 + ΛR)wSt LS,R,t,
subject to (1.15), taking the patent price (QRt ) and skilled wage rate (w
St ) as given.
The wage in the innovation sector is affected proportionally again by a cost para-
meter ΛR. When ΛR > ΛI , the labour market for the innovation sector is more
distorted than the labour market for imitation sector, meaning that it is compara-
tively more expensive to hire skilled workers in innovation than unskilled workers in
imitation within the economy. This specification of ΛR > ΛI is consistent with the
general finding documented in Haaland and Wooton (2001).16
For an interior solution for skilled labour employment in innovation to exist
(LS,R,t > 0), the first-order condition is given by
wSt =1
1 + ΛR
QRt ΦR
t
Lt, (1.16)
15A more accurate modelling approach would be to scale the variable by number of domesticexperts, but such top domestic experts is usually very small or non-existent in a developing econ-omy. Instead, we introduce a foreign-to-domestic innovation expertise ratio, Ψt = nFV,t/ θS,R,t,where θS,R,t = LS,R,t/Lt, later as a proxy measure to compare across policy outcomes.
16In their studies, Haaland and Wooton (2001) examine the effects of labour market rigidities,especially redundancy payments, on MNCs’choice of investment destination. They document that,those sectors with relatively less certainty in production, such as the innovation sector, tend tohave higher degree of labour market rigidities.
36
which, using (1.14), can be rewritten as
wSt = (1
1 + ΛR)(QRt
Lt)(nFV,t)
ψR1 [1 + ψR2 (mIt
mRt
)]MRt . (1.17)
Final Output
The final output sector is a perfectly competitive market consisting of firms pro-
ducing final good. There is a continuum of identical domestic firms involving in the
production of a homogenous final good, indexed by i ∈ (0, 1). Production by indi-
vidual domestic firm i uses firm-specific private capital, Kit , skilled labour, LS,Y,i,t,
unskilled labour, LU,Y,i,t, and composite intermediate input, X it .
The production function of individual domestic firm i takes the form of a stan-
dard Cobb-Douglas specification:
Y it = (LS,Y,i,t)
βS(LU,Y,i,t)βU (X i
t)γ(Ki
t)α[
Kt
(Lt)ι]%, (1.18)
where % > 0, ι > 0, α ∈ (0, 1), βS ∈ (0, 1), βU ∈ (0, 1), γ ∈ (0, 1), and α + (βS +
βU)+γ = 1 to reflect constant returns to scale in firm-specific inputs LS,Y,i,t, LU,Y,i,t,
X it , and K
it .
The economy-wide aggregate stock of private capital, Kt =∫ 1
0Kitdi, asserts a
conventional Arrow-Romer type of externality on each individual firm i’s production,
at a magnitude of %. However, it is subject to a congestion effect of ι due to total
population size, Lt.
The composite intermediate input exhibits constant returns to scale with respect
to basic and sophisticated intermediate inputs. The composite intermediate inputs
required for individual firm’s production, X it , in (1.18) is written as
X it = [
∫ MIt
0
(xIs,t)ηds]ν/η · [
∫ MRt
0
(xRs,t)ηds](1−ν)/η, (1.19)
where xIs,t, s ∈ (0,M It ) refers to basic intermediate inputs, xRs,t, s ∈ (0,MR
t ) sophis-
37
ticated intermediate inputs, ν ∈ (0, 1) the share of basic intermediates in composite
intermediates, η ∈ (0, 1) and 1/(1 − η) > 1 the price elasticity of demand for each
intermediate input (in absolute terms).17
Faced with competitive markets for private inputs, the optimisation problem of
firms in the final output sector is to maximise profits, ΠY,it , with respect to private
capital, skilled labour, unskilled labour, and the quantities of all intermediate inputs,
taking factor prices and aggregate level of M It , M
Rt , LS,Y,t, LU,Y,i,t, and Lt at any
time t as given:
maxKit ,L
S,Y,it ,LU,Y,it ,xI,is,t,x
Us,t
Πit = P Y
t Yit − (1 + ΛY )wSt LS,Y,i,t − (1 + ΛY )wUt LU,Y,i,t
−(rt + δ)Kit −∫ MI
t
0
P I,st xIs,tds−
∫ MRt
0
PR,st xRs,tds,
where P Yt is the price of final good normalised to unity, P I,s
t (PR,st ) is the price of
basic (sophisticated) intermediate good s, wSt (wUt ) the skilled (unskilled) wage rate,
rt the net rental rate of private capital, and δ ∈ (0, 1) the rate of depreciation for
private capital. A third labour market distortion parameter ΛY is introduced to
capture the additional cost faced by firms induced by sector-specific labour market
rigidity, and is assumed to affect in the same manner the use of both skilled and
unskilled labour in production of final good.
Given that all firms in final output production are identical and demand the
same quantity of each inputs, profit maximisation in a symmetric equilibrium yields
rt = αYtKt
− δ, (1.20)
wSt =βS
1 + ΛY
YtLS,Y,t
, wUt =βU
1 + ΛY
YtLU,Y,t
, (1.21)
17Similar to final output elasticities α, βS , βU , and γ, the coeffi cient ν is fixed initially at aconstant value, though it is endogenised in the sensitivity analysis section later using a generalisedlogistic curve.
38
xκs,t = (γνκZκtPκ,st
)1/(1−η), s = 1, ...Mκt , (1.22)
Zκt = Yt/
∫ Mκt
0
(xκs,t)ηds, (1.23)
where κ = I, R, νI = ν, and νR = 1− ν. (1.20) is the standard first-order condition
that determines the real interest rate, (1.21) states the two first-order conditions for
wages that arisen from the labour demand of firms in the sector, while (1.22) gives
the demand function for the two types of intermediate goods.
Given that both the technology and demand for all specific intermediate type
(either basic or sophisticated) are the same, the equilibrium for both intermediate
types are symmetric too. In a symmetric equilibrium,∫MI
t
0(xIs,t)
ηds = M It (xIt )
η and∫MRt
0(xRs,t)
ηds = MRt (xRt )η. The composite intermediate inputs can then be written
as
Xt = [(M It )1/ηxIt ]
ν [(MRt )1/ηxRt ]1−ν . (1.24)
To derive an expression for the aggregate final output of the economy, the number
of firms engaged in the production of final good is normalised to unity, Yt =∫ 1
0Y it di,
which implies that the aggregate skilled and unskilled labour used in the final output
sector are given by LS,Y,t =∫ 1
0LS,Y,i,tdi and LU,Y,t =
∫ 1
0LU,Y,i,tdi respectively. Using
(1.18), the aggregate final output Yt can be written as
Yt = (LS,Y,t)βS(LU,Y,t)
βU (Xt)γ(Kt)
α[Kt
(Lt)ι]%. (1.25)
Finally, the law of motion for the private capital is given by the standard form
of:
Kt = It − δKt, (1.26)
where It is the aggregate private investment by the normalised number of firms.
39
Intermediate Goods
The intermediate goods sector is monopolistically competitive, and consists of two
sub-sectors of: (i) intermediate input producers producing basic inputs, based on
blueprints produced by the imitation sector; (ii) intermediate input producers pro-
ducing sophisticated inputs, based on blueprints produced by the innovation sector.
Consider first producers of basic intermediate inputs, xI,st , s = 1, ...M It . Each firm
specialises in producing one unit of horizontally-differentiated basic intermediate
input. To obtain the rights to produce, each producer pays an imitative blueprint
price, RIt , in each period to the firm that produces the relevant blueprint in the
imitation sector.
Faced with a monopolistically competitive market structure, each basic inter-
mediate input firm maximises profits by setting price P I,st for good s, given the
perceived demand function, (1.22) for its good. In a symmetric equilibrium, and
using also (1.23), profits are then expressed as
ΠIt = (P I
t − 1)[γνYt/PIt M
It (xIt )
η]1/(1−η).
The solution yields an optimal price of
P I,st =
1
η, ∀s = 1, ...M I
t . (1.27)
The associated quantity demanded at the equilibrium price, P It = P I,s
t is
xIs,t = (γηνZIt )1/(1−η),∀s,
which is equal to
xIt = γην(YtM I
t
), (1.28)
40
in a symmetric equilibrium.
The maximum profit in a current period t is then given by
ΠIt = (1− η)γν(
YtM I
t
). (1.29)
Standard arbitrage implies that the blueprint price must be equal to the present
discounted stream of profits. For simplicity, we follow Agénor and Canuto (2015b)
and assume that all the profits of an imitative blueprint, excluding capital gain, go
into the imitative blueprint price, RIt set in equilibrium. This yields
RIt = ΠI
t . (1.30)
The sub-sector for the production of sophisticated intermediate inputs assumes a
similar market structure. Before producing its specialised sophisticated inputs, each
firmmust purchase a patented blueprint from the innovation sector. Unlike imitative
blueprints, patented blueprints are infinitely-lived. Each sophisticated intermediate
input firm sets its price to maximise profits, given the perceived demand function,
(1.22) for its good. In a symmetric equilibrium, and using also (1.23), profits are
then expressed as
ΠRt = (PR
t − 1)[γ(1− ν)Yt/PRt M
Rt (xRt )η]1/(1−η).
The solution yields an optimal price of
PR,st =
1
η, ∀s = 1, ...MR
t , (1.31)
with an associated quantity demanded at the equilibrium price, PRt = PR,s
t of
xRt = γη(1− ν)(YtMR
t
). (1.32)
41
The maximum profit is then given by
ΠRt = (1− η)γ(1− ν)(
YtMR
t
). (1.33)
To derive the equilibrium price of a patent for sophisticated input, QRt , recall that
standard no-arbitrage condition requires that the rate of return on private capital
must equal to the rate of return on the exclusive holding of an innovative blueprint
for sophisticated intermediate inputs. The latter is equal to the sum of the profit
rate and the rate of capital gain from a change in the patent price over time. This
gives
rt =ΠRt
QRt
+QRt
QRt
,
which can be rearranged to yield
QRt = rtQ
Rt − ΠR
t . (1.34)
1.3.2 Foreign Sector
In each period of time, for any host economy of interest, investment flows charac-
terised by total number of firms in three different modes of foreign MNC subsidiaries
are determined for any individual host economy. The three types of FDI mode are
Non-mandated, Horizontal, and Vertical FDI.18 A foreign firm consists of an expert
or professional that brings specific know-how into the host economy. Specifically,
each foreign firm is one individual and the fixed know-how brought into the host
economy is essentially specific processes that are only available in the foreign source
country. For example, this means a Vertical MNC would come in the form of an
innovation expert bringing sophisticated know-how, while a Horizontal MNC would
18In addition to the literature reviewed, the classification of FDI is also supported by an em-pirical estimation exercise implemented. See Section 1.4 and 1.7.1 for further details.
42
be in the form of an imitation expert bringing standardisation know-how. By de-
finition, the FDI composition of a particular host economy in any period t would
therefore equal the composition of foreign experts in the economy. For simplicity, we
assume no cross-border trade in the model. Also, the role of Non-mandated MNCs
(NFP,t) in production of a middle-income host economy with an innovation sector is
deemed insignificant and therefore not examined, though they are still modelled as
a default base entry mode of foreign MNCs.
Stylised Framework to explain ‘Internalisation advantage’:
To characterise the mechanics of foreign subsidiaries’deployment, we use a three-
staged, nested Dixit-Stiglitz CES objective function framework adapted from Allan-
son and Montagna (2005) and Brambilla et al. (2009). In each period, it is assumed
that there is a mass of foreign subsidiaries, j = 1, ..., NF , entering the host economy,
with the salaries/profits of the experts/subsidiaries assumed, for simplicity, to be
paid by the planner of the foreign source economy.19
Specifically, in the first stage, the planner of the foreign source economy deter-
mines the allocation of aggregate salary expenditure for experts deployed across all
developing host economies. Based on a standard Cobb-Douglas value maximisation
specification, max uFt = z%H,tz1−%q,t , in each time period, the exogenously given aggre-
gate salary expenditure (IF ) is allocated between salary expenditure for experts in
our host economy of interest (zq) and for simplicity, other host economies collectively
(zH). This yields yFt = (1− %)IFt , where yF is the total salary expenditure allocated
for the specific host economy examined and (1 − %) the corresponding share. By
definition, yFt = wFNF,t, where wF is some exogenously given wage rate paid by
19A more conventional approach is to specify that the salaries/profits of foreign ex-perts/subsidiaries to be determined in the host economy. However, as applicable to most actualinstances in real life, experts of MNC subsidiaries deployed to developing economies for assignmentsdo receive their remuneration from the headquarters. In addition, unlike models treating FDI ascapital stock, our main emphasis is on heterogeneous FDI compositions and how such choice isaffected by skills distribution of a host economy. The usual returns motive is therefore abridgedand simplified as an exogenous salary expenditure paid by foreign planner to the entire pool offoreign experts.
43
the foreign headquarter and NF,t is the total number of foreign experts in the host
economy studied.
Having determined the allocation in the first stage, a stylised institutional ap-
proach is specified in the second stage. ‘Investment’in the host economy is assumed
to be in terms of the intermediate variety it is matched to. Collectively, the pool of
foreign experts deployed to the specific host economy forms a representative value
function over a composite of intermediate varieties, with a further layer of ‘shadow in-
vestment quality’ascribed to capture the preference of foreign experts to be matched
to workers of higher productivity, even among the same variety type that they are
matched to.20
Specifically, the value function is given by
UFt = (
∫ NF
j=0
[
∫ MIt
s=0
γ1,t(xIs,FH,t)
σF−1σF ds+
∫ MRt
s=0
γ2,t(xRs,FV,t)
σF−1σF ds]
θF−1θF dj)
θF
θF−1 ,
(1.35)
whereM It ,M
Rt denote the imitative and innovative varieties over Horizontal, x
Is,FH,t,
and Vertical investments, xRs,FV,t; σF and θF are elasticities of substitution within
and between intermediates, with σF > θF > 1 assumed as in Brambilla et al. (2009).
γ2,t and γ1,t represent foreign preferences for investment of Vertical and Horizontal
MNC respectively.21 ,22
Solving the optimisation problem yields a series of theoretical investment de-
mand functions and shadow investment prices for each variety s and productivity
difference-induced quality j.23
20By construction, the ‘quality difference’between investments in a host country for the foreignexperts in this model reflects solely the perceived difference in productivity among the domesticworkers.
21As shown later, foreign preferences are endogenous to the state of industrial development ofa host economy, providing a key feedback channel of the host economy’s industrial state to FDIvia the product market dimension. Nevertheless, it is taken as given by the pool of foreign expertswhen solving for the maximisation problem in every period.
22Since not all destinations of host economies have an innovation sector, we can set xRs,t = 0 inthe value function if we were to model a host economy without an innovation sector.
23The general expression of the theoretical demand functions, as well as associated shadowinvestment price indices, are provided in the Appendix.
44
FDI Compositions in Host Economy
In stage three, for a given number of foreign firms (NF,t) entering the host economy
of interest in each period t, each firm’s dynamic entry decision is modelled as a static
decision in opting for investment mode.24 Upon entry, foreign firms first assume a
Non-mandated MNC mode and to simplify matters, no subsequent exit is allowed.
Further, in each period t, a firm can opt to stay and operate as Non-mandated
MNC (incurring a basic ‘doing-business’cost of F0); incurring additional cost, F1
on top of F0 to upgrade into Horizontal FDI mode; or incur F0 + F2 to operate as
a Vertical MNC. All three costs, F0, F1, and F2 are expressed as a fraction of some
theoretical baseline price corresponding to the default Non-mandated investment, P0
(which is normalised to one). Further, F2 > F1 > F0 is assumed. Since each foreign
subsidiary is essentially a foreign expert, these mean foreign subsidiaries have the
option to ‘upgrade’and bring in an expert with more advanced know-how in every
period, by incurring additional operation cost to operate in the host economy.25
As stated, unlike domestic firms, each foreign expert coming in with know-how
perceives heterogeneity among productivity of domestic workers. This asymmetry
leads to a ‘productivity requirement’-induced information cost component, 1/$,
that is implicitly priced by foreign experts when deciding on the choice of oper-
ational mode. This productivity is a transformation of ability. For simplicity, a
one-to-one relationship is assumed, where $ = a/a, with a being value along the
ability distribution of the host economy and 1 < a <∞ some exogenously specified
constant value. 1/$ is therefore also characterised by a Pareto distribution. Due to
persistence, for those who have become skilled, it is assumed that a more able indi-
vidual pre-skills acquisition would remain more productive over another individual
24Heterogeneous foreign firms are assumed to behave in a homogenous manner within the sameFDI type.
25Consistent with the nature of most common ‘doing-business’costs surveyed, such as time toacquire permits and number of administrative procedures in transactions, these costs are treatedas deadweight losses in this model, instead of being fees collected by the government of the hosteconomy.
45
with lower ability pre-skills acquisition, resulting in a Melitz (2003) type of sorting
of foreign subsidiaries on 1/$.26
Specifically, for any intermediate variety s at time t, we can express an optimal
shadow price of investment (from the perspective of foreign experts) as a function
of $, that is,
Ps,t =
(σF
σF − 1
)($s,t) , (1.36)
priced at σF/(σF − 1) > 1 times of $s,t.27
This implies that, for any investment of variety s, the larger the ‘productivity
requirement’-induced information cost is (lower $s,t), the lower is the theoretical
investment price ascribed by the foreign experts.
Further, as both an additional novel feature and to ensure the solution space is
bounded from below (a ∈ [am,∞)), a second source of asymmetry between Vertical
and other MNCs is introduced. As seen later, it turns out that this technical fea-
ture actually allows us to provide a theoretical proposition that is consistent with
Braconier et al. (2005) and the ‘race-to-the-bottom’literatures in explaining the
empirical documentation of limited high value-added Vertical MNC activities in de-
veloping economies, despite most developing governments competing for their inward
presence. Specifically, when a foreign subsidiary is confronted with the decision to
upgrade to Vertical mode, the cost associated with the productivity requirement is
subject to a parameter φ, such that (1/$)ˆ(−φ) > 0, φ ≤ 0 is now priced by the
foreign experts to reflect the increasing diffi culties in telling apart and identifying
the best (highest productivity) among the brightest of skilled workers. To explain
intuitively, say for example, as a given value of a gets smaller (1/$ gets larger)
26By virtue of the persistence assumption, productivity therefore assumes the same continuousprobability distribution of ability. As such, given random matching, the perceived productivitydifferences by the foreign experts would naturally lead to a sorting of all the foreign experts alongthe same distribution in each period. A mechanism for updating expectation is therefore notnecessary in this instance.
27Given that the perceived quality difference among investment is driven by perceived hetero-geneity among productivity of domestic workers, this price is implicit in nature and reflects the‘value’placed by foreign experts on a specific intermediate variety s.
46
and smaller (note that if from the supply side, it means the actual quantity of
skilled labour in host economy gets larger), a negative value for parameter φ would
indicate increasing diffi culties in identifying and matching to the most productive
skilled workers. In other words, as the pool of skilled workers gets larger in the host
economy, it gets harder to tell apart the brightest from among the skilled workers.
The two dichotomous features discussed in the foreign sector characterise the
stylised ‘internalisation’framework that determines FDI compositions in this model.
Equation (1.36), together with theoretical investment demand functions across dif-
ferent varieties, allow us to express individual value function for a typical foreign
expert j opting for either Non-mandated (πFP ), Horizontal (πFH), or Vertical (πFV )
operational mode. Imposing zero profit conditions for foreign experts across the
three types (πFP ($FP ) = 0, πFH($FH) = πFP ($FH), πFH($φFV ) = πFV ($φ
FV )),
and given that on aggregate, Pj = Ps = LI is assumed in symmetric equilibrium28,
the three minimum threshold values for MNCs’internalisation decision in any period
t can be expressed as
$FP,t =aFP,ta
=
[F0(
(σF − 1)σF−1/(σF )σF−1(yFt )−1)P θF−1F,t
]1/(1−σF )
, (1.37)
$FH,t =aFH,ta
=
[F1(
(σF − 1)σF−1/(σF )σF−1(yFt )−1)P θF−1F,t [γσ
F
1,t (LI)σF−θF − 1]
]1/(1−σF )
,
(1.38)
$FV,t =aFV,ta
=
[(F2 − F1)(
(σF − 1)σF−1/(σF )σF−1(yFt )−1)P θF−1F,t (LI)σF−θ
F[γσ
F
2,t − γσF
1,t ]
]1/[φ(1−σF )]
,
(1.39)
where F0, F1, F2 are the ‘doing-business’costs (in proportion of P0 = 1); σF , θF ,
yFt , φ, γ1,t, γ2,t are as defined earlier; and PF,t is a theoretical aggregate shadow
28This means the shadow price indices for the implicit ‘investment price’of between- (Ps) andwithin-variety (Pj) are equalised, and assumed to be taken as given by the individual experts.As explained further in the Appendix, for ease of modelling, we proxy this by a time-invariantstructural parameter, the Lerner Index, LI, which generalises market competitiveness– hence areflection of the implicit value of investment– in the host economy.
47
investment price index that is substituted out later.
To calculate the shares of foreign firms by FDI type, recall that the sorting of
foreign firms follows that of 1/$. We know that the cumulative distribution function
of a typical Pareto distribution z, takes the form of F (z) = 1 − (zmin/z)χ for some
minimum of z, zmin. Let F (1/$) = F (a/a). Further, by assuming that there is
no exit option for MNCs, we can set aFP = a/amin∀t, where a/amin denotes some
minimum threshold value of entry by foreign firms (a large value along the ability
distribution of host economy). At any time t, the proportion of the three types of
foreign firms can then be computed as
nFP,t =NFP,t
NF,t
= [F (1/$FH,t)− F (1/$FP,t)] (1.40)
= [1− (aFH,taFP
)χ] ,
nFH,t =NFH,t
NF,t
= [F (1/$FV,t)− F (1/$FH,t)] (1.41)
= [(aFH,taFP
)χ − (aFV,taFP
)χ],
nFV,t =NFV,t
NF,t
= [1− F (1/$FV,t)] (1.42)
= (aFV,taFP
)χ,
where aFP , aFH , aFV give the host economy-specific threshold values of entry (for
Non-mandated, Horizontal, and Vertical MNCs respectively). While nFH,t in (1.41)
is determined by both aFH,t and aFV,t, given fixed aFP , (1.42) shows that the lower
the value of aFV (therefore the stricter the entry criteria for Vertical FDI), the
smaller share of Vertical MNCs in the host economy. Also, (1.40) shows that the
lower the value of aFH (therefore stricter criteria for Horizontal FDI), the larger the
48
share of Non-mandated MNCs.29
Some straightforward algebraic manipulations using (1.37)-(1.39) allow us to
substitute out yFt and PF,t, and establish two threshold conditions of
aFH,t =
[F0
F1
((LI)σF−θF (γ1,t)
σF − 1)
]−1/(1−σF )
aFP , (1.43)
and
aFV,t =
[F2 − F1
F0
1
(LI)σF−θF
[γσF
2,t − γσF
1,t ]
]1/[φ(1−σF )]
a1/φFP a
(φ−1)/φ, (1.44)
respectively.
In addition, a feedback channel from the state of industrial development in a
host economy to FDI composition is introduced. Given that FDI inflows into the
Southeast Asian regions are found empirically to follow a Weibull distribution by
Gander et al. (2009), we simplify by modelling the two foreign preference parameters
γ1and γ2 using a Weibull distribution, governed by a hazard function of
γ1 = [1− h(γ2;ωk, ωλ)]γ2 (1.45)
= [1− (ωkωλ
(γ2
ωλ)ωk−1)]γ2,
where h(γ2;ωk, ωλ) denotes the hazard rate of γ230, and ωk and ωλ are the shape
and scale parameter respectively. As γ1 is given by the expected value of E(γ2), this
allows us to endogenise foreign preferences to become QF , a demand-side feedback
channel depending on the state of industrial development in a host economy, and
29Indirectly, these imply that the distribution of foreign experts in the host economy is influencedby a Pareto distribution. In the absence of an actual empirical distribution, and given that theelement of ability is unobserved in terms of real world data, this is a reasonable assumption.
30This means we assume that foreign investment preference in Horizontal MNC would reduceover time relative to the investment preference in the mode of Vertical MNC. While this assumptionseems arbitrary, it provides a reasonable simplification that allows for feedback of industrial statein the host economy to FDI composition through only a single foreign preference channel.
49
rewrite (1.43) and (1.44) as
aFH,t =
[F0
F1
((LI)σF−θF (QF
t −Θ1(QFt )ωk)σ
F − 1)
]−1/(1−σF )
aFP , (1.46)
and
aFV,t =
[F2 − F1
F0
1
(LI)σF−θF
[(QFt )σF − (QF
t −Θ1(QFt )ωk)
σF]
]1/[φ(1−σF )]
a1/φFP a
(φ−1)/φ,
(1.47)
respectively, where Θ1 = (ωk/ωλ)(1/ωλ)ωk−1. For tractability, we assume that the
foreign MNCs set QF = mIt in each period.
31
Finally, using (1.40)-(1.42), (1.46), and (1.47), we can derive the expressions for
nFH,t and nFV,t as
nFH,t = a−χFP (aFH,tχ − nFV,t.aχFP ) (1.48)
=
(aFH,taFP
)χ− nFV,t
=
[F0
F1
((LI)σF−θF (QF
t −Θ1(QFt )ωk)σ
F − 1)
]−χ/(1−σF )
− nFV,t,
and
nFV,t =(a
1/φFP a
(φ−1)/φ)χ [F2 − F1
F0
1
(LI)σF−θF
[(QFt )σF − (QF
t −Θ1(QFt )ωk)
σF]
]χ/[φ(1−σF )]
,
(1.49)
respectively, where Θ1 = (ωk/ωλ)(1/ωλ)ωk−1 and QF
t = wmmIt (wm is a multiplica-
tive constant).
31The use of mIt in the feedback channel as a proxy that reflects the state of industrial develop-
ment in a developing host economy is consistent with studies such as Yusuf and Nabeshima (2009).It also provides a more general feature given that there are developing host economies that haveonly imitation production. Note that the industrial composition ratio, mt = mI
t /(mRt +mI
t ) can beused in an alternative specification, though it comes with a lot more complications. Specifying QF
as being driven by the dynamics of the industrial ratio– hence a complicated expression with thedynamics from both state variables, mI
t and mRt – would make the subsequently derived expres-
sions for nFH,t and nFV,t analytically intractable. The same tractability consideration explainsthe rationale for using the stationary variable of mI
t instead of MIt .
50
As a result of the perceived heterogeneity of productivity among workers, and the
assumed one-to-one transformation of productivity from ability (due to persistence),
the determination of nFH,t and nFV,t in any period t is therefore driven by the sorting
process along the same ability distribution, and depends on threshold ability values,
aFH,t and aFV,t. Naturally, these lead to some degree of direct tradeoffbetween nFH,t
and nFV,t, as can be seen in (1.48), though it is also possible that an economy can
gain in both nFH,t and nFV,t.
1.3.3 Government and Market-clearing Conditions
Government
All public policies in this chapter are assumed to be financed by reallocating spending
within the budget, so that the tax rate remains the same and the overall balance
remains. As such, we can assume a simplified government sector. A balanced budget
is maintained, and the government cannot issue bonds to borrow. At each time t,
the government taxes on final output at the rate τ to finance its expenditure Gt, as
in
Gt = τYt. (1.50)
Market Equilibrium Conditions
Final Good Market Equilibrium For the final good market, as noted earlier,
under symmetry,∫MI
t
0xIs,tds = M I
t xIt and
∫MRt
0xRs,tds = MR
t xRt . The final good
market-clearing condition is given by
Yt = Ltcat +M I
t xIt +MR
t xRt + It +Gt. (1.51)
Using (1.28), (1.32), and (1.50), equation (1.51) is rewritten as
It = Ltcat − (1− γη − τ)Yt, (1.52)
51
which represents the private investment level in the economy at any time t.
Labour Markets Equilibrium In order for the market for skilled labour to clear,
note that skilled workers are employed in either the production sector for final good
or innovative blueprints. Market equilibrium is
LS,Y,t + LS,R,t = LS,t,
which equals to
θS,Y,t + θS,R,t = θS,t, (1.53)
when expressed as a proportion of total population (divided by Lt).
To clear the labour market for unskilled workers, recall that unskilled workers
are employed in either the production sector of final good or imitative blueprints.
Market equilibrium is
LU,Y,t + LU,I,t = LU,t,
equivalent to the ratio terms of
θU,Y,t + θU,I,t = θU,t, (1.54)
when expressed as a proportion of total population (divided by Lt).
For the foreign sector, in any given period t, the shares of foreign experts or
subsidiaries in Non-mandated, Horizontal, and Vertical mode in the host economy
should sum up to one, with nFP,t derived residually. This means
nFP,t = 1− nFH,t − nFV,t , nFP,t ≥ 0. (1.55)
52
1.3.4 Dynamic System and Steady State
Dynamic System
Before presenting the overall dynamic system of the economy, to generate endoge-
nous growth, we impose the following knife-edge conditions:
Assumptions: βS + βU − %ι = 0, (γ/η) + α + % = 1.
Specifically, first, define mIt = M I
t /Kt and mRt = MR
t /Kt. Using (1.28) and
(1.32), (1.24) is written as:
Xt = (γηνν(1− ν)1−ν)(mIt )ν(1−η)/η(mR
t )(1−ν)(1−η)/η(YtKt
)(Kt)1/η.
Substituting the expression into (1.25), and let θS,Yt = LS,Yt /Lt and θU,Yt =
LU,Yt /Lt, give
Yt = (θS,Yt )βS
(θU,Yt )βU
LβS+βU−%ιt (1.56)
×
(γηνν(1− ν)1−ν)(mIt )ν(1−η)/η(mR
t )(1−ν)(1−η)/η(YtKt
)
γ(Kt)
(γ/η)+α+%.
(Lt)0 = 1 if and only if βS + βU − %ι = 0. The level of output, Yt, is linear to
the private capital stock, Kt, if and only if (γ/η) + α + % = 1.
The dynamic system of the economy is characterised by a differential algebraic
system consisting of four first-order differential equations and seven static equations.
The four differential equations are
mRt
mRt
= (nFV,t)ψR1 [1 + ψR2 (
mIt
mRt
)](θS,t − θS,Y,t)− (1− γη − τ)(YtKt
) + zCt + δ, (1.57)
mIt
mIt
= (nFH,t)ψI1 [1 +ψI2nFV,t(
mRt
mIt
)](θU,t− θU,Y,t)− (1− γη− τ)(YtKt
) + zCt + δ, (1.58)
zCtzCt
= n+ [σα− (1− γη − τ)](YtKt
) + zCt − σ(ρ+ δ) + δ, (1.59)
53
QRt
QRt
= [α(YtKt
)− δ]− (1− η)γ(1− ν)(YtKt
)(1
QRt
)(1
mRt
), (1.60)
of which mIt and m
Rt are backward-looking state variables, while z
Ct and QR
t are
forward-looking jump variables.
The seven static equations are
YtKt
=Θ2
[(θS,Y,t)βS(θU,Y,t)β
U]−1/(1−γ)
(mI
t )ν(1−η)/η(mR
t )(1−ν)(1−η)/ηγ/(1−γ)
, (1.61)
θS,Y,t =βS(1 + ΛR)
(1 + ΛY )(YtKt
)[QRt (mR
t )]−1(nFV,t)−ψR1 [1 + ψR2 (
mIt
mRt
)]−1, (1.62)
θU,Y,t =βU(1 + ΛI)
(1 + ΛY )(1− η)νγ(nFH,t)
−ψI1 [1 + ψI2nFV,t(mRt
mIt
)]−1, (1.63)
θU,t = 1− aχm[βU
βS(1− Γ)
θS,Y,tθU,Y,t
]−χ/ξ, (1.64)
θS,t =χaχmχ− 1
[βU
βS(1− Γ)
θS,Y,tθU,Y,t
](1−χ)/ξ, (1.65)
nFH,t =
[F0
F1
((LI)σ−θ(wmmIt −Θ1(wmm
It )ωk)σ
F − 1)
]−χ/(1−σF )
− nFV,t, (1.66)
nFV,t =(a
1/φFP a
(φ−1)/φ)χ
(1.67)
×[F2 − F1
F0
1
(LI)σF−θF
[(wmmIt )σF − (wmmI
t −Θ1(wmmIt )ωk)
σF]
] χ
[φ(1−σF )]
,
where Θ1 = (ωk/ωλ)(1/ωλ)ωk−1 and Θ2 = (γηνν(1− ν)1−ν)γ/(1−γ).
Finally, to calculate the final output growth rate of the economy at any time t
during the transition, first log-differentiates (1.61) with respect to time. Then, with
further substitution of the log-differentiated version of equations (1.62) and (1.63),
54
and rearranging of terms, we obtain
YtYt
=KPt
KPt
+
[γν(1− η)
(1− γ)η(1− βS
1− γ −βS(1 + ψR1 )(−χωkσF )
(1− γ)[φ(1− σF )])−1
]mIt
mIt
(1.68)
+(βU
1− γ )(1− βS
1− γ )−1 θU,Yt
θU,Yt
+[γ(1− ν)(1− η)
(1− γ)η− βS
1− γ ](1− βS
1− γ )−1mRt
mRt
−(βS
1− γ )(1− βS
1− γ )−1 QRt
QRt
−[(βSψR21− γ )(1− βS
1− γ )−1(1 + ψR2mIt
mRt
)−1][mIt
mRt
(mIt
mIt
− mRt
mRt
)]
Steady State
The steady-state equilibrium is defined as an equilibrium path where the growth rate
of the aggregate representative households’consumption (nt + (cat /cat )), the growth
rate of the private capital stock (Kt/Kt), the growth rate of imitative blueprints
(M It /Mt), and the growth rate of innovative blueprints (MR
t /MRt ) are all equal,
whereas the imitative blueprint price (RIt ), the patent price (Q
Rt ), rate of return on
private capital (rt), real prices (PI,st , PR
t ), and the aggregate shadow investment price
index (PF,t) are constant. From the five static conditions in domestic sectors, (1.61)-
(1.65), and the two equations determining number of Horizontal MNCs (foreign
experts with standardisation know-how) and Vertical MNCs (foreign experts with
sophisticated know-how), (1.48) and (1.49), we also know that Yt/Kt , θS,Yt , θU,Yt ,
θUt , θSt , nFH,t, and nFV,t are constant. These imply that: (i) final output grows at
the same constant rate as private capital stock in the steady state, which in turn
means that private consumption is also growing at a same constant rate; (ii) labour
supplies grow at the same rate as the population growth rate in the steady state;
and (iii) the number of foreign experts in imitation, nFH,t, and innovation, nFV,t,
are constant.
In the steady state, these constancies indicate that the innovative blueprint-
55
private capital ratio (mRt ), imitative blueprint-private capital ratio (m
It ), as well
as the private consumption-private capital ratio (zCt ) are constant, resulting in
mRt = mI
t = zCt = QRt = 0. Hence, the left-hand side (LHS) of equations (1.57)-
(1.60) can be set equal to zero to derive steady-state values, mI , mR, zC , and QR.
Given the non-linearities associated withmRt andm
It , complete reduced form expres-
sions for mI , mR, zC , and QRare not presented analytically, but instead determined
numerically.
The complexity of the model means that saddlepath stability cannot be estab-
lished analytically, though local stability in the vicinity of computationally derived
steady states can be established for selected configurations of model parameters
using numerical techniques. Nonetheless, since it cannot be fully established ana-
lytically, some configurations of the model may result in the model being locally
indeterminate. This necessitates the use of a computational method solving for a
two-point boundary value problem in any policy experiment, such as the relaxation
algorithm proposed by Trimborn et al. (2008).32 Unlike conventional forward shoot-
ing methods (see Judd (1998)), finite-horizon discrete time approximation methods
(see Mercenier and Michel (1994) for examples), or the backward integration method
(Brunner and Strulik 2002), the relaxation algorithm is more effi cient in dealing with
high dimensional systems and therefore allows us to trace out the unique transitional
dynamics numerically for each of the policy experiments implemented. Likewise, lo-
cal saddlepath stability is also established numerically by calculating the eigenvalues
of the Jacobian of the linearised system for each simulation case considered.
Lastly, note that an alternative regime involving smaller version of the system can
be derived to characterise those developing host economies that have only imitation
sector, similar to the one derived in Agénor and Dinh (2013). The outcome of this
32The relaxation algorithm is a specific type of finite-difference method designed to overcometypical problems faced when solving multi-dimensional continuous time growth models. In additionto approximating the system of differential equations with finite-difference equations on a mesh ofpoints in time, the algorithm also applies a typical error minimisation procedure of shooting methodwhen calculating the time path of solutions. See Trimborn et al. (2008) for a full description ofthe algorithm.
56
would depend on the interactions of the different threshold values along the ability
distribution of the host economy. Specifically, if aFV,t < at < aFH,t < aFP , there
is non-zero supply of skilled workers in the economy but no foreign expert operates
as a Vertical MNC, therefore all skilled workers can only work in the final output
sector. There is only imitation sector in the economy, with non-zero presence of
Horizontal MNCs. However, this case is not examined.
1.4 Model Parameterisation
To illustrate possible impacts of policies, the model is parameterised for an upper-
middle income country with both innovation and imitation sectors, as well as having
non-zero presence of multinationals with Vertical FDI mode. Malaysia, a Southeast
Asian economy that has successfully positioned herself as part of the global pro-
duction value chain of foreign MNCs yet remains entrapped in middle-income trap,
is chosen as the studied economy. In spite of the presence of some leading foreign
MNCs, Malaysia has experienced a lack of success in stimulating knowledge spillovers
and technological diffusion, due primarily to the low innovation absorption capabili-
ties of domestic firms (Yusuf and Nabeshima 2009). Further, as highlighted in stud-
ies such as Flaaen et al. (2013) and Zeufack and Lim (2013), Malaysia is presently
undergoing an ambitious industrial transformation process towards innovation-led
growth using a series of human capital and investment programmes. It is therefore
perfect for the context of this study.
On the household side, the annual discount rate, ρ, and the elasticity of intertem-
poral substitution, σ, are set at fairly conventional values of 0.04 and 0.27 (Agénor
and Montiel 2008). L0 is normalised to unity, with the constant population growth
rate, n, set at the five-year average of 1.73 percent as in 2008-12. The supply of
skilled labour is measured in effi cient units of human capital, and is therefore ad-
justed for average ability. For parameterisation purposes, and given that firm-level
57
distribution of skills (hence also include training expenditure) in Malaysia is gen-
erally not reported in surveys (Sander and Hanusch 2012), the number of effective
skilled labour in the model is defined as the number of workers with tertiary ed-
ucation. The parameterisation strategies for the remaining household parameters
would therefore focus on producing an initial share of skilled workers, θS at 0.240,
given the other fairly standard production parameters used for other sectors. This
involves assuming initial skills acquisition cost, Γ, to be high at 25 percent of skilled
wages, though given the recent establishment of meso-organisations for human cap-
ital development, such as Pembangunan Sumber Manusia Berhad, the effi ciency of
training, ξ is set highly at 0.9. For the ability distribution, both the lower bound
value, am and the Pareto index parameter, χ, is set at a minimum value that would
still satisfy the mathematical properties of χ > 2 and am > 1.
In the imitation sector, for ψI1, the parameter measuring the spillover from the
presence of Horizontal MNCs, Lim (2015), in an empirical study using Productivity
and Investment Climate Survey (PICS) dataset for Malaysia, obtains econometric
estimates in the range of 0.20 − 0.35 for a foreign ownership dummy. The upper
estimate is used in our parameterisation to reflect reasonable strength of spillover
in the imitation sector, therefore ψI1 = 0.35. On the multiplicative parameter of
ψI2, we set ψI2 = −0.3 for the initial baseline to reflect a mildly negative tradeoff
between the productivity of domestic imitators and the cross-term of leading foreign
innovation experts and innovative blueprint stock.33,34
In the innovation sector, for ψR1 , based on case studies such as Rasiah (2012),
a slightly stronger effect of foreign MNCs’ presence on indigenous innovation in
comparison to ψR1 is to be expected, leading to the setting of ψR1 = 0.40. The
33As discussed earlier in the sub-section for Imitation sector, the parameter ψI2 can be inter-preted as either a direct negative effect on imitators’productivity as the size of innovation growsor a positive productivity spillover from leading foreign innovators to domestic imitators, as docu-mented econometrically by Kam (2013). Sensitivity analysis is therefore implemented to examinethe steady-state implications under both cases. For the benchmark case, the negative sign is chosento be consistent with the industrial transformation thesis examined in this chapter.
34Small values for ψI2, irrespective of the sign, are used for the calibration and sensitivity analysis.Large value of ψI2 is destabilising to the model, and this is obvious from the equation for mI
t .
58
stepping stone effect parameter measuring the marginal externality associated with
stock of imitative blueprints, ψR2 , is set initially to a high value of 9.5 to reflect the
historically established industrial base in Malaysia, such as the global electronic and
electrical components manufacturing hubs documented by Kharas et al. (2010) in
Penang, though sensitivity analysis reported later will further assess the effect of a
change in this parameter on the degree of industrial transformation.35
In the final output sector, the elasticity of production with respect to private
capital, α, is set at a fairly standard value of 0.3 (Agénor 2011). The elasticity of
output with respect to composite intermediate goods, γ, is set at 0.3, which is double
the value of 0.15 used by Agénor and Alpaslan (2014) for a low-income economy to
reflect the industrial state of Malaysia, though it remains slightly lower than the 0.36
used by Funke and Strulik (2000) and Sequeira (2011) for developed economies. By
implication of the constant returns-to-scale assumption, that leaves a total of 0.4
between skilled and unskilled labour. Both Agénor and Dinh (2013) and Agénor and
Alpaslan (2014) set the elasticity of production with respect to unskilled labour, βU ,
at 0.2 for low-income economies. To adjust for Malaysia’s middle-income country
status while based on similar proportions to βS, the parameter βU is set at 0.15,
which leaves βS = 0.25. The relative share of basic intermediate in the composite
intermediate inputs, Xt, as measured by ν, is set at 0.57. By comparison, Agénor
and Alpaslan (2014) use a high value of 0.90 for low-income economies. As we might
expect ν to change as industrial transformation takes place over time, a specific sub-
section on endogenous ν is presented later as part of the sensitivity analysis. Lastly,
following Agénor and Dinh (2013), the depreciation rate for private capital, δ, is set
at 0.068.
On the three hiring cost mark-up parameters introduced across the labour-
employing sectors, an initial state with the order of innovation, imitation, and final
35It is worth noting from (1.14) that the stepping stone effect is specified as entering multiplica-tively into the productivity equation of the innovation sector. The multiplicative form is selectedin consistent with relevant studies such as Agénor and Dinh (2013).
59
output sector in terms of rigidity is calibrated, in consistent with observations docu-
mented in Sander and Hanusch (2012). In the model of Agénor and Khazanah team
(2012), the labour cost mark-up parameter in the knowledge-intensive sector (their
model does not distinguish between imitation and innovation) is set at 0.10. We set
this as the value for ΛI , with ΛY = 0.05 being half of it while ΛR = 0.20 doubles
the value to reflect greater diffi culties in hiring workers for the innovation sector.
In the intermediate goods sectors, the substitution parameter η for domestic
production is set at 0.39 to capture a lower elasticity of substitution between inter-
mediate inputs, in comparison to the 0.54 used by Funke and Strulik (2000) or the
0.61 used by Iacopetta (2011), but similar to the non-competitive scenario studied
in Sequeira (2011). In our views, this captures the unique context of the Malaysian
industry very well– a highly specialised global electrical and electronic component
manufacturing hub, and part of the production network of large foreign MNCs.
Regarding the vastly simplified government, the tax rate on final output, τ , is
set equal to 0.25, which corresponds to the average effective tax rate on the output
of Malaysia used by Agénor and Khazanah team (2012). Table 1.1 summarises the
parameter values for the host economy.
Moving on to the foreign sector, in the representative objective function for for-
eign experts in the host economy, recall that the elasticities of substitution abide
by the assumption of σF > θF > 1, as in Brambilla et al. (2009). The between-
variety elasticity, σF , is first set arbitrarily at 2. The across-variety elasticity for
foreign preference, θF , is then set at 1.64, which is calibrated to reflect a correspond-
ing substitution parameter of 0.61, the value used by Agénor and Alpaslan (2014)
for substitution parameter in the production side. This is deliberately calibrated
to reflect the different preferences of foreign experts who come in with different
know-how, though the combination of calibrated values for σF and θF is reasonably
consistent with studies using nested utility framework. As stated, the normalisation
of P0 = 1 is applied. The parameterisation for the Lerner Index, LI, is based on
60
the average empirical estimates of profit margin, 0.2544, for Malaysian manufac-
turing firms in Zeufack and Lim (2013). A simple approximation measure for LI
is just 1 − 0.2544 = 0.7456. For the basic doing-business cost of F0, a value of
0.2733 is calibrated, based on the average cost of business start-up procedures as a
percentage of real GDP per capita reported in the 2004, 2006, and 2008 version of
World Bank Doing Business Surveys. For F1 and F2, given the imposed assumption
of F2 > F1 > F0, F1 = 0.33 and F2 = 0.40 are set, which imply that the cost
incurred by foreign subsidiaries to come in with experts with standardisation and
sophisticated know-how would be one-third and forty percent of the baseline price,
P0 = 1. As policy scenarios involving cuts in F1and F2 are examined extensively
in simulation exercises later, these initial calibrated values are intended to reflect
an initial situation where it is expensive for foreign experts to operate in the host
economy. In terms of the asymmetric cost parameter, φ = −1 is conveniently set to
reflect a constant rate of decreasing return associated with 1/$.36
The total number of foreign experts entering into the host economy, NF,t in each
period is normalised to one. In terms of the parameters in the Weibull process used
to model the evolution of foreign preferences, the shape parameter, ωk, and the
scale parameter, ωλ, are set equal to 1 and 2 respectively. For the shares of the
three different types, the FDI compositions for Malaysia are estimated using data
from the U.S. Bureau of Economic Analysis (BEA). Indeed, the compositions of
inward FDI stock from the United States (U.S.) for different East Asian economies
are presented in Figure 1.1. Due to the constraints of existing FDI statistics classi-
fication (by broad industry or country, not MNCs’operations or value chain), the
breakdown based on American MNCs’foreign affi liates from BEA is used, as it is
the only national agency with suffi ciently long time series of such details.37 Based
36For robustness check, we experimented with an increasing rate of decreasing return (φ < −1),and a decreasing rate of decreasing return (0 < φ < −1). For the range of φ values where the modelstill solves, the calibration of φ does not produce significant difference to the policy experimentresults in the next section.
37Ideally, the availability of firm-level enterprise survey data on an annual basis would allowus to adopt the approach of Lim (2015) to distinguish the three types of FDI modes. In the
61
on the estimates, the initial proportion of Non-mandated (nFP ), Horizontal (nFH),
and Vertical MNCs (nFV ) are calibrated to equal 0.3099, 0.6737, and 0.0164 respec-
tively.38 To obtain these initial values for the FDI compositions in an initial steady
state that is saddlepath stable, it turns out that the constant value a, and the con-
stant term, wm in the international product market dimension feedback channel are
set simultaneously at 9.55 and 3.6 respectively. Lastly, using the expression for LI
adopted from Allanson and Montagna (2005), we estimate the initial value of aFP
at 24.656.
To establish that the initial steady state is consistent with aFV < aFH < aFP ,
first, rearranging (1.42) would allow us to calculate the threshold value of entry
for Vertical FDI, aFV , to equal 3.155. Then, given the values for aFV , aFP , the
initial steady-state value for nFH , and other calibrated parameters, the threshold
value for Horizontal FDI, aFH , can be calculated by rearranging (1.41), yielding
aFH = 23.392 < aFP . The theoretical condition of aFV < aFH < aFP is therefore
satisfied in the initial steady state. The parameter values used for the foreign sector
are summarised in Table 1.2.
For the main variables of interest, parameterisations for the initial steady-state
labour proportions work as follows. As stated, from data, we know θS = 0.240.
Further, based on estimated statistics on the percentage share of R&D researchers
in Malaysia, the share of effective skilled labour in innovation, θS,R, is estimated
at 0.045. These imply that θS,Y = 0.195. Knowing the initial values for θS and
θS,Y , as well as the calibrated values for am, χ, ξ, βS, βU , we can rearrange (1.65)
to calculate for the absolute share of unskilled labour in final output production,
θU,Y , which equals 0.0231. Then, rearranging (1.64), the share of unskilled labour in
absence of such data, the classification is based largely on Markusen’s (1998), as well as those ofBrainard (1997) and Braconier et al. (2005), and the financial and operating data of majority-owned nonbank foreign affi liates of U.S. is used to estimate for the composition of MNCs. SeeSection 1.7.1 for further details.
38The respective shares of the foreign MNCs in Malaysia are based entirely on the preliminaryempirical estimation exercise using US BEA dataset, as presented in Figure 1.1. For a maturedindustralising economy with historically significant presence of foreign multinationals, it is realisticthat Horizontal MNCs have much more significant presence than Nonmandated MNCs.
62
the population, θU , would equal 0.9856. By implication, the proportion of unskilled
labour working in the imitation sector can then be calculated as equal to 0.9625.39
For the parameterisation of the industrial composition ratio, the average of
Malaysia’s share of high technological exports as percentage of total manufactured
exports is calculated for the year between 2008 and 2011, yielding 0.4164. The in-
dustrial composition ratio measures the ratio, mt = mIt/(m
Rt +mI
t ), which means its
initial steady-state value would equal 1−0.4164 = 0.5836. To measure the degree of
innovation expertise in host economy, the foreign-to-domestic innovation expertise
ratio, Ψt, is defined as the ratio of the number of foreign experts with sophisti-
cated know-how to the number of skilled workers in innovation sector. Recalling
that both NF,t and Lt are normalised to one in the model, we can therefore write
Ψt = nFV,t/ θS,R,t to compute for the innovation expertise ratio in each period. The
initial steady-state value of Ψt turns out to be 0.3672.40
Finally, for the initial steady-state growth rate of final output, a multiplicative
constant is introduced to yield both an initial annual growth rate for final output
and private capital stock to equal 4.3 percent per annum, which corresponds to the
average growth rate for Malaysia in the period of 2008-13. By implication of the
properties of the initial steady state, private consumption growth is also equal to
4.3 percent.
1.5 Policy Experiments
Similar to Agénor and Dinh (2013) and Agénor and Alpaslan (2014), policy out-
comes concerning the industrial structure (measured by the industrial composition
ratio of mt = mIt/(m
Rt + mI
t )) and total skilled workforce expansion (measured by39Following Agénor and Dinh (2013), we introduce inertia in the labour adjustment process to
prevent unrealistic jumps in the transitional dynamics. The relevant static equations are thereforesolved as dynamic equations in their partial adjustment form, though these are merely nuances innumerical simulations that make no material difference to the actual solutions.
40In the absence of data on the embodied human capital of experts, we retain the calculated ratiothat is based on nFV,t and θS,R,t. Alternatively, we can also introduce a multiplicative constantto normalise the value to an index, though this will not make material difference.
63
both skilled labour share, θS,t, and skilled labour in innovation, θS,R,t) are the key
policy indicators to be examined. To measure progress on the deepening of domestic
innovation expertise, the foreign-to-domestic innovation expertise ratio, Ψt, is intro-
duced as it provides a more meaningful policy interpretation than the individual
measure of share of Vertical MNCs, nFV,t, and share of skilled labour in innovation,
θS,R,t.
Given that the key interest here is industrial transformation (a long-term policy
reform issue and therefore needs to be analysed independent of business cycle in-
fluence), and the fact that FDI, unlike portfolio investment, is stable over time, all
policy experiments considered are permanent in nature. Policies considered in ad-
dition to foreign investment liberalisation measures are in the broad area of human
capital policies, specifically a permanent reduction in skills acquisition cost and a
permanent removal of labour market rigidity-induced cost mark-up in the innovation
sector. In addition, to ensure that households do not permanently lose out due to
transformation, the long-run steady-state effect on aggregate private consumption
growth (Ct/Ct) is also evaluated, with a policy option considered to be acceptable
only if the growth rate is sustained or increases in the steady state.41 Individual
policies are first discussed, followed by different variations of composite policy pack-
ages. These are then followed by a specific sensitivity analysis involving endogenous
technological change, where the parameter ν is made endogenous to the state of
industrial transformation.41When solving for the continuous time dynamic problems over the entire infinite time horizon,
the numerical method of relaxation algorithm allocates mesh points unevenly such that the timedifference between result observations generated increasingly widens over time. The steady-stateresult therefore would dominate other observations along the time path in any integrable mea-sure like the conventional welfare calculations. Higher steady-state growth in aggregate privateconsumption therefore necessarily reflects improvement in welfare.
64
1.5.1 Individual Policies
Skills Acquisition Cost
Consider first a permanent reduction in skills acquisition cost, Γ, from 0.25 to 0.18.
This may be thought of as a subsidy scheme designed to reduce the cost of pursuing
advanced skills, obtained by reallocating spending within the budget, so that the
tax rate remains the same and the overall balance remains.
The cost reduction associated with skills acquisition induces more workers to
invest in skills. This leads to an expansion in both the proportion of skilled labour
employed in the final output and the innovation sectors. At first, the increase
in skilled labour supply lowers skilled wages. At the same time, the rise in skilled
employment promotes activity in both innovation and final output production, which
would raise the marginal product of unskilled workers and consequently, unskilled
wages. This nets offsome of the skills acquisition incentive, resulting in ‘scaling back’
for both effective shares of total skilled labour and those employed in innovation.
The respective absolute deviations from the initial steady state are 0.69 and 0.13
percentage points respectively.
The innovation sector expands while the imitation sector contracts, leading to a
decline in the industrial composition ratio by 0.43 percentage points. Similar to θS,
the initial contraction of imitative varieties is more significant than the end steady-
state effect. However, as the ratio of skilled and unskilled employment is ultimately
tied to the relative wage ratio, the eventual ‘scale-back’of unskilled employment
causes the industrial composition ratio to settle at just a slightly lower level than the
initial steady state. This is the same for the proportion of foreign innovation experts
with sophisticated know-how, nFV , where despite uneven paths along the transition,
long-run permanent changes are negligible. In terms of the relative measure of
foreign-to-domestic innovation expertise ratio, Ψ declines from 0.3672 to 0.3527.
This indicates a small deepening of relative domestic innovation expertise by 3.9
65
percent. Lastly, the steady-state effect on aggregate private consumption growth is
negligible.
In Table 1.4, additional sensitivity analysis on key elasticity parameters in both
the innovation (ψR1 and ψR2 ) and imitation (ψ
I1 and ψ
I2) sectors, are also carried out
for this shock. It can be seen that the impact on industrial transformation is more
profound the larger the learning effect (ψR2 ) is, as the economy benefits from the
greater strength of the stepping stone from imitation. The difference for the other
variables are generally negligible. These results are largely consistent with those in
Agénor and Dinh (2013), where strong learning effects mean greater improvement
in the productivity of innovation workers. In the case of ψI2, if the externality
associated with the cross term, nFV,tMRt , is specified instead as a positive feedback
to imitation, the industrial transformation outcomes are similar to the benchmark
case though the gain in domestic innovation expertise is smaller.
Reducing Hiring Cost in Innovation Sector
Next, consider the individual policy of a permanent reduction in the hiring cost
mark-up associated with the employment of skilled researchers. In Malaysia’s con-
text, this policy may be interpreted as bringing about similar effects to the type of
initiatives implemented by the semi-statutory body of TalentCorp Malaysia in re-
cent years.42 Specifically, a permanent reduction in ΛR from 0.2 to 0.0 (a 100 percent
reduction in labour cost mark-up in the innovation sector) is simulated. Simulation
results for the four main variables of interest are presented in Figure 1.5.
While changes in the industrial composition ratio and effective skilled labour
share appear to be largely similar to Figure 1.4, the policy effects here operate
mainly through the skilled labour reallocation channel. As skilled workers become
relatively more expensive in the production of final output, more skilled labour are
42TalentCorp Malaysia was established on 1 January 2011 under the Prime Minister’s Depart-ment of Malaysia to formulate initiatives to address the availability of talent in line with the needsof the country’s economic transformation.
66
employed in the innovation sector. However, similar to skills acquisition cost cut,
there is a secondary effect that mitigates the expansion, resulting in the hump shaped
patterns observed for effective skilled labour. The decline in the cost of skilled labour
in innovation tends to raise the unskilled-skilled wage ratio, which would then take
away some of the skills acquisition incentive associated with the initial expansion
of the innovation sector. More specifically, the re-allocation of skilled labour away
from θS,Y to θS,R would result in θS,R increasing by 0.72 percentage points at end
steady state, while θS,Y declines by 0.58 percentage points. Overall, total effective
skilled labour share expands by 0.14 percentage points.
Even though the ‘scale-back’ in innovation sector expansion observed earlier
(with skills acquisition cost cut) remains in action, it is less significant as the link
with the relative wage ratio adjustment is less direct here. The reduction in ΛR
leads to a proportionate decline in the effective hiring cost of skilled labour in inno-
vation, but given that both ΛY and ΛI stay the same, the unskilled wage adjustment
mechanism resulting in subsequent disincentive in skills acquisition is less in action
here. As such, the expansion in innovation relative to imitation is more effective,
therefore resulting in a larger permanent reduction of 3.25 percentage points in the
industrial ratio, m. Similar to the results associated with skills acquisition cost
cut, the steady-state effect on the proportion of foreign experts with sophisticated
know-how, nFV , is negligible. However, the policy impact on the relative measure
of Ψ is much larger due to the strong reallocation effect, where domestic innovation
expertise improves considerably relative to foreign expertise in the host economy
(Ψ declines from 0.3672 to 0.3119, which indicates a relative deepening of domestic
innovation expertise by 15.1 percent). Lastly, in the steady state, aggregate private
consumption growth increases marginally by 0.1 percentage points from the initial
baseline.
Two other sensitivity results are presented in Figure 1.5, where the transitional
pattern of shock associated with a larger stepping stone effect, ψR2 = 15.5, is mostly
67
similar to the benchmark case (other than a steeper decline in industrial ratio by
3.83 percentage points). Similar to the skills acquisition cost cut, when the exter-
nality associated with the cross-term of foreign innovation experts and innovative
blueprint stock (nFV,tMRt ) is specified as having positive feedback (ψ
I2 = 0.3) to
imitation (instead of a negative tradeoff as in the benchmark parameterisation), a
more favourable outcome is observed for the industrial composition ratio (m declines
by 3.3 percentage points) without the corresponding decline in the share of Vertical
MNCs, nFV,t. This suggests that, in terms of domestic labour market and skills
expansion policies, slightly favourable industrial transformation outcomes can be
achieved when there is positive externality from the foreign innovation experts to
the productivity of domestic imitators.
Lastly, the experiment with a simultaneous cut in ΛY also yields results with
similar transitional patterns, with deviations observed in variables generally smaller.
This is due to the cut in ΛY producing a mitigating effect because of: (i) less
skilled final output worker reallocating to the innovation sector, and (ii) smaller
skills acquisition incentive due to cut in ΛY also reduces effective cost of hiring
unskilled labour in the final output sector. Nonetheless, the steady-state effect of
a rise in effective skilled labour share is actually larger with the additional ΛY cut
due to the effects of overall skilled labour expansion outweighing that of point (ii)
mentioned above.
Foreign Investment Liberalisation Measures
In the model context, the policy measures considered here involve a permanent
reduction in the ‘doing-business’costs for foreign experts, namely the basic doing-
business cost, F0; the additional cost incurred by foreign subsidiaries of Horizontal
nature, F1; and the additional cost incurred by Vertical operation with leading
foreign innovation experts, F2. The reduction of these costs may be interpreted as
an outcome from some specific targeted investment liberalisation or deregulation
68
measure implemented by the host economy.
First, we consider individual effects associated with each of the three fixed costs.
Recall that F0 is incurred on all types of foreign experts in the host economy, while
F1and F2 are additional costs incurred by the specific type of foreign experts. Pre-
dictably, a cut in the basic cost of F0 would unambiguously bring about positive
effects on both nFH and nFV . Nonetheless, for the add-on cost of F1 and F2, by
implication of the foreign sector specification, as well as owing to the asymmetric
nature of the perceived productivity difference from the perspective of foreign in-
novation experts, the policy experiments produce some interesting results that may
partly help to explain the phenomenon often observed in real life, where competing
host economies offering the best financial incentives often do not end up attracting
the best foreign innovation experts with frontier know-how.43
Simulations on F2: Consider a permanent reduction of F2 from 0.40 to 0.37,
which is a three percent reduction in terms of the baseline theoretical price (equiv-
alently, in relative terms, a 7.5 percent drop from the initial 0.40). While a host
economy may intend to attract more foreign experts with sophisticated know-how by
reducing the additional cost incurred on them, this results in an adverse signalling
effect where the proportion of foreign subsidiaries in Vertical mode is reduced. A
reduction in F2 would ceteris paribus, be expected to result in an expansion of the
perceived investment value for a typical foreign experts j with sophisticated know-
how. Nevertheless, given the equi-profit condition used to derive threshold value
for Vertical MNCs, aFV , the asymmetric productivity term, $φFV , would have to
adjust, as seen from (1.39). The reduction in F2 puts a downward pressure on $FV
(and increases the information cost associated with perceived productivity differ-
43These are summarised in studies on FDI policy competition, such as Oman (2000) and Blom-ström (2002). In essense, this branch of the literature argues that the quality of the enablingenvironment of investment (for examples, human capital quality), especially for foreign firms withinvestments in technological leadership areas, affects a country’s ability to attract quality FDImore than direct investment incentives. Indeed, it can be costly and counterproductive to offerinvestment incentives if the ‘fundamentals’of the potential host economy are bad.
69
ence, 1/$FV ), and this results in a lower and stricter threshold value for Vertical
MNCs, aFV . Foreign subsidiaries are therefore less willing to operate with experts
in sophisticated know-how in the host economy, resulting in a reduction of nFV .
Intuitively, these effects may be interpreted as follows. While typical direct in-
vestment incentives may be attractive to new firms, consistent with Horstmann and
Markusen (1996), the reduction in F2, without an accompanying cut in F0, can lead
to an adverse signalling type of outcome. Given the asymmetric structure speci-
fied for the internalisation decision of a typical foreign innovation expert in Vertical
MNC mode, foreign subsidiaries in the host economy would face increasing diffi cul-
ties in discriminating the best among the most productive ones. This productivity
uncertainty associated with the asymmetric cost structure of a typical Vertical MNC
means a smaller F2 in (1.39), resulting in existing foreign subsidiaries of the host
economy being relatively more wary of the information cost associated with per-
ceived productivity difference for a typical Vertical operation, 1/$FV (compares to
1/$FH), therefore preferring the alternative of bringing in an expert with standard-
isation know-how. In the benchmark simulation, nFH increases by 4.4 percentage
points while nFV drops by 0.5 percentage points. While the decline of nFV seems
to be counter-intuitive, it actually corroborates well with the findings in the OECD
comparative study on tax holidays, which presents cases where the elimination of
such narrowed incentives did lead to long-term improvements in FDI performance
for certain developing economies (OECD 2008). Likewise, it is also consistent with
the empirical findings documented in the various ‘race-to-the-bottom’studies, where
the sole implementation of incentives targeting only the ‘big names’often results in
adverse signalling outcome.44
The expansion in nFH further creates a secondary effect: it leads to an expansion
44Examples of the ‘race-to-the-bottom’studies include Oman (2000), Blomström (2002), Vogeland Kagan (2004), and Olney (2013). These studies document similar adverse signalling effectsof narrowed FDI-promoting policies. In the context of the analysis here, a cut in F2, without anaccompanying F0 cut, is viewed adversely by foreign subsidiaries as a signal of shortage in domesticinnovation expertise and lower productivity of domestic workers they are going to be matched to.
70
in imitative goods relative to innovative goods in the host economy due to a rise
in productivity of imitation. This results in industrial composition ratio, m, rising
by 5.6 percentage points (see Figure 1.6). The corresponding increase in unskilled
workers hired in imitation, θU,I , given a fixed number of unskilled workers, θU , means
a fall in the unskilled workers employed in final output production, θU,Y . The rela-
tive wage ratio is determined in the final output sector, which hires both skilled and
unskilled workers. A decline in θU,Y , ceteris paribus, results in an increase of the
unskilled-skilled wage ratio. This in turn disincentivises skills acquisition and sub-
sequently, employment in the innovation sector. In the steady state, this is reflected
as a decline in θS and θS,R by 0.36 and 0.09 percentage points respectively.45 Never-
theless, as the decline in θS,R is much milder relative to nFV , the relative domestic
innovation expertise in the host economy improves, with Ψ declining from 0.3672
to 0.2563. This indicates a relative deepening of domestic innovation expertise by
30.2 percent, though much of this is driven by the significant drop of foreign experts
with sophisticated know-how in the host economy. Lastly, in the steady state, as
imitation-based varieties remain the main intermediate type used in final output pro-
duction, the expansion in innovative varieties raises aggregate final output growth
by 0.2 percentage points. By implication of an increase in final output-to-private
capital ratio (Yt/Kt) and therefore rt as in (1.20), aggregate private consumption
grows by the same percentage points too.
Other sensitivity results concerning this specific shock are summarised in Table
1.4, where the adverse signalling steady-state effects associated with F2 cut are
consistently observed, with the effects on m being stronger the higher ψR1 (greater
reliance of domestic innovation in Vertical MNCs), or the higher ψR2 (greater learning
associated with the stepping stone effect) is. Indeed, the simulation results are
largely consistent with the Malaysian experience over the past two decades, where
45The simulation results (a decrease in nFV and a relatively cheaper skilled wage) is consistentwith the empirical findings of Braconier et al. (2005), who document that Vertical MNC activitiestend to get larger (smaller) when the skilled labour become relatively expensive (cheaper).
71
the Malaysian administration had been among the most active ‘open-door’regime
with respect to offering all forms of targeted incentives to attract foreign firms at the
global frontier, yet failed to attract many of such foreign firms (Yusuf and Nabeshima
2009).
Simulations on F1: Next, consider a permanent reduction of F1 from 0.33 to
0.30. The same three percent reduction in terms of the baseline theoretical price
is maintained, though it is equivalent to a 9.1 percent drop from the initial 0.33 in
relative terms. While the steady-state effects presented in Table 1.4 show largely
opposite results to the previous cut in F2, the underlying operating mechanism for
a reduction in F1, without an accompanying cut in F0, is slightly different. Unlike
the F2 cut, in the primary sorting channel, a direct investment incentive in the form
of a F1 cut would bring about positive effects to both nFH and nFV . As seen from
(1.38), a reduction in F1 would bring about an increase in $FH (or equivalently,
a reduction in information cost associated with perceived productivity difference,
1/$FH). This in turn would result in a relaxation of the threshold value of entry
for a Horizontal mode of operation, aFH , therefore providing greater incentive for
foreign experts with standardisation know-how to come into the host economy. This
is what would have been expected in the previous shock if there is no asymmetric
cost structure for Vertical FDI (arising from the growing diffi culty in identifying the
best among the most productive talents at the ‘deeper ends’of ability distribution,
as aFV gets more restrictive). In (1.39), given fixed F2, the reduction in F1 widens
the comparative cost gap, F2 − F1. In this case, the asymmetric cost structure for
Vertical MNCs brings about a positive signalling effect, therefore resulting in higher
$FV (or equivalently, a reduction in 1/$FV ). This leads to a relaxation of the
threshold value of entry for Vertical MNCs, aFV , and provides greater incentives for
foreign experts with sophisticated know-how to come into the host economy.
The share of foreign innovation experts, nFV , increases, and this then results
72
in an expansion of the innovation sector relative to the imitation sector, hence a
drop in the industrial composition ratio, m. As the flow of innovation production
increases, there are more skilled workers hired in the innovation sector. Given initial
fixed supply of skilled labour, this reallocates skilled labour away from final output
production, which then puts downward pressure on the unskilled-skilled wage ratio,
wU/wS. This creates greater incentives for skills acquisition. In the steady state, the
shares of effective skilled labour, θS, and those employed in innovation, θS,R, expand
by 0.38 and 0.09 percentage points respectively. Overall, the steady-state effect for
the industrial composition ratio, m, is a decline of 3.33 percentage points. In terms
of the foreign-to-domestic innovation expertise ratio, Ψ increases from 0.3672 to
0.4103, indicating a growing reliance on foreign innovation expertise in the host
economy.
In terms of sensitivity analysis, it can be observed from Table 1.4 that the out-
come of industrial transformation is more favourable when either of the four elasticity
parameters in the blueprint-production sectors examined is larger. This is notable
for the two parameters in the innovation sector (ψR1 and ψR2 ). Nevertheless, in all four
cases, the disadvantage of this specific policy shock is that it is achieved through
a growing reliance on foreign experts in innovation expertise since nFV grows at
a larger magnitude than θS,R. This is most apparent for the case where there is
positive feedback from the cross-term of nFV,tMRt to the productivity of imitation
(ψI2 = 0.3), with the foreign-to-domestic innovation expertise ratio, Ψ, increases by
more than the benchmark case. In addition, it can also be seen from Figure 1.7
that the transition paths are more volatile in this case, since Vertical MNCs are
not only driving innovation but also having a positive spillover to imitation, hence
more complicated dynamics are observed. Results on steady-state effects for other
sensitivity analysis are also presented in Table 1.4.
73
Simulations on F0: Next, consider a permanent reduction of F0 from 0.2733 to
0.2433. While the same three percent reduction is maintained, this is equivalent to
an 11 percent cut from its initial value. This may be interpreted as an economy-wide
liberalisation attempt aimed at reducing general administrative cost for all foreigners
in the host economy. As F0 is the basic cost involved for all foreign MNCs, ceteris
paribus, this would create incentives for foreign firms to adopt an improved mode of
operation and bring in foreign experts with more advanced know-how. Given that
nFP is treated as a residual, this would result in an unambiguous increase for both
nFH and nFV . For Vertical MNCs, the reduction in total cost required to be paid
every period (F0 + F2) means there will be an unambiguous increase of nFV in the
steady state, of 0.2 percentage points. Similarly, for Horizontal MNCs, the reduction
in total cost required to be paid every period (F0 +F1) results in an increase of nFH
by 3.8 percentage points.
The increase in both nFH and nFV leads to an expansion for both the imitation
and the innovation sector, though the latter grows more in relative terms. Specifi-
cally, the industrial composition ratio, m, declines by 1.34 percentage points in the
steady state. As the innovation sector expands relatively faster than the imitation
sector, more skilled workers are relocated out of final output production compared
to unskilled workers’reallocation to imitation. This tends to put a downward pres-
sure on the relative wage ratio, wU/wS (recall that it is determined by a function
of θS,Y /θU,Y ). This then creates greater skills acquisition incentives and leads to an
increase in the effective supply of skilled labour. Specifically, in the steady state,
these effects translate to moderate expansions in θS and θS,R. The relatively small
increase in θS,R comparing to nFV also means that the foreign-to-domestic inno-
vation expertise ratio, Ψ, increases from 0.3672 to 0.4111. In relative terms, this
means domestic innovation expertise deteriorates by 12 percent, indicating a growing
reliance on foreign innovation experts in the host economy.
In terms of sensitivity analyses presented in Table 1.4, cases with larger parame-
74
ters in the innovation sector (ψR1 = 0.8 and ψR2 = 15.5) would produce more effec-
tive industrial transformation results, underlying the importance of the strength of
learning effects in the innovation sector– the former (ψR1 ) denoting the direct learn-
ing from foreign experts in Vertical mode, the latter (ψR2 ) denoting the stepping
stone effect from imitative knowledge– to drive industrial transformation. Again,
for the case where ψI2 is positive, the transition paths display cyclical properties and
overshooting patterns. This suggests that, If a less volatile transition path for the
industrial transformation process is desired, an environment where the growth of
Vertical MNCs in the innovation sector would phase out the domestic imitators will
be more supportive.
Indeed, given that F0 is calibrated based on the basic doing-business cost and
therefore captures the institutional quality aspect faced by foreign investors, a nat-
ural extension is to examine whether the policy results observed are influenced by
the initial parameterised value, and whether there exists any threshold value. To do
these, within the F0 ∈ [0.159, 0.519] range where the model still solves, we simulate
the experiment repeatedly across a grid of four decimal places. Specifically, for the
basic doing-business cost, F0, the same 11 percent cut is simulated for the different
initial values. It is noted that there are clear level effects where the higher the ini-
tial cost of doing-business is, the larger the deviation is observed for the industrial
composition ratio, mt. It turns out that there is a threshold value for the initial
doing-business cost, below which final output growth effect is negative. This value
is F0 = 0.2964. The key policy implication from this analysis is therefore that, the
usefulness of broad-based investment liberalisation measure in promoting output
growth in a developing host economy may depend on its initial doing-business cost
as measured by the World Bank. For Malaysia, its initial doing-business cost is
below the threshold value for her to rely solely on F0 cut to promote output growth.
75
1.5.2 Composite Policy Reform Programmes
A key goal that policymakers in developing economies often seek to achieve when
implementing composite reform programmes involves identifying the best combina-
tion of individual policies to reap the benefits of policy complementarities. The main
premise of this study is that a composite programme delivering the best outcome
of industrial transformation, overall skills expansion, and a deepening of domestic
innovation expertise, while simultaneously attaining positive changes in final output
and aggregate private consumption growth rates, will be the preferred composite
programme. The key complementarity between labour and foreign investment lib-
eralisation policies is best illustrated here, since a successful deepening of domestic
innovation expertise– relative to foreign expertise– in the host economy would see
a reduction in the foreign-to-domestic innovation expertise ratio, Ψ.
Consider three different composite policy reform programmes, which combine
the policies of a skills acquisition cost cut (Γ from 0.25 to 0.18), the innovation
sector-specific labour market reform (ΛR from 0.2 to 0.0), and different combina-
tions of the three foreign investment liberalisation measures discussed. Specifically,
Composite Programme A combines both the skills acquisition cost and innovation
sector labour cost mark-up reductions with a simultaneous reduction in F0, F1, and
F2 by 0.03. Composite Programme B combines the proxies for education and labour
market policies with a proportionate cost-cutting programme where F0 is reduced
by 0.01, F1 reduced by 0.03, and F2 reduced by 0.05, while Composite Programme
C combines the Γ and ΛR reductions with a proportionate cost-cutting programme
tilted towards providing basic investment incentives to all foreigners (F0 reduced by
0.05, F1 reduced by 0.03, and F2 reduced by 0.01).
The results of the three composite policy reform packages implemented in the
benchmark model are illustrated in Table 1.5 and Figure 1.9. The transitional
paths of the key policy variables examined largely conform to what would have
been expected when the effects of the individual policies are combined. Both the
76
simultaneous foreign cost-cutting programme and the proportionate cost-cutting
programme with F0 cut by 0.05 produce positive deviation in the share of Vertical
MNC, nFV , in the steady state. At the same time, the skills acquisition-stimulating
cost reduction measures of Γ and ΛR cuts would create greater incentives for labour
to not only undergo training, but also work in the innovation sector. The increase
in skilled labour supply would initially put a downward pressure on skilled wages.
However, due to the overall increase in skilled employment occurring in both the
innovation (θS,R) and final output sector (θS,Y ), a secondary effect would also be
at play: the expansion of innovative blueprints relative to imitative blueprints, and
conversely, the varieties of sophisticated intermediate inputs relative to basic inputs.
This shift towards innovation raises the productivity of labour in that sector, which
magnifies the initial effect. Nonetheless, the increase in the supply of skilled labour
in final output production would also raise marginal product of unskilled workers,
which then raises unskilled wages. This then mitigates the initial effect on incentives
to acquire skills, and the labour market adjustment dynamics are reflected in the
humped and U-shaped pattern associated with θS and m (as well as Ψ) in Figure
1.9.
The decline in imitative varieties would further feed back into the foreign firms’
internalisation process, which creates a tertiary dynamic that is then reflected in
the cyclical pattern of m and Ψ in Figure 1.9. The decline in imitative varieties
makes the host economy less attractive as a host to Horizontal MNCs, but at the
same time improves the incentive for foreign innovation experts with sophisticated
know-how to enter. In the case of Composite Programme A, this therefore mitigates
the initial decline in nFV and results in an overall steady-state increase of nFV , while
in the case of Composite Programme C, it further leads to growth in the share of
foreign innovation experts in the host economy. Overall, while the host economy
would experience improvements in both industrial composition (a decline in m)
and relative domestic innovation expertise (a decline in Ψ) under both Composite
77
Programme A and Composite Programme C, the balanced Composite Programme A
would be the better programme as it sustains aggregate private consumption growth
whereas Composite Programme C would lead to a slight decline in the steady state.
In contrast, the Composite Programme B results in largely opposite results. The
share of foreign experts in the Vertical MNC mode, nFV , would decline in the steady
state due to the adverse signalling effects associated with the large F2 cut. This then
results in ‘reverse transformation’towards imitation, less incentive to acquire skills
and work in innovation sector, hence a drop in both effective skilled workers, θS, and
those employed in the innovation sector, θS,R. In terms of steady-state aggregate pri-
vate consumption growth, Composite Programme B predictably delivers the largest
gain of 0.22 percentage points, but unlike the preferred Composite Programme A,
this is maintained by not making much progress in industrial transformation.
Tables 1.5 and 1.6 present additional simulation results for nine sensitivity tests.
While steady-state effects for other key variables are also documented, we focus on
the industrial composition ratio (m) and the foreign-domestic innovation expertise
ratio (Ψ), the two key indicators of interest. When the elasticity of blueprint produc-
tion with respect to foreign experts in either the innovation (ψR1 ) or imitation sector
(ψI1) is calibrated at a higher value, Composite Programme C (which depends more
on the inflow of foreign innovation experts to drive industrial transformation) would
see a larger decline inm at the cost of a largerΨ. On the other hand, while the policy
effects on both indicators are milder under Composite Programme A when foreign
experts have a greater influence on the host economy’s design activities (hence ‘tak-
ing away’some of the effectiveness of the human capital and labour market policies),
the more balanced reform programme continues to have the edge over Composite
Programme C for the gains made in the deepening of domestic innovation exper-
tise, as well as sustaining growth rates in private consumption. Similar results are
also observed when sensitivity analysis is implemented with a positive externality
specification for the parameter, ψI2. In a nutshell, the relatively balanced Composite
78
Programme A would tend to deliver more effective industrial transformation out-
comes compared to Composite Programme B, while being much better at promoting
the deepening of domestic innovation expertise in the host economy when compared
to Composite Programme C. The results from these policy experiments are gener-
ally consistent with the consensus views surveyed and documented in Saggi (2002)
and Faeth (2009), where evidence on the direct role of FDI in promoting indigenous
knowledge activities are mixed, but their indirect impacts on domestic economy tend
to be positive if their presence leads to a deepening of innovation expertise among
domestic agents.
Meanwhile, when the externality parameter associated with learning effects in
both the innovation sector (the stepping stone effect from the stock of imitative
goods, ψR2 ) is calibrated at a higher value, the steady-state effects on both the
industrial composition ratio (m) and foreign-domestic innovation expertise ratio
(Ψ) are unambiguously more effective in all three composite programmes. As an
illustration, Figure 1.10 presents results on the steady-state deviations of m across
different combinations of ψR2 and ψI2, and the strong effects associated with a larger
stepping stone observed are consistent with findings in Agénor and Dinh (2013) and
Agénor and Alpaslan (2014).
In terms of other parameters, an interesting case to discuss is when the substi-
tution parameter for intermediate goods production is parameterised at a higher
value, specifically η = 0.54 as in Funke and Strulik (2000). This indicates greater
substitutability between intermediate goods in domestic production. In this case,
the effectiveness of Composite Programme A and Composite Programme C in driving
industrial transformation becomes lower, with m declining, and θS and θS,R increas-
ing at lower rates. The lower substitutability between intermediates effectively takes
away the effectiveness of policies in expanding innovative varieties as it implies that
each unit of intermediate input is priced lower. For each gain from expansion of in-
novative varieties, the associated benefits to improving skills acquisition incentives
79
will also be lower, hence resulting in smaller gains of effective skilled labour and
those employed in the innovation sector. In terms of domestic innovation expertise,
even though the indicator of Ψ declines more (compared to the benchmark case),
this relative deepening is spurious as it is attained when both labour market and
FDI-promoting policies become less effective under this scenario.
Lastly, for the balanced and generally less volatile Composite Programme A,
Table 1.8 illustrates the benefits of the implementation of composite packages. In
comparison to the ‘sum of parts’from aggregating steady-state effects of all individ-
ual policies, the implementation of a composite reform programme clearly exhibits
policy complementarity. The decline in the industrial composition ratio, the ex-
pansion of effective skilled labour and those employed in the innovation sector, as
well as the increase in the share of foreign experts with sophisticated know-how
(Vertical MNC) in the host economy, are of larger magnitude compared to when
merely summing up effects from all the individual policies implemented in isolation.
However, the fact that there is an increase in the number of foreign innovation ex-
perts under the composite programme means the relative measure of Ψ declines by
less. Likewise, the positive steady-state deviation in aggregate private consumption
growth– growing at the same rate as final output in the steady state– is actually
slightly lower under the composite programme. This is due to the fixed share of
basic inputs (in composite intermediate inputs, ν) used in final output production
(biased towards imitation-based basic input), therefore leading to less expansionary
effects from policies. Nevertheless, as would be seen in the next sub-section, when ν
is allowed to change over time, the composite policy programme would generate even
more complementarity and attain the desired outcome in all indicators examined.
80
1.5.3 Endogenous Technological Change and Policy Com-
plementarities
In addition, we consider endogenous change in the industrial production structure.
As pointed out by Agénor and Dinh (2013), as the process of industrial transforma-
tion gradually takes place over time, the share of basic inputs in composite inter-
mediate inputs, ν, is expected to change. Nonetheless, endogenising a production
parameter and linking it to a non-linear variable using a standard S-curve within
a high-dimension system could easily pose a convergence problem. To overcome
this problem, a generalised logistic curve is used to model ν endogenously to the
change in the industrial composition ratio, mt, with the critical parameter on rate of
technological diffusion gradually increased in typical fashion of sensitivity analysis.
The generalised logistic curve is specified as
νt = f(mt) = νm +(νM − νm)
[1 + exp−ζ(mt −mI)]1/υ, νt ≥ νm, (1.69)
where νm, νM ∈ (0, 1) represents the lower and upper bounds (asymptotes) of νt
respectively, ζ is the technological diffusion rate, υ > 0 is the corresponding asymp-
tote value for diffusion, and mI is the inflection point for the industrial composition
ratio. For the purposes of this particular sensitivity analysis, the parameterisations
of νm = 0.1, νM = 0.9, and mI = 0.55 are applied, all of which are reasonable values
for a typical S-curve. The parameter ζ is set at 1.0 to 5.0, which indicates a sensi-
tivity analysis of diffusion rates ranging from 100 to 500 percent, and the parameter
υ is calibrated to maintain initial steady-state values at νt = 0.57, mt = 0.5836, and
Ψt = 0.3672 for the different cases of ζ.
The three composite policy reform programmes are examined again, with steady-
state effects for the key variables of interest presented in Table 1.7. As expected, for
all three of the composite programmes, endogenising νt generates more sensitive re-
sults, and the higher the diffusion rate, ζ considered, the greater the steady-state ef-
81
fects documented. The additional gains amplify the policy complementarity effects.
For example, at the highest ζ value examined (ζ = 5.0), Composite Programme A
would lead νt to decline from 0.57 to 0.496. This would result in an impressive
reduction of −7.8 percentage points in the industrial composition ratio (in compar-
ison, in the benchmark model with fixed ν, m declines by 4.9 percentage points),
and expansion of θS and θS,R by 1.95 and 1.82 percentage points respectively. In
terms of the deepening of domestic innovation expertise, the foreign-domestic in-
novation expertise ratio, Ψ decreases more significantly too despite both θS,R and
nFV having increased. At the same time, the steady-state effect on aggregate pri-
vate consumption growth would be higher too, growing by 0.21 percentage points.
The final output growth rate increases from 4.3 to 4.5 percentage points. These
indicate ‘across-the-board’overall gains, underlying the significance of endogenous
technological change in magnifying the benefits of policy complementarity between
the labour market and FDI-promoting policies. In fact, notwithstanding the fact
that Composite Programme C would come with even more volatility, the model
with endogenous ν and ζ ≥ 2.0 would allow the composite programme to produce
a steady-state increase in aggregate private consumption growth, which has been
the shortcoming of this option when implementing the composite programmes in
the benchmark model. These greater benefits of policy complementarity in a model
where the share of intermediate inputs in production is allowed to change can be
seen in Table 1.8.
1.6 Concluding Remarks
This chapter develops an imitation-innovation model with heterogeneous labour and
foreign MNCs to examine industrial transformation for a developing host economy.
With FDI modelled at the disaggregated level of foreign experts, we formalise a
MNC composition-determination framework that explains Dunning’s ‘internalisa-
82
tion advantage’(1977) as being driven by the presence of asymmetric views on the
productivity of domestic workers. As productivity is a transformation of ability,
the skills acquisition decision and foreign subsidiaries’operational mode choice are
determined along the same ability distribution in the model. These, coupled with
the modelling of an additional asymmetry between Vertical MNCs and other MNCs,
enable the model to be parameterised and analysed to produce policy experiment
results that are consistent with some well-documented stylised facts in the FDI
literature.
We examine the transitional dynamics of various policies. The results show that
the implementation of foreign investment liberalisation measures in a typical de-
veloping host economy is not a matter of straightforward provision of investment
incentives. Indeed, in the presence of asymmetries, our results find that an invest-
ment liberalisation measure that is balanced and targeting all types of foreign firms is
more innovation- and skills acquisition-promoting than disproportionate ones biased
towards selected types of foreign firms. Overall, the results show the importance
of combining human capital and FDI-promoting policies in promoting industrial
transformation, especially if the government of a host economy intends to minimise
disruption of industrial transformation. Furthermore, results from the sensitivity
analysis conducted with endogenous technological change support the conventional
belief that governments of developing economies should strive to undertake measures
in improving the technological diffusion rate within the economy.
By design, the model provides a base framework for future research, notably
a stage-of-development modelling exercise similar to Chen and Funke (2013) that
would allow for post hoc examination of historical development paths of selected de-
veloping economies moving from pure imitation-based to fully industrialised econ-
omy. The key model features (linking of heterogeneous human capital and FDI
along the same ability distribution, with the latter modelled as experts) are novel
contributions. However, there remain limitations that future research can address.
83
For this reasonably complicated high-dimensional model, some policy elements are
not pursued, largely as a self-contained measure to ease computational burden, but
are obviously aspects for extensions. For instance, the role of fiscal policy in the
model is minimal. Second, while the model establishes indirect feedback from the
skills channel to FDI composition, a direct feedback channel of human capital to
FDI is not modelled. For future research, notably in a model with Lucas type of
disembodied human capital and more traditional modelling of FDI as capital, this
would obviously be worth examining.
1.7 Appendix
1.7.1 Estimation of FDI composition data
The data on direct investment and multinational enterprises compiled by the U.S.
Bureau of Economic Analysis (BEA) is used to estimate the FDI composition data
presented in Figure 1.1. The presented figures are therefore reflecting only the
patterns of the operating behaviour of majority-owned nonbank foreign affi liates of
the United States. The BEA dataset is conventionally used in almost all of the
literature for cross-country comparison of FDI behaviour, but it is important to
note that the presented FDI composition data is therefore influenced by factors
such as U.S.’s bilateral trade relationship with a particular host economy, and that
the aggregate direct investment patterns of the U.S. do not necessarily reflect the
direct investment patterns of other major developed economies.
On the dataset, BEA surveys of U.S. direct investment abroad are tabulated
based on reported financial and operating data of foreign affi liates. Unlike most
broad level foreign direct investment datasets based solely on Balance of Payments
statistics, this database focuses not only on annual direct investment position data,
but also data on the activities of multinational enterprises. The BEA maintains an
annual time series, though not all the variables have data recorded for the entire
84
period of 1982-98. Based on BEA’s definition, a foreign affi liate is a foreign business
enterprise in which there is a direct ownership stake, or more specifically, with at
least a ten percent equity stake. A majority-owned affi liate is a foreign business
enterprise in which the U.S. entity would have at least 51 percent equity holdings.
While statistical classification does distinguish between parent companies, majority-
, and minority-owned foreign affi liates, not all statistics are made available. For
instance, there is no research and development (R&D) expenditure and value added
data available for minority-owned foreign affi liates, therefore limiting the scope of
our coverage to only the majority-owned affi liates as it is practically impossible to
know how much of the aggregate amount of minority-owned affi liates’sales belong
to horizontally or vertically-integrated enterprises.
Conceptually, to estimate the different MNC composition from an aggregate
figure reported by country, we first distinguish between the shares of foreign affi liates
with horizontally- and vertically-integrated operations. Most empirical studies in the
literature adopt a sales-based classification approach in the tradition of Horstmann-
Markusen-Venables (HMV), where Horizontal MNC is set up in a host economy to
serve solely the markets of the host economy, while affi liates with Vertical FDI mode
engage in exports. We follow the same tradition and calculate the export-to-value
added ratio (which is a more accurate measure than export-to-sales) for each host
country, and use the ratio calculated for each year to separate the share of Horizontal
FDI from the combined shares of Vertical FDI and Non-mandated FDI.
Based on the theoretical definition of Markusen (1998), in any particular loca-
tion/host economy, the MNC with a vertical operation is the most skilled intensive,
followed by a MNC with a horizontal operation, and lastly plants of MNCs, which in
our context, are non-mandatory investment commitments. This, coupled with our
relatively strict definition for Vertical FDI as consisting of only firms with innovation
at the world frontier, allow us to make a further assumption consistent with the the-
oretical framework of HMV and ours to estimate the shares of Vertical MNCs: Only
85
firms with Vertical FDI mode conduct R&D in the host economy, while for other
modes of operation, R&D is conducted at the headquarters or parents company.
Given the significance of production fragmentation in the Asia and Pacific region
(Athukorola 2005), the ratio of affi liates’R&D expenditure over the aggregate R&D
expenditure spent by parent companies for the entire region is used. Hence, the
shares of Vertical FDI from the earlier residual figures for all vertically integrated
operations is determined by the R&D expenditure spent in the host economy, ad-
justed by the total spent by parent companies for the overall region.
Lastly, consistent with how we define Non-mandated FDI as including techno-
logical licensing, the series for royalties and technological licensing fees received by
U.S. parent companies are added to the estimated Non-mandated FDI amount for
each host economy. Time series on the composition of each FDI type from 1999-2008
are then computed for each host country. These can subsequently be applied to the
FDI statistics on U.S. direct investment position abroad (on a historical cost basis)
compiled directly from Balance of Payments to obtain the estimated time series for
(U.S.) FDI compositions in monetary amount for each destination country.
As seen in Figure 1.1, conditioned on having to use the HMV interpretation, the
estimated FDI compositions do capture the actual heterogeneous industrial structure
of the East Asian economies well (as described in Amsden (2001)). The predomi-
nant shares of Horizontal MNCs in most economies are consistent with the HMV
prediction. Both Singapore and Taiwan have much larger shares of Non-mandated
MNCs than Horizontal MNCs despite their developed status due to their histori-
cal role of being component parts’manufacturing platform. Japan, Singapore, and
South Korea have the largest shares of Vertical FDI, though there does not ap-
pear to be a consistent pattern (and in the case of Japan, Vertical FDI appears
to have declined in recent years). The two unique characteristics associated with
Taiwan and South Korea commonly documented in the technological capabilities
literature– the industrial success of the former is driven by small and medium en-
86
terprises conducting basic, non-cutting edge activities; the latter does not involve
much in the value chains of regional MNCs’ production fragmentation– are also
reflected in the estimated FDI compositions.
1.7.2 Technical Notes
As noted in the text, the dynamic system of the model economy consists of nine
equations (four differential equations, five static equations) that determine the evo-
lutions of mRt , m
It , Q
Rt , and z
Ct .
Domestic Sectors
First, starting from the final good market equilibrium, the expression for private
investment It, (1.52) is substituted into (1.26) to get
Kt = [(1− γη − τ)Yt − LtCat ]− δKt,
or equivalentlyKt
Kt
= [(1− γη − τ)(YtKt
)− zCt ]− δ, (1.70a)
where zCt = LtCat /Kt.
Differentiating zCt with respect to time, we have zCt /z
Ct = n+ (Ca
t /Cat )− Kt/Kt.
This, combined with (1.5), (1.20), and (1.70a), delivers the first-order differential
equation for zCt :
zCtzCt
= n+ [σα− (1− γη − τ)](YtKt
) + zCt − σ(ρ+ δ) + δ. (1.71)
Next, for the imitation sector, substituting (1.11) and (1.54) into (1.12) gives
M It
M It
= (nFH,t)ψI1 [1 + ψI2nFV,t
MRt
M It
]θU,I,t,
or equivalently,
87
M It
M It
= (nFH,t)ψI1 [1 + ψI2nFV,t
MRt
M It
](θU,t − θU,Y,t). (1.72)
Given that mIt/m
It = M I
t /MIt − Kt/Kt, the first-order differential equation for
mIt is obtained by combining (1.72) and (1.70a):
mIt
mIt
= (nFH,t)ψI1 [1 + ψI2nFV,t
mRt
mIt
](θU,t − θU,Y,t)− (1− γη − τ)(YtKt
) + zCt + δ. (1.73)
From the first-order condition (1.13) in the imitation sector, and using (1.11),
wUt = (1
1 + ΛI)(RIt
Lt)(nFH,t)
ψI1 [1 + ψI2nFV,tMR
t
M It
]M It . (1.74)
This is then equated to the first-order condition (1.21) in the final good sector,
wUt = ( βU
1+ΛY)Yt/LU,Y,t to get
YtM I
t
=1 + ΛY
βU(1 + ΛI)RIt θU,Y,t(nFH,t)
ψI1 [1 + ψI2nFV,tMR
t
M It
]. (1.75)
From (1.28)-(1.30), RIt = (1 − η)νγ(Yt/M
It ). Substituting the expression into
(1.75) and replacing Yt/M It out of the equation yields a solution for the proportion
of unskilled labour (of total population) in the production of the final good, θU,Y,t:
θU,Y,t =βU(1 + ΛI)
(1 + ΛY )(1− η)νγ(nFH,t)
−ψI1 [1 + ψI2nFV,tmRt
mIt
]−1. (1.76)
Similarly, for the innovation sector, rewrite (1.14) as
ΦRt = (nFV,t)
ψR1 [1 + ψR2 (mIt
mRt
)]MRt . (1.77)
Then, substitute (1.77) into equation (1.15) to get
MRt
MRt
= (nFV,t)ψR1 [1 + ψR2 (
mIt
mRt
)](θS,t − θS,Y,t). (1.78)
88
Given that mRt /m
Rt = MR
t /MRt − Kt/Kt, the first-order differential equation for
mRt is obtained by combining (1.78) and (1.70a):
mRt
mRt
= (nFV,t)ψR1 [1 + ψR2 (
mIt
mRt
)](θS,t − θS,Y,t)− (1− γη − τ)(YtKt
) + zCt + δ. (1.79)
Substituting (1.78) and (1.28) into the first-order condition (1.16), yields
wSt = (1
1 + ΛR)(QRt
Lt)(nFV,t)
ψR1 [1 + ψR2 (mIt
mRt
)]MRt . (1.80)
This is then equated to the first-order condition (1.21) in the final good sector,
wSt = ( βS
1+ΛY)Yt/LS,Y,t. Rearranging the expression and dividing by Kt yields
YtKt
=(1 + ΛY )
βS(1 + ΛR)QRt θS,Y,t(nFV,t)
ψR1 [1 + ψR2 (mIt
mRt
)]mRt , (1.81)
which can be rearranged to obtain an equation for the proportion of skilled labour
(of total population) in the production of the final good, θS,Y,t:
θS,Y,t =βS(1 + ΛR)
(1 + ΛY )(YtKt
)[QRt (mR
t )]−1(nFV,t)−ψR1 [1 + ψR2 (
mIt
mRt
)]−1. (1.82)
The fourth and last differential equation to derive is for QRt . From (1.20), (1.33),
and (1.34),
QRt
QRt
= [α(YtKt
)− δ]− (1− η)γ(1− ν)(YtKt
)(1
QRt
)(1
mRt
). (1.83)
The static equation for Yt/Kt is given by (1.61), which is derived after applying
the assumptions of βS + βU − %ι = 0 and (γ/η) + α + % = 1:
YtKt
=Θ2
[(θS,Yt )βS(θU,Yt )β
U]−1/(1−γ)
(mI
t )ν(1−η)/η(mR
t )(1−ν)(1−η)/ηγ/(1−γ)
. (1.84)
The remaining two static equations to derive are for θS,t and θU,t. Before that,
89
from the first-order conditions (1.21), the relative wage ratio is given by:
wUtwSt
= (βU
βS)(LS,Y,tLU,Y,t
) = (βU
βS)(θS,Y,tθU,Y,t
). (1.85)
Then, substituting the wage ratio (1.85) into (1.8), the threshold cognitive ability
can be expressed as
at = [βU
βS(1− Γ)
θS,Y,tθU,Y,t
]1/ξ. (1.86)
Substituting this expression for at into both (1.9) and (1.10) would yield
θU,t = 1− aχm[βU
βS(1− Γ)
θS,Y,tθU,Y,t
]−χ/ξ, (1.87)
and
θS,t =χaχmχ− 1
[βU
βS(1− Γ)
θS,Y,tθU,Y,t
](1−χ)/ξ, (1.88)
respectively for θU,t and θS,t.
Foreign Sector
The derivations of key equations in the foreign sector work as follows. Given the
nested objective function, (1.35), even though shadow investment prices are taken as
given by experts, we can similarly write a theoretical aggregate shadow investment
price index, PF , as in Brambilla et al. (2009). This is given by
PF = P 1−θF0 + (
∫ NF
0
[
∫ MIt
0
γσF
1,tP1−σFs,FH,tds+
∫ MRt
0
γσF
2,tP1−σFs,FV,t]
1
1−σF )1−θF ds]dj1
1−θF ,
(1.89)
where P0 is the default baseline price associated with Non-mandated FDI, while the
remaining expressions give various implicit, theoretical shadow investment prices
across different intermediate varieties.
Given the utility specification, a series of theoretical investment demand func-
tions for variety s and ‘productivity requirement’-induced shadow quality j can be
90
derived as
qj,s,t = γσF
κ,t
(yFtP Ft
)(Pj,tP Ft
)−θF (Pj,tPs,t
)σF, (1.90)
where κ = 1 or 2, and P F , Ps, Pj are the host economy-specific theoretical shadow
price indices for the aggregate, variety s, and shadow investment quality j respec-
tively.
Next, by assuming that each firm is small within its group, let firms choose
‘investment prices’to maximise (Ps,t− $s,t)qj,s,t, the standard monopolistic compe-
tition pricing condition of constant mark-up would yield (1.36), which is rewritten
here as
Ps,t =
(σF
σF − 1
)($s,t) . (1.91)
Given (1.35) and (1.89), we substitute (1.91) into (1.90) to yield the theoretical
investment demand of expert with shadow quality (perceived productivity) prefer-
ence j matched to product variety s, as in
qj,s,t =
(σF
σF − 1
)−σFγσ
F
κ,t$−σFj,s,t P
θF−1F,t P σF−θF
j,t yFt . (1.92)
If the ‘market’ for foreign experts’ productivity preference is in equilibrium,
symmetry implies that the average quality preference for the collective pool of foreign
experts of a same variety s would equalise, qj,s,t = qs,t.46
(1.91) and (1.92) allow us to express an evaluative ‘individual value’ function
for a typical foreign expert j in variety s of either Non-mandated, Horizontal, or
Vertical investment mode as
πFP,j,t($j,s,t) =
((σF − 1)σ
F−1yFt
(σF )σF−1
)($FP,j,s,t)
1−σFP θF−1F,t P σF−θF
0 − F0, (1.93)
46Given that the shadow quality difference is driven by perceived heterogeneity among pro-ductivity of domestic workers and therefore implicit in nature, the symmetric equilibrium resultimposed for the collective experts of a same variety pool therefore holds on average for variety s.
91
πFH,j,t($j,s,t) =
((σF − 1)σ
F−1yFt
(σF )σF−1
)γσ
F
1,t ($FH,j,s,t)1−σFP θF−1
F,t P σF−θFj,t − (F0 + F1),
(1.94)
πFV,j,t($j,s,t) =
((σF − 1)σ
F−1yFt
(σF )σF−1
)γσ
F
2,t ($FV,j,s,t)1−σFP θF−1
F,t P σF−θFj,t − (F0 + F2),
(1.95)
respectively, where the value functions of the three types of experts depend on both
the variable components of $l, l = FP, FH,FV , and the fixed cost components of
F0, F1, F2.
As the role of Non-mandated FDI in domestic production is not modelled in this
study, the default baseline investment price, P0 is normalised to one. Further, with
qj,s,t = qs,t derived earlier, in equilibrium, the standard symmetry assumption also
implies that the shadow price index for quality (perceived productivity) j within the
same variety s would be the same for the collective pool of workers matched to variety
s, hence Pj = Ps is imposed. Drawing on the idea of Allanson and Montagna (2005),
we can then use a simplification approach for these indices where a time-invariant
parameter generalising the degree of pricing competition in host economy, Lerner
Index, LI is introduced as a structural parameter in place of Pj = Ps = θF
θF−1$,
where θF is simply written in place of σF in the (1.91) expression for some average
value of $. Substituting in the expression for the average value, $, we can further
express LI as
LI = Pj = Ps =θF
θF − 1
χ
χ− 1
amaFP.amin
, (1.96)
where the definition $ = a/a, the mean ability expression of a Pareto distribution
(z = χzmin/(χ− 1) for some F (z)), and the assumption aFP = a/amin∀t are applied
(the second part of the expression in right-hand side is used for the parameterisation
of aFP ).
Next, define $FP = g(aFP ), $FH = g(aFH), $FV = g(aFV ), where g(a) = a/a,
as the respective cut-off values for each FDI type. Accounting for the asymme-
try of productivity requirement-induced cost (where $s,t = $j,s,t holds for Non-
92
mandated and Horizontal MNCs, and $φs,t = $φ
j,s,t holds for Vertical MNCs), we
set πFP ($FP ) = 0, πFH($FH) = πFP ($FH), and πFH($φFV ) = πFV ($φ
FV ) using
(1.93)-(1.95). The three cut-off productivity values for MNCs’internalisation deci-
sion (also expressed in ability value) in any period t are given by
$FP,t =aFP,ta
=
[F0(
(σF − 1)σF−1yFt /(σF )σF−1
)P θF−1F,t (1)σF−θ
F
]1/(1−σF )
, (1.97)
$FH,t =aFH,ta
=
[F1(
(σF − 1)σF−1yFt /(σF )σF−1
)P θF−1F,t [γσ
F
1,t (LI)σF−θF − (1)σF−θ
F]
]1/(1−σF )
,
(1.98)
$FV,t =aFV,ta
=
[(F2 − F1)(
(σF − 1)σF−1yFt /(σF )σF−1
)P θF−1F,t (LI)σF−θ
F[γσ
F
2,t − γσF
1,t ]
]1/[φ(1−σF )]
.
(1.99)
where (1.96) and the assumption, P0 = 1 are substituted in.
As seen in the expressions for (1.97)-(1.99), the cut-off productivity values are
therefore determined by both the constant cost components of F0, F1, and F2, as
well as other preference paramters in the objective function of the collective foreign
experts.
To calculate the number of foreign firms by FDI type (or share of FDI category),
recall that the sorting of foreign firms follow that of 1/$. We know that the cu-
mulative distribution function of a typical Pareto distribution z, takes the form of
F (z) = 1− (zmin/z)χ for some minimum of z, zmin. Let F (1/$) = F (a/a). Further,
by assuming that there is no exit option for MNCs, we can set aFP = a/amin∀t,
where a/amin denotes some minimum threshold value of entry by foreign firms (a
large value along the ability distribution of host economy). At any time t, the
93
proportion of the three types of foreign firms can be computed as
nFP,t =NFP,t
NF,t
= [F (1/$FH,t)− F (1/$FP,t)] (1.100)
= [F (a/aFH,t)− F (a/aFP,t)]
= 1− (aFH,taFP
)χ − 1 + (aFPaFP
)χ, where aFP = a/amin∀t
= [1− (aFH,taFP
)χ] ,
nFH,t =NFH,t
NF,t
= [F (1/$FV,t)− F (1/$FH,t)] (1.101)
= [F (a/aFV,t)− F (a/aFH,t)]
= [1− (aminaFV,t
a)χ − (1− (
aminaFH,ta
)χ)]
= [(aFH,taFP
)χ − (aFV,taFP
)χ],
nFV,t =NFV,t
NF,t
= [1− F (1/$FV,t)] (1.102)
= [1− F (a/aFV,t)]
= (aFV,taFP
)χ,
where aFP , aFH , aFV give the host economy-specific threshold values of entry for
Non-mandated, Horizontal, and Vertical FDI.
Using (1.97)-(1.99), the variables yFt and PF,t can easily be substituted out.
Specifically, dividing (1.98) by (1.97), and (1.99) by (1.97), the following two thresh-
old conditions can be derived after some straightforward algebraic manipulations:
aFH,t =
[F0
F1
((LI)σF−θF (γ1,t)
σF − 1)
]−1/(1−σF )
aFP , (1.103)
94
and
aFV,t =
[F2 − F1
F0
1
(LI)σF−θF
[γσF
2,t − γσF
1,t ]
]1/[φ(1−σF )]
a1/φFP a
(φ−1)/φ. (1.104)
For the feedback channel of the state of industrial development in a host economy
to the determinant of FDI composition, the evolution of the two foreign preferences,
γ1and γ2 are modelled using a Weibull distribution, governed by a hazard function
of
γ1 = [1− h(γ2;ωk, ωλ)]γ2 (1.105)
= [1− (ωkωλ
(γ2
ωλ)ωk−1)]γ2,
where h(γ2;ωk, ωλ) denotes the ‘out-of-taste’rate of γ2, and ωk and ωλ are the shape
and scale parameter respectively. As γ1 is given by the expected value of E(γ2), we
can rewrite (1.103) and (1.104) as
aFH,t =
[F0
F1
((LI)σF−θF (QF
t −Θ1(QFt )ωk)σ
F − 1)
]−1/(1−σF )
aFP , (1.106)
and
aFV,t =
[F2 − F1
F0
1
(LI)σF−θF
[(QFt )σF − (QF
t −Θ1(QFt )ωk)
σF]
]1/[φ(1−σF )]
a1/φFP a
(φ−1)/φ,
(1.107)
where Θ1 = (ωk/ωλ)(1/ωλ)ωk−1. QF , a measure of the state of industrial develop-
ment of the host economy, is now written in place of the preference parameters. As
not all developing economies host an innovation sector, the foreign source economy
therefore evaluates all host economies for offshore investment by setting QF = mIt
in each period. This yields QFt = wmm
It , where wm is a multiplicative constant.
Lastly, substituting (1.102) into (1.101), and use the two expressions of (1.106)
95
and (1.107), the equation for nFH,t can be derived as
nFH,t = a−χFP (aFH,tχ − nFV,t.aχFP ) (1.108)
=
(aFH,taFP
)χ− nFV,t
=
[F0
F1
((LI)σ−θ(wmmIt −Θ1(wmm
It )ωk)σ
F − 1)
]−χ/(1−σF )
− nFV,t,
with nFV,t given by
nFV,t =(a
1/φFP a
(φ−1)/φ)χ× (1.109)[
F2 − F1
F0
1
(LI)σF−θF
[(wmmIt )σF − (wmmI
t −Θ1(wmmIt )ωk)
σF]
]χ/[φ(1−σF )]
.
The four differential equations (1.71), (1.73), (1.79), and (1.83), and the seven
static equations (1.76), (1.82), (1.84), (1.86), (1.87), (1.108), and (1.109) form the
dynamic system of the model economy.
In the steady state, mIt = mR
t = zCt = QRt = 0, while Yt/Kt , θ
S,Yt , θU,Yt , θUt , θ
St ,
nFH,t, and nFV,t are constant. This implies that the growth rate of final output is
the same as the growth rate of the private capital stock in the steady state, which
in turn means the growth rate of aggregate private consumption is the same.
To estimate the dynamics of output growth rate during the transitions, Kt in
(1.84) is first moved to the RHS to derive an expression for Yt:
Yt =Θ2
[(θS,Yt )βS(θU,Yt )β
U]−1/(1−γ)
(mI
t )ν(1−η)/η(mR
t )(1−ν)(1−η)/ηγ/(1−γ)
Kt. (1.110)
Log-differentiating (1.110) gives
YtYt
=KPt
KPt
+ [γν(1− η)
(1− γ)η]mIt
mIt
+ [γ(1− ν)(1− η)
(1− γ)η]mRt
mRt
+ (βU
1− γ )θU,Yt
θU,Yt+ (
βS
1− γ )θS,Yt
θS,Yt.
(1.111)
Next, substituting time derivative equation of (1.82) into θS,Yt /θS,Yt , we get
96
YtYt
=KPt
KPt
+ [γν(1− η)
(1− γ)η]mIt
mIt
+ [γ(1− ν)(1− η)
(1− γ)η]mRt
mRt
+ (βU
1− γ )θU,Yt
θU,Yt
+(βS
1− γ ) YtYt− KP
t
KPt
− QRt
QRt
− mRt
mRt
− ψR1nFV,tnFV,t
− ψR2
(1 + ψR2mItmRt
)[mIt
mRt
(mIt
mIt
− mRt
mRt
)].
Since we can easily derive nFV,t/nFV,t as
nFV,tnFV,t
=
((−χωkσF )
φ(1− σF )
)mIt
mIt
,
the expression for final output growth is then
YtYt
=KPt
KPt
+
[γν(1− η)
(1− γ)η(1− βS
1− γ −βS(1 + ψR1 )(−χωkσF )
(1− γ)[φ(1− σF )])−1
]mIt
mIt
(1.112)
+(βU
1− γ )(1− βS
1− γ )−1 θU,Yt
θU,Yt+ [γ(1− ν)(1− η)
(1− γ)η
− βS
1− γ ](1− βS
1− γ )−1mRt
mRt
− (βS
1− γ )(1− βS
1− γ )−1 QRt
QRt
−[(βSψR21− γ )(1− βS
1− γ )−1(1 + ψR2mIt
mRt
)−1][mIt
mRt
(mIt
mIt
− mRt
mRt
)].
97
1.8 Tables and Figures
Table 1.1: Calibrated Parameter Values: Benchmark for Host Economy
Parameter Value DescriptionHouseholds
ρ 0.04 Annual discount rateσ 0.27 Elasticity of intertemporal substitutionn 0.0173 Population growth rateξ 0.9 Productivity parameter (effi ciency of skills acquisition)Γ 0.25 Skills acquisition cost (in proportion of skilled wage)χ 2.001 Pareto index, breadth of ability distribution in host economy
Imitation sectorψI1 0.35 Elasticity wrt number of foreign experts in Horizontal modeψI2 −0.3 Externality, Vertical MNCs and innovative blueprintΛI 0.1 Cost mark-up due to labour market distortions
Innovation sectorψR1 0.4 Elasticity wrt number of foreign experts in Vertical modeψR2 9.5 Stepping stone effect, from stock of imitative goodsΛR 0.2 Cost mark-up due to labour market distortions
Final Outputα 0.3 Elasticity with respect to private capitalβU 0.15 Elasticity with respect to unskilled labourβS 0.25 Elasticity with respect to skilled labourγ 0.3 Elasticity wrt composite intermediate inputν 0.57 Share of basic input in composite intermediate input
ΛY 0.05 Cost mark-up due to labour market distortionsδ 0.068 Rate of depreciation, private capital
Intermediate goodsη 0.39 Substitution parameter for production, intermediate goods
Governmentτ 0.25 Effective tax rate on final output
98
Table 1.2: Calibrated Parameter Values: Benchmark for Foreign sector
Parameter Value Description
σF 2.0 Elasticity of foreign preference, between varietiesθF 1.64 Elasticity of foreign preference, across varietiesP0 1.0 Baseline price, Non-mandated FDI’s investmentLI 0.7456 Lerner Index, proxy for pricing competitionF0 0.2733 Basic doing-business cost incurred on foreign expertsF1 0.33 Additional cost incurred on Horizontal MNCF2 0.40 Additional cost incurred on Vertical MNCa 9.55 Constant value linking productivity to abilityφ −1.0 Asymmetric cost parameter, Vertical MNC-specificωk 1.0 Shape parameter, Weibull functionωλ 2.0 Slope parameter, spread of Weibull distributionwm 3.6 Constant, feedback to foreign preference
Table 1.3: Calibrated Parameter Values for Generalised Logistic Curve
Parameter Value Description
νM 0.9 Upper bound for ν (asymptotes)νm 0.1 Lower bound for ν (asymptotes)υ 1.272 Corresponding asymptote value for diffusionζ 1.0 Diffusion ratemI 0.55 Inflection point for industrial composition ratio
99
Table 1.4: Individual Policies: Steady-state Effects
(Absolute deviations from baseline)Benchmark Initial values Γ cut ΛR cut F0 cut F1 cut F2 cut
m 0.5836 -0.0043 -0.0325 -0.0134 -0.0333 0.0560θS 0.2400 0.0069 0.0014 0.0003 0.0038 -0.0036θSR 0.0446 0.0013 0.0072 0.0001 0.0009 -0.0009C/C 0.0430 0.0003 0.0009 -0.0002 -0.0022 0.0022nFV 0.0164 -0.0002 -0.0002 0.0020 0.0023 -0.0052Ψ 0.3672 -0.0145 -0.0553 0.0439 0.0431 -0.1109
Sensitivity Test 1 - ψR1 = 0.8m 0.5836 -0.0031 -0.0276 -0.0159 -0.0414 0.0761θS 0.2400 0.0068 0.0015 0.0000 0.0030 -0.0016θSR 0.0446 0.0012 0.0072 -0.0001 0.0008 -0.0004C/C 0.0430 0.0002 0.0009 0.0002 -0.0017 0.0009nFV 0.0164 -0.0002 -0.0002 0.0016 0.0023 -0.0052Ψ 0.3672 -0.0138 -0.0544 0.0373 0.0438 -0.1117
Sensitivity Test 2 - ψI1 = 0.7m 0.5836 -0.0045 -0.0326 -0.0135 -0.0334 0.0562θS 0.2400 0.0068 0.0014 0.0002 0.0037 -0.0035θSR 0.0446 0.0012 0.0072 0.0001 0.0009 -0.0009C/C 0.0430 0.0003 0.0010 -0.0001 -0.0022 0.0021nFV 0.0164 -0.0001 -0.0001 0.0020 0.0023 -0.0052Ψ 0.3672 -0.0124 -0.0533 0.0447 0.0432 -0.1109
Sensitivity Test 3 - ψI2 = 0.3m 0.5836 -0.0046 -0.0330 -0.0136 -0.0336 0.0568θS 0.2400 0.0068 0.0013 0.0002 0.0037 -0.0034θSR 0.0446 0.0012 0.0072 0.0000 0.0009 -0.0008C/C 0.0430 0.0003 0.0010 -0.0001 -0.0021 0.0020nFV 0.0164 -0.0001 0.0000 0.0021 0.0025 -0.0053Ψ 0.3672 -0.0119 -0.0513 0.0466 0.0468 -0.1153
Sensitivity Test 4 - ψR2 = 15.5m 0.5836 -0.0051 -0.0383 -0.0158 -0.0391 0.0682θS 0.2400 -0.0044 0.0015 0.0003 0.0039 -0.0038θSR 0.0446 0.0013 0.0072 0.0001 0.0010 -0.0009C/C 0.0430 0.0003 0.0009 -0.0002 -0.0022 0.0022nFV 0.0164 -0.0003 -0.0003 0.0019 0.0022 -0.0051Ψ 0.3672 -0.0159 -0.0576 0.0423 0.0413 -0.1095
Relevant parameter values for the benchmark: ψR1 = 0.4, ψI1 = 0.35, ψI2 = −0.3, andψR2 = 9.5.The specific individual shocks considered: Γcut from 0.25 to 0.18; ΛR cut from 0.2to 0.0; F0 cut from 0.2733 to 0.2433; F1 cut from 0.33 to 0.30; and F2 cut from 0.40to 0.37.
100
Table 1.5: Composite Reform Programmes: Steady-state Effects
(Absolute deviations from initial steady-state)Benchmark Initial values Composite A Composite B Composite C
m 0.5836 -0.0489 -0.0048 -0.0830θS 0.2400 0.0092 0.0067 0.0121θSR 0.0446 0.0089 0.0082 0.0097C/C 0.0430 0.0007 0.0022 -0.0010nFV 0.0164 0.0007 -0.0036 0.0039Ψ 0.3672 -0.0477 -0.1256 0.0063
Sensitivity Test 1 - ψR1 = 0.8m 0.5836 -0.0461 0.0115 -0.0886θS 0.2400 0.0088 0.0079 0.0107θSR 0.0446 0.0088 0.0085 0.0093C/C 0.0430 0.0008 0.0014 -0.0003nFV 0.0164 0.0008 -0.0035 0.0040Ψ 0.3672 -0.0456 -0.1252 0.0102
Sensitivity Test 2 - ψI1 = 0.7m 0.5836 -0.0493 -0.0050 -0.0837θS 0.2400 0.0090 0.0066 0.0066θSR 0.0446 0.0088 0.0082 0.0096C/C 0.0430 0.0008 0.0023 -0.0009nFV 0.0164 0.0010 -0.0035 0.0043Ψ 0.3672 -0.0427 -0.1238 0.0149
Sensitivity Test 3 - ψR1 = 0.8, ψI2 = 0.3m 0.5836 -0.0486 0.0116 -0.0938θS 0.2400 0.0084 0.0079 0.0100θSR 0.0446 0.0087 0.0085 0.0091C/C 0.0430 0.0011 0.0014 0.0000nFV 0.0164 0.0011 -0.0035 0.0047Ψ 0.3672 -0.0386 -0.1253 0.0256
Sensitivity Test 4 - ψI1 = 0.7, ψI2 = 0.3m 0.5836 -0.0501 -0.0050 -0.0835θS 0.2400 0.0088 0.0066 0.0115θSR 0.0446 0.0088 0.0082 0.0096C/C 0.0430 0.0009 0.0023 -0.0007nFV 0.0164 0.0012 -0.0035 0.0046Ψ 0.3672 -0.0386 -0.1239 0.0199
Relevant parameter values for benchmark: ψR1 = 0.4, ψI1 = 0.35, ψI2 = −0.3, ψR2 =9.5, η = 0.39, and ωk = 1.0.The composite policy reform programs considered: (i) Composite A - Γcut from0.25 to 0.18, ΛR cut from 0.2 to 0.0, F0 cut from 0.2733 to 0.2433, F1 cut from 0.33to 0.30, and F2 cut from 0.40 to 0.37;(ii) Composite B - Γcut from 0.25 to 0.18, ΛR cut from 0.2 to 0.0, F0 cut from 0.2733to 0.2633, F1 cut from 0.33 to 0.30, and F2 cut from 0.40 to 0.35;(iii) Composite C - Γcut from 0.25 to 0.18, ΛR cut from 0.2 to 0.0, F0 cut from0.2733 to 0.2233, F1 cut from 0.33 to 0.30, and F2 cut from 0.40 to 0.39.
101
Table 1.6: Composite Reform Programmes: Steady-state Effects (continue)
Initial values Composite A Composite B Composite CSensitivity Test 5 - ψR2 = 15.5m 0.5836 -0.0571 -0.0058 -0.0955θS 0.2400 0.0096 0.0068 0.0128θSR 0.0446 0.0090 0.0083 0.0099C/C 0.0430 0.0006 0.0022 -0.0012nFV 0.0164 0.0005 -0.0037 0.0034Ψ 0.3672 -0.0526 -0.1268 -0.0040Sensitivity Test 6 - ψR1 = 0.8, ψR2 = 15.5m 0.5836 -0.0541 0.0141 -0.1006θS 0.2400 0.0092 0.0080 0.0113θSR 0.0446 0.0089 0.0086 0.0095C/C 0.0430 0.0007 0.0014 -0.0005nFV 0.0164 0.0006 -0.0035 0.0035Ψ 0.3672 -0.0493 -0.1259 -0.0001Sensitivity Test 7 - η = 0.54m 0.5836 -0.0303 -0.0066 -0.0481θS 0.2400 0.0092 0.0078 0.0109θSR 0.0446 0.0095 0.0091 0.0099C/C 0.0430 0.0004 0.0012 -0.0006nFV 0.0164 0.0006 -0.0040 0.0042Ψ 0.3672 -0.0533 -0.1374 0.0098Sensitivity Test 8 - ωk = 1.2m 0.5836 -0.0483 -0.0083 -0.0798θS 0.2400 0.0094 0.0000 0.0127θSR 0.0446 0.0090 0.0080 0.0099C/C 0.0430 0.0005 0.0026 -0.0014nFV 0.0164 0.0003 -0.0028 0.0025Ψ 0.3672 -0.0557 -0.1085 -0.0200
102
Table 1.7: Sensitivity Analysis: Endogenous Technological Change with CompositeReform Programmes: Steady-state Effects
(Absolute deviations from initial steady-state)Initial values Composite A Composite B Composite C
100% diffusion rate, ζ = 1.0m 0.5836 -0.0535 -0.0054 -0.0902θS 0.2400 0.0105 0.0068 0.0143θSR 0.0446 0.0101 0.0083 0.0118C/C 0.0430 0.0009 0.0022 -0.0007
Ψ 0.3672 -0.0566 -0.1262 -0.0112ν 0.5700 -0.0097 -0.0010 -0.0164
200% diffusion rate, ζ = 2.0m 0.5836 -0.0585 -0.0060 -0.0978θS 0.2400 0.0121 0.0070 0.0169θSR 0.0446 0.0116 0.0085 0.0142C/C 0.0430 0.0011 0.0022 -0.0003
Ψ 0.3672 -0.0670 -0.1271 -0.0310ν 0.5700 -0.0215 -0.0021 -0.0361
300% diffusion rate, ζ = 3.0m 0.5836 -0.0643 -0.0067 -0.1059θS 0.2400 0.0141 0.0072 0.0200θSR 0.0446 0.0134 0.0086 0.0171C/C 0.0430 0.0013 0.0023 0.0002
Ψ 0.3672 -0.0791 -0.1280 -0.0528ν 0.5700 -0.0358 -0.0035 -0.0592
400% diffusion rate, ζ = 4.0m 0.5836 -0.0709 -0.0076 -0.1141θS 0.2400 0.0165 0.0075 0.0238θSR 0.0446 0.0155 0.0089 0.0205C/C 0.0430 0.0017 0.0023 0.0008
Ψ 0.3672 -0.0931 -0.1294 -0.0764ν 0.5700 -0.0532 -0.0055 -0.0859
500% diffusion rate, ζ = 5.0m 0.5836 -0.0780 -0.0087 -0.1217θS 0.2400 0.0195 0.0078 0.0280θSR 0.0446 0.0182 0.0092 0.0243C/C 0.0430 0.0021 0.0023 0.0015
Ψ 0.3672 -0.1090 -0.1309 -0.1010ν 0.5700 -0.0739 -0.0080 -0.1153
103
Table 1.8: Policy Complementarities: Comparison across Composite Programme A
(Absolute deviation from initial steady-state)m θS θSR C/C Ψ
Sum of Parts:Γcut -0.0043 0.0069 0.0013 0.0003 -0.0145ΛR cut -0.0325 0.0014 0.0072 0.0009 -0.0553F0 cut -0.0134 0.0003 0.0001 -0.0002 0.0439F1 cut -0.0333 0.0038 0.0009 -0.0022 0.0431F2 cut 0.0560 -0.0036 -0.0009 0.0022 -0.1109Aggregate effects -0.0275 0.0087 0.0086 0.0011 -0.0937
Composite A (fixed ν) -0.0489 0.0092 0.0089 0.0007 -0.0477
Composite A (endogenous ν)- ζ = 1.0 -0.0535 0.0105 0.0101 0.0009 -0.0566- ζ = 2.0 -0.0585 0.0121 0.0116 0.0011 -0.0670- ζ = 3.0 -0.0643 0.0141 0.0134 0.0013 -0.0791- ζ = 4.0 -0.0709 0.0165 0.0155 0.0017 -0.0931- ζ = 5.0 -0.0780 0.0195 0.0182 0.0021 -0.1090
The specific individual policy shocks considered: Γcut from 0.25 to 0.18; ΛR cutfrom 0.2 to 0.0; F0 cut from 0.2733 to 0.2433; F1 cut from 0.33 to 0.30; and F2 cutfrom 0.40 to 0.37.
104
Figure 1.1: Estimated FDI Composition fromU.S. to selected East Asian Economies,1999-2008
105
Figure 1.2: Production and Labour Allocation in Host Economy
Figure 1.3: Foreign Sector
106
Figure 1.4: Policy Experiment for Skills Acquisition Cost Cut
0.06
0.05
0.04
0.03
0.02
0.01
0
0.01
0.02
Permanent Cut in from 0.25 to 0.18(Absolute deviations from baseline)
Time
Effective skilled labour share
Time
ForeignDomestic Innovation Expertise Ratio
0
0.002
0.004
0.006
0.008
0.01
0.012
0.014
0
0.001
0.002
0.003
0.004
0.005
0.12
0.1
0.08
0.06
0.04
0.02
0
0.02
Industrial composition ratio
Effective skilled labour in innovation
Time Time
Y = 0.9
Y = 0.9 Y = 0. 99Y = 0. 8Baseline
u
@
107
Figure 1.5: Policy Experiment for Labour Hiring Cost-mark up Reduction in theInnovation Sector
0.05
0.04
0.03
0.02
0.01
0
0.01
Permanent Cut in from 0.2 to 0.0(Absolute deviations from baseline)
Time
Effective skilled labour share
Time
ForeignDomestic Innovation Expertise Ratio
0.001
0
0.001
0.002
0.003
0.004
0.005
0.001
0
0.001
0.002
0.003
0.004
0.005
0.006
0.007
0.008
0.009
0.12
0.1
0.08
0.06
0.04
0.02
0
0.02
Industrial composition ratio
Effective skilled labour in innovation
Time Time
Baseline
CR
CY = 0.05, f2R = 15.5CY = 0.0, f2
R = 9. 5CY = 0.05, f2R = 9.5
108
Figure 1.6: Policy Experiment for Investment Incentive targeted only at VerticalMultinationals
0.01
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
Permanent Cut of F2 by 0.03(Absolute deviations from baseline)
Time
Effective skilled labour share
Time
ForeignDomestic Innovation Expertise Ratio
0.004
0.003
0.002
0.001
0
0.001
0.001
0.0008
0.0006
0.0004
0.0002
0
0.0002
0.2
0.15
0.1
0.05
0
0.05
Industrial composition ratio
Effective skilled labour in innovation
Time Time
Baseline f1R = 0.4, f2
I = 0.3f1R = 0.4, f2
I = ?0.3 f1R = 0.8, f2
R = ?0.3
109
Figure 1.7: Policy Experiment for Investment Incentive targeted only at HorizontalMultinationals
0.06
0.05
0.04
0.03
0.02
0.01
0
0.01
Permanent Cut of F1 by 0.03(Absolute deviations from baseline)
Time
Effective skilled labour share
Time
ForeignDomestic Innovation Expertise Ratio
0.001
0
0.001
0.002
0.003
0.004
0.0002
0
0.0002
0.0004
0.0006
0.0008
0.001
0.05
0
0.05
0.1
0.15
0.2
Industrial composition ratio
Effective skilled labour in innovation
Time Time
Baseline f1I = 0.35, f2
I = 0.3f1I = 0. 35, f2
I = ?0. 3 f1I = 0.70, f2
I = ?0. 3
110
Figure 1.8: Policy Experiment for economy-wide Investment Liberalisation for AllForeign Multinationals
0.02
0.015
0.01
0.005
0
0.005
Permanent Cut of F0 by 0.03(Absolute deviations from baseline)
Time
Effective skilled labour share
Time
ForeignDomestic Innovation Expertise Ratio
0.0002
0.0001
0
0.0001
0.0002
0.0003
0.00015
0.0001
5E5
0
5E5
0.0001
0.01
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
Industrial composition ratio
Effective skilled labour in innovation
Time Time
Baseline f1I = 0.35, f2
I = 0.3f1I = 0. 35, f2
I = ?0. 3 f1I = 0.70, f2
I = ?0. 3
111
Figure 1.9: Policy Experiments for Composite Policy Reform Programmes
0.15
0.1
0.05
0
0.05
Benchmark Calibration(Absolute deviations from baseline)
Time
Effective skilled labour share
Time
ForeignDomestic Innovation Expertise Ratio
0
0.005
0.01
0.015
0.02
0
0.005
0.01
0.015
0.25
0.2
0.15
0.1
0.05
0
0.05
0.1
0.15
Industrial composition ratio
Effective skilled labour in innovation
Time Time
Composite A
Composite B
Composite C
Composite A
Composite B
Composite C
Composite A
Composite A
Composite B
Composite C
Composite B
Composite C
112
Figure 1.10: Industrial Composition Ratio - Composite Policy Reform ProgrammeA (Absolute deviation from baseline)
Note: ψI2 is the elasticity of imitative blueprint with respect to the cross-term offoreign innovation expertsand stock of innovative blueprint, and ψR2 is the elasticity of innovative blueprintwith respect toimitative-to-innovation blueprint ratio measuring the stepping stone effect
113
Chapter 2
Unemployment, Growth andWelfare Effects of Labour MarketReforms
2.1 Introduction
The impact of labour market reforms on unemployment and economic growth has
been the focus of a large theoretical and empirical literature. From an analytical
perspective, important issues in this context are the modelling of the production
structure and the causes of mismatches between supply and demand in the labour
market. Accounting for innovation activities for instance, is critical to study the
role of human capital accumulation, knowledge externalities, and the distribution
of skills as sources of growth and employment; and the modelling of labour market
rigidities is essential to explain unemployment. These rigidities have taken the form
of government legislation on minimum wages, mandated firing costs, unemployment
benefits, collective bargaining (Daveri and Tabellini (2000), Varga et al. (2014),
Bhattacharyya and Gupta (2015) and Chang and Hung (2016)), search and matching
frictions in the Mortensen-Pissarides tradition (Zagler (2009) and Cacciatore and
Fiori (2016)), and effi ciency wages (van Schaik and de Groot (2000), Meckl (2001,
2004), Bucci et al. (2003), Parello (2011), and Zagler (2011)).1 A key result from
1Some of these contributions also account for the existence of an innovation sector, albeit (asdiscussed next) in a partial manner.
114
the literature is that the relationship between growth and unemployment may be
weak, both in the short run and in the long run.
However, the existing literature suffers from four major shortcomings. First,
except for a few contributions– such as Cacciatore and Fiori (2016), albeit in a
business cycle setting– most of the literature neglects transitional dynamics. As a
result, the dynamic tradeoffs that may be associated with labour market reforms,
that is, the possibility of conflicting effects in the short and the longer run in terms of
their impact on either unemployment or growth specifically, cannot be ascertained.
Second, almost none of the existing models considers the supply side of the labour
market. In particular, the distribution of the labour force across levels of education,
and how it changes over time, are seldom explicitly analysed.2 This creates a major
diffi culty in terms of understanding how the labour market adjusts in response to
shocks, how it interacts with the process of economic growth, and how public policy
can affect unemployment and its composition. Third, only a few contributions
(including again, Cacciatore and Fiori (2016)) study the impact of labour market
reforms on welfare and the possibility that growth and welfare effects may move in
opposite directions. This may help to understand (organised) resistance to reform.
Moreover, these conflicting effects may also have a temporal dimension, which can be
studied only if transitional dynamics are accounted for. Finally, there have been few
attempts to assess quantitatively– in terms of unemployment, growth, or welfare–
the benefits of a simultaneous implementation of labour market reforms, compared
to a piecemeal approach, and the scope for exploiting policy externalities to mitigate
the welfare cost of reforms. This matters because the impact of a specific policy may
depend on whether other policies are implemented at the same time. Ignoring policy
externalities is a potential source of bias.
2Some models introduce a work-leisure trade-off into workers’ utility functions (thereby ac-counting for the intensive margin of labour supply), but the distribution of the labour force acrossskills (the extensive margin) is kept constant. Other contributions do introduce disembodied humancapital in the Lucas-Uzawa tradition, but also fail to account for the heterogeneous distribution ofskills in the labour force.
115
The purpose of this chapter is to address all of these issues, using an OLG en-
dogenous growth model with a heterogeneous labour force, final good and innovation
sectors, labour market rigidities, and structural unemployment. To model wage for-
mation in final good production, where activity involves more routine tasks and
effort is fully observable, trade unions are introduced; but to model wage formation
in the innovation sector, an effi ciency wage specification is adopted. This approach,
as argued elsewhere in the literature, is better suited than standard search models
of the Mortensen-Pissarides type to understand the link between wages and pro-
ductivity in innovation activities. Indeed, in these activities, firms cannot monitor
researchers’effort perfectly; the key issue for an employer is thus to mitigate in-
centives to shirk and encourage creativity. A natural approach is thus to use an
effi ciency wage framework, in this case linking effort and wages. As a result, persis-
tent uncompetitive wage differentials for highly-skilled workers may emerge across
sectors.3
While the balanced growth path is solved, the complexity of the model pre-
cludes a full analytical characterisation of its dynamic properties. The model is
therefore parameterised to perform an extensive range of quantitative simulations.
Importantly, this parameterisation is performed for two sets of countries that are
characterised by a range of labour market rigidities (including high minimum wages
and active trade unions) and have recorded high structural unemployment rates in
recent years: a group of high-income European countries and a group of middle-
income Latin American countries. Numerical policy experiments therefore allow for
the systematic comparison of the impact of labour market reforms in two signifi-
cantly different economic environments. The impacts of these reforms are not only
assessed on unemployment, growth, and welfare, but also on the misallocation of
3It can be argued that, in the presence of globalisation, there may be an eroding case for the useof effi ciency wage framework. While this may be true for the production of routine task, a Shapiro-Stiglitz effi ciency wage framework remains useful in modelling the research-based innovation sector,where the efforts of workers remain largely unobservable and the monitoring intensity is mainly alocalised issue.
116
talent, a situation where individuals with abilities that are high enough to operate
in the innovation sector end up instead performing routine production tasks. In an
innovation-driven economy this is costly for society as a whole, even though these
individuals are (like everybody else) utility maximisers.
In addition to evaluating the effects of single policy experiments, composite pro-
grammes are considered and examined so to understand the extent to which policy
externalities may mitigate the adverse effects of individual reforms. The cases where
composite reform programmes are combined with skills expansion, as well as an in-
crease in public investment in infrastructure, are also considered. Such investments
have been advocated in a number of developed and developing countries in the after-
math of the global financial crisis– not only as a short-term Keynesian response due
to their demand-side effects, but also as a fundamental step to improve productivity
due to their supply-side effects (see for instance, LSE Growth Commission (2013)
and International Monetary Fund (2016)).
To preview the results, we find that labour market reforms entail a two-way
causality between growth and unemployment: growth tends to lower unemploy-
ment, through its impact on labour demand; but unemployment may lower growth
because it reduces (through its wage signalling effects) incentives to acquire skills
and constrains the ability to expand innovation activities– —a key engine of growth.
Individual labour market reforms may generate a weak correlation between growth
and unemployment, as predicted in a number of existing studies; in addition, they
may have conflicting effects on growth and welfare in the long run. To some extent,
this tradeoff can be tempered by exploiting policy externalities. But to avoid cre-
ating an oversupply of specialised workers, governments must refrain from adopting
policies (such as drastic reductions in effective tuition fees) that contribute to gen-
erating large numbers of university graduates; improving the quality of education
may prove more effective.
In addition, public investment in infrastructure may help to boost employment
117
and mitigate the oversupply problem, partly by promoting innovation activities.
Finally, a comparison of the sum of the long-run effects in terms of growth, un-
employment and welfare of each individual policy in a composite programme with
those associated with the same composite programme suggests that, if unemploy-
ment or social welfare matters more than growth to policymakers, comprehensive
reform programmes may generate negative externalities. With limited political cap-
ital, overly ambitious labour market reform programmes may therefore be costly
and ineffective.
The remainder of the chapter is organised as follows. Section 2.2 presents the
model. Section 2.3 defines the balanced growth equilibrium and Section 2.4 char-
acterises its properties. Section 2.5 describes the parameterisation of the model for
“typical”high- and middle-income countries with distorted labour markets and high
unemployment. Section 2.6 considers a variety of individual labour market policies
(including a reduction in the minimum wage and a reduction in unemployment ben-
efit rates), as well as policies aimed at promoting acquisition of skills. Section
2.7 considers composite reform programmes involving a combination of these poli-
cies, with and without increases in public investment on infrastructure. Section 2.8
provides a sensitivity analysis with respect to all experiments, and the results ob-
tained are quantitatively and qualitatively robust to a significant range of parameter
changes. The final section provides some concluding remarks.
2.2 The Model
The economy that we consider is populated by individuals with different innate
abilities, who live for two periods, adulthood and old age. Population is constant
at N . Each individual is endowed with one unit of time in each period of life. In
old age, time is allocated entirely to leisure. There are four production sectors: a
manufacturing sector, which produces a homogeneous final good with routine tasks,
118
an intermediate goods sector, an innovation sector, which creates designs used for
producing intermediate goods, and an education sector, which allows individuals to
acquire advanced training. The final good is produced by combining both private
and public inputs, and is used for consumption, private and public investment, and
the production of intermediate goods. The public input consists of infrastructure
and is provided free of charge. However, it is subject to congestion. Production in
the innovation sector combines public and private inputs as well, but workers’effort
is not observable.
Firms in the final good and innovation sectors are perfectly competitive whereas
those in the intermediate goods sector are monopolistically competitive, producing
(as in Romer (1990)) differentiated varieties of goods. The total number of blueprints
existing at a certain point of time coincides with the number of intermediate input
varieties available, and represents the stock of (nonrival) knowledge.
Two categories of labour are available, untrained (with only basic education) and
specialised (with advanced education).4 Workers are born untrained and must decide
at the beginning of adulthood whether or not to become specialised. Acquiring
advanced education requires both time and pecuniary costs. While all specialised
workers can work in the final good sector, only those with the highest ability can work
in the innovation sector, as for instance, in Böhm et al. (2015). Rigidities prevail in
all segments of the labour market and unemployment emerges in equilibrium.
2.2.1 Individuals
Individuals have identical preferences but are born with different abilities, indexed
by a. Ability is instantly observable by all and follows a continuous distribution
with density function f(a) and cumulative distribution function F (a), with support
(0, 1). For tractability, a is assumed to be uniformly distributed on its support.
Each individual maximises utility and decides whether to engage in market work as4Formally there are only two periods in the model, but implicitly there is a first period where
basic education is acquired.
119
an untrained worker or (after training) as a specialised worker.
Specifically, an adult with ability a can enter the labour force at the beginning
of period t as an untrained worker and earn the wage wUt , which is independent
of the worker’s ability. Alternatively, the individual may choose to first spend a
fraction ε ∈ (0, 1) of his/her time endowment at the beginning of adulthood in higher
education, incurring a cost tct > 0, and then enter the labour force for the remainder
of the period as a specialised worker, earning either the wage wSYt if employed in
the final good sector, or wSRt if employed in the innovation sector. During training,
workers earn no income. All individuals can either be employed (superscript E) or
unemployed (superscript L). If employed, an untrained individual can work only
in the final good sector. All specialised individuals can work in that sector as well,
but only those with the highest level of ability, a > aR, can potentially work in the
innovation sector. The threshold ability level aR is taken to be constant, consistent
with the assumption that, for any given population, the spread of individuals along
the ability continuum is largely determined by nature.5 If unemployed, individuals
earn an unemployment benefit, bht , h = U, S, which is not taxable.
Let ch,jt|t+n denote consumption at period t + n of an individual h = U, SY, SR,
either employed or unemployed, j = E,L, born at the beginning of period t, with
n = 0, 1. The individual’s discounted utility function is given by
V h,jt = ηC ln ch,jt|t +
ln ch,jt|t+1
1 + ρ, h = U, SY, SR, j = E,L (2.1)
where ρ, ηC > 0 are the common discount rate and preference parameter, respec-
tively.6
5Hypotheses such as the Flynn effect in the psychological science literature do suggest that IQscores tend to improve as the share of the skilled population grows (see Flynn (2007)). However,this remains a contentious subject of research and in the absence of conclusive evidence we treataR as fixed.
6Because leisure does not enter the utility function, the opportunity cost of unemployment issimply the wage foregone. Another specification that can be considered is to allow for differentpreference for leisure between the specialised and the untrained workers, where the former is allowedto value leisure time. However, given the already-complicated model, this is not explored in thecontext of this chapter.
120
The period-specific budget constraints are given by
cU,jt|t + sUjt =
(1− τ)wUt
bUt
if j = Y
if j = L, (2.2)
ch,jt|t + sht =
(1− τ)(1− ε)wht − tct
(1− ε)bSt − tct
if j = E, h = SY, SR
if j = L(2.3)
ch,jt|t+1 = (1 + rt+1)sht , h = U, SY, SR, j = E,L (2.4)
where sh,jt is savings, 1 + rt+1 the gross rate of return between periods t and t + 1,
and τ ∈ (0, 1) the tax rate.
An individual finds it optimal to train if and only if his expected earnings as a
specialised worker, adjusted for the time and pecuniary costs of training, exceeds
the expected earnings of an untrained worker:
(1−ε)[ζSYt (1−τ)wSYt +ζSRt (1−τ)wSRt +ζSLt bSt ]−tct ≥ ζUYt (1−τ)wUt +ζULt bUt , (2.5)
where the going wage, or the unemployment benefit, is weighted by the respec-
tive probability of being either employed or unemployed, ζht ∈ (0, 1), for h =
SY, SR, SL, UY, UL.7 In specifying (2.5), we assume for simplicity that an indi-
vidual knows if his/her ability is above or below the threshold aC and can therefore
decide whether to acquire specialised skills or not at the beginning of adulthood,
but finds out whether his/her ability is at or above aR > aC only after undergo-
ing training. Put differently, this specification captures the idea that an individual
discovers whether he/she is “super smart”only upon college graduation– a sensible
assumption in practice.8
7Equation (2.5) is assumed to hold as a strict inequality for the individual with the highestability, that is, a = 1, otherwise nobody would choose to become specialised. Also, as will becomeclearer later, for the probabilities, ζUYt + ζULt = 1, and ζSYt + ζSRt + ζSLt = 1.
8Without this assumption two separate conditions, one for those with a > aR (which wouldtake the form shown in (2.5), given that these individuals can work anywhere) and one for thosewith a < aR (which would exclude the wage in the innovation sector in calculating the expected
121
The training cost is proportional to the expected specialised wage when employed
and varies inversely with the individual’s ability, which determines how fast (or how
well) he or she can learn:
tct = µ(1− ε)(1− τ)(ζSYt wSYt + ζSRt wSRt )/aχ, (2.6)
with µ, χ ∈ (0, 1). The assumption on the productivity parameter χ ensures that
the effect of ability on training costs is subject to diminishing returns.
As shown in the Appendix for this chapter, the threshold level of ability aCt such
that all individuals with ability higher than aCt choose to undergo training is given
by
aCt = µ1/χ
1− (1− ζULt )(1− τ)wUt + ζULt bUt − (1− ε)ζSLt bSt
(1− τ)(1− ε)(ζSYt wSYt + ζSRt wSRt )
−1/χ
. (2.7)
This equation plays an important role in understanding the dynamics of the
labour market; it shows that labour market outcomes (which are partly influenced
by public policy) have a direct impact on the decision to acquire training, through
their effect on expected, rather than actual, wages.
The productivity of untrained workers is constant regardless of ability and is
normalised to unity. Given (2.7), the raw supply of untrained labour, NUt , is equal
to the number of individuals in the population who choose not to undergo training:
NUt = N
∫ aCt
0
f(a)da = aCt N. (2.8)
The raw supply of specialised workers with ability a ∈ (aCt , aR) is N
∫ aRaCt
f(a)da =
(aR−aCt )N . However, the average productivity of these workers equals (aCt +aR)/2;
specialised wage) would be required. This would complicate significantly the analysis, withoutadding much additional insight.
122
thus, the effective supply of specialised labour with a ∈ (aCt , aR) can be defined as
(aR − aCt )(aCt + aR)
2N =
(aR)2 − (aCt )2
2N. (2.9)
As noted earlier, among specialised workers, only those with ability a ∈ (aR, 1)
can operate in the innovation sector; thus, the (effective) supply of labour to that
segment of the market, NRt , is
NRt =
(1− aR)(aR + 1)
2N =
1− (aR)2
2N. (2.10)
Adding (2.9) and (2.10), the total (effective) supply of specialised workers, NSt ,
is
NSt =
1− (aCt )2
2N. (2.11)
However, workers with the highest ability are also able to work in the final good
sector, at the same wage as other specialised workers there. Assuming that all work-
ers with ability greater than aR seek employment in innovation activities first, the
supply of specialised labour to manufacturing is not given by NSt −NR
t , but rather
by NSt − NSR
t , where NSRt ≤ NR
t is the actual (demand-determined) level of em-
ployment in the innovation sector. Hence, while NSRt is determined by the labour
demand from firms in the innovation sector (specifically, the first-order conditions
(2.27) and (2.28)), NRt is therefore primarily determined by nature. To the extent
that NRt > NSR
t , there is misallocation of talent, in the sense that individuals with
abilities that are high enough to operate in the innovation sector may end up per-
forming routine tasks in manufacturing. In our numerical experiments, we measure
talent misallocation by the share of “overqualified”workers in the final good sector,
defined as max[0, (NRt −NSR
t )/NSYt ], where NSY
t is the actual employment in that
sector.
123
2.2.2 Final Good
final good production by firm i, Y it , requires the use of specialised labour, N
SYi,t ,
untrained labour, NUYi,t , private capital, K
Pi,t, aggregate public capital, K
Gt , and the
combination of intermediate inputs, xi,s,t, with s ∈ (0,Mt).
The production function is specified as
Y it = [
KGt
(KPt )ζK N ζN
]ω[(1− ε)NSYi,t ]β
S
(NUYi,t )β
U
(KPi,t)
α[
∫ Mt
0
xηi,s,tds]γ/η, (2.12)
where βS, βU , α, γ ∈ (0, 1), ω > 0, ζK , ζN > 0, γ = 1 − (βS + βU) − α, η ∈ (0, 1)
and 1/(1− η) > 1 is (the absolute value of) the price elasticity of demand for each
intermediate good, and KPt aggregate private capital. Constant returns therefore
prevail with respect to private inputs, and public capital is subject to congestion,
measured by aggregate private capital and population.
Assuming full depreciation, firm i’s profits are defined as
ΠYi,t = Y i
t −∫ Mt
0
P st xi,s,tds− (1 + ς t)[w
SYt (1− ε)NSY
i,t + wUt NUYi,t ]− rtKP
i,t,
where ς t > 0 is the firm’s contribution rate to the unemployment insurance scheme,
based on its total wage bill.
Each firm maximises profits subject to (2.12) with respect to labour, private
capital, and quantities of intermediate goods xi,s,t, ∀s, taking factor prices and Mt
as given. This yields, in standard fashion,
wSYt = (βS
1 + ς t)
Yi,t(1− ε)NSY
i,t
, wUt = (βU
1 + ς t)Yi,tNUYi,t
, (2.13)
rt = α(Yi,tKPi,t
), (2.14)
xi,s,t = (γZi,tP st
)1/(1−η), s = 1, ...Mt, (2.15)
124
Zi,t = Yi,t/
∫ Mt
0
(xi,s,t)ηds. (2.16)
2.2.3 Intermediate Goods
As in Romer (1990), intermediate goods firms produce inputs based on blueprints
produced by the innovation sector. Each firm produces one, and only one, horizontally-
differentiated good, using the same technology used to produce the final good. Pro-
duction of each unit of intermediate goods costs one unit of final output. Similar to
Chapter 1, the intermediate goods sector provides an intermediary channel where
the prices of each variety of intermediate goods are set, as well as the transformation
of knowledge stock into more production-relevant forms.
Each producer must purchase a patented design from the innovation sector. Once
the patent fee Qt is paid, each producer sets its price to maximise profits, given the
perceived demand function for its good (2.15), which determines marginal revenue.
Under a symmetric equilibrium, profits are given by ΠIt = (Pt−1)xt or, using (2.15)
and (2.16), ΠIt = (Pt − 1)[γYt/PtMtx
ηt ]
1/(1−η). In standard fashion, the solution
yields the optimal price as
P st =
1
η. ∀s = 1, ...Mt (2.17)
Using (2.15), the quantity demanded at this price is xs,t = (γηZt)1/(1−η), ∀s, that
is, noting that under symmetry∫Mt
0xηs,tds = Mtx
ηt ,
xt = γη(YtMt
), (2.18)
with maximum profit given by
ΠIt = (1− η)γ(
YtMt
). (2.19)
Intermediate-input producing firms last only one period, and patents are auc-
125
tioned off randomly to a new group of firms in each period. Thus, each firm holds
a patent only for the period during which it is bought, implying monopoly profits
during that period only; yet patents last forever. By arbitrage, therefore,
Qt = ΠIt . (2.20)
2.2.4 Innovation Sector
Firms in the innovation sector use only high-ability specialised labour, in quantity
(1− ε)NSRt . There is no aggregate uncertainty and the production technology is
Mt+1 −Mt = ARt [eRt (1− ε)NSR
t
N]λ, (2.21)
where eRt is the level of effort and ARt productivity, which depends on access to
public infrastructure and, consistent with the standing-on-shoulder effect (see Jones
(2005)), the stock of knowledge:
ARt = (kGt )φR1Mt, (2.22)
with kGt = KGt /K
Pt and φR1 > 0. Thus, in terms of effi ciency units of labour,
effort and workers are perfect substitutes. Because of duplication effects there are
diminishing marginal returns to labour, so that λ ∈ (0, 1).9 Access to public capital
is subject to (proportional) congestion, measured by private capital. In addition, to
eliminate scale effects, as in Dinopoulos and Segerstrom (1999) innovation diffi culty
is measured in terms of population size.
Effort is modelled following the simple specification developed in Agénor and
Aizenman (1999). In deciding how much effort to provide at t, researchers evaluate
a period utility function, UR(wSRt , 1 − eRt ), which depends on the after-tax wage
9See Gancia and Zilibotti (2005) for a discussion. Empirical estimates of λ are discussed later.
126
earned, (1− τ)wSRt , and the disutility of effort, 1− eRt :
UR[(1− τ)wSRt , 1− eRt ] = ln[((1− τ)wSRt )δR(1− eRt )1−δR ], (2.23)
where δR ∈ (0, 1).10 Let π denotes the probability that a researcher is caught shirk-
ing, in which case he is fired and ends up being either employed in manufacturing,
at the going wage wSYt , or unemployed, collecting the benefit bSt . In line with the
standard Shapiro-Stiglitz shirking model, we assume that it is related one-to-one
with the intensity with which firms in the innovation sector choose to monitor their
workers. Thus, although given at the level of each individual researcher, π (or,
equivalently here, monitoring intensity) is in principle a choice variable at the level
of the firm, which would normally vary inversely with unit monitoring costs. In turn,
these costs may depend on both firm-specific characteristics (the required number of
supervisors for particular tasks, for instance) and sector- or economy-wide factors.
The level of effort provided is either eRt , when employed and not shirking, or the
minimum eRm ∈ (0, 1), when shirking while employed. The optimal level of effort is
such that the utility derived from working without shirking (as given by (2.23)) is
at least equal to the expected utility of shirking:
UR[(1− τ)wSRt , 1− eRt ] ≥ π ln[(ζSYt (1− τ)wSYt + ζSLt bSt )δR(1− eRm)1−δR ] (2.24)
+(1− π) ln[((1− τ)wSRt )δR(1− eRm)1−δR ],
where the latter is defined as a weighted average of the expected income earned if
caught shirking and fired with probability π (either working at the alternative wage
wSYt , with probability ζSYt , or unemployed, with probability ζSLt , and earning the
benefit bSt ) and if not caught with probability 1− π (earning the going wage wSRt ).
10While at the household level, agents face no disutility to work, at the individual level spe-cialised workers in the innovation sector does have a choice in determining their optimal level ofresearch efforts.
127
In either case, for simplicity, the worker provides the minimum effort level eRm.
In equilibrium, workers are indifferent between shirking and not shirking; condi-
tion (2.24) therefore holds with equality and can be solved to give
eRt = 1− (1− eRm)(ζSYt (1− τ)wSYt + ζSLt bSt
(1− τ)wSRt)ψ, (2.25)
with ψ = πδR/(1 − δR). Thus, an increase in the expected wage in the innovation
sector relative to its opportunity cost raises the level of effort. For a given wage
ratio, an increase in the probability of getting caught shirking (a rise in π) raises
also the level of effort.11
Using (2.21), and taking the patent fee and productivity as given, the firm’s
problem is to maximise profits by setting both wages and employment:
maxNSRt ,wSRt
ΠRt = QtA
Rt [eRt (1− ε)NSR
t
N]λ − (1 + ς t)w
SRt (1− ε)NSR
t , (2.26)
subject to (2.25). The first-order conditions are given by
λ(NSRt )λ−1(eRt )λ(1− ε)λQtA
Rt
Nλ= (1 + ς t)(1− ε)wSRt , (2.27)
λ(eRt )λ−1QtARt
Nλ[(1− ε)NSR
t ]λψ(1− eRt )
wSRt= (1 + ς t)(1− ε)NSR
t . (2.28)
These equations can be combined to give
wSRt = κR(ζSYt wSYt + ζSLt bSt ), (2.29)
where κR = (1 − τ)−1[(1 + ψ)(1 − eRm)]1/ψ > 1.12 Thus, the effi ciency wage is
proportional to, and higher than, the (expected) opportunity cost of working in the
innovation sector. At the optimal wage, the equilibrium level of effort is constant at
11If effort is independent of relative wages (ψ = 0), or if wages are continuously equal in bothsectors, then eRt = eRm.
12The Solow condition can be established by combining (2.27) and (2.28), which yieldswSRt (eRt )′/eRt = 1, where (eRt )′ = deRt /dw
SRt = ψ(1− eRt )/wSRt .
128
eR = 1− (1− eRm)(κR)−ψ > 0.
2.2.5 Government
The government operates both a general budget and an unemployment insurance
fund.13 It cannot issue bonds and must run balanced accounts in both cases. To
finance its general outlays, the government levies a tax on wages at the rate τ .
These outlays consist of investment in infrastructure, GIt , and spending on other
(not directly productive) items, GOt . It imposes no fees for its services.
The government’s general budget is given by
GIt +GO
t = τwUt NUYt +NSY
t [(1− ε)wSYt − tct] +NSRt [(1− ε)wSRt − tct]. (2.30)
Shares of spending are constant fractions of government revenues:
Git = υiτwUt NUY
t +NSYt [(1−ε)wSYt −tct]+NSR
t [(1−ε)wSRt −tct], i = I, O (2.31)
where υi ∈ (0, 1). Combining (2.30) and (2.31) therefore yields
υI + υO = 1. (2.32)
Let θht , h = UY, SY, SR, denote the proportion of employed individuals of cate-
gory h in the adult population N , and let θht , h = UL, SL, denote the unemployment
rate (again, in proportion of N) of labour category h; the unemployment insurance
fund’s budget is given by
(bUt θULt + bSt θ
SLt )N = ς twUt θUYt + (1− ε)(wSYt θSYt + wSRt θSRt )N,
13The unemployment insurance funds are generally funded by firms’payroll taxation, whereasgeneral budget is financed by taxes on workers’ wages. The different nature of the two taxesnecessitates the separation of the two funding mechanisms.
129
which implies
ς t =bUt θ
ULt + bSt θ
SLt
wUt θUYt + (1− ε)(wSYt θSYt + wSRt θSRt )
. (2.33)
Thus, all else equal, a higher benefit rate (bUt or bSt ) raises the payroll contribution
rate, thereby reducing labour demand. In turn, the reduction in labour demand
(through a fall in employment ratios) mitigates the initial increase in the contribution
rate at the initial unemployment and wage rates.
Assuming full depreciation, the stock of public capital evolves according to
KGt+1 = ϕGI
t , (2.34)
where ϕ ∈ (0, 1) is an effi ciency parameter, which measures the extent to which
investment outlays translate into actual public capital (Agénor 2012).
To ensure the existence of a nondegenerate solution, the unemployment benefit
is set as a linear function of the level of per capita income, so that
bht = κhYtN, (2.35)
where κh ∈ (0, 1), with h = U, S, is the benefit indexation parameter.
2.2.6 The Labour Market
Wages in the final good sector are set through a right-to-manage bargaining process
between a centralised trade union and firms. The union’s objective is to maximise
the expected current income of both types of workers in manufacturing, subject to
wage and employment targets.14
Specifically, the union sets wUt and wSYt with the objective of maximising a utility
14The union’s optimisation problem is static, in the sense that when it formulates its wagedemands it takes the existing capital stock as given and does not internalise the effect of futurewages on the firm’s decision to accumulate capital– and thus future labour demand. This istantamount to assuming sequential wage bargaining and the absence of reputational links acrossperiods.
130
function that depends on deviations of both employment and wages from their target
levels, subject to the manufacturing sector’s demand schedule for each type of labour.
Normalising the employment target to zero, the union’s utility function takes the
standard form
Vht = (wht − whTt )ξ
h
(Nht )1−ξh ,
where h = UY, SY , ξh ∈ (0, 1), and Nht is given in (2.13). The term whTt measures
the union’s target wage, whereas ξh reflects the relative importance that the union
attaches to wage deviations from that target. Maximising this function with respect
to wht gives the actual wage as a mark-up (which is increasing in ξh) over the target
wage:15
wht = (1− ξh
1− 2ξh)whTt . (2.36)
The target wage for untrained workers is related positively to a government-
imposed minimum wage, wUMt , and negatively to the unemployment rate for that
category of labour, θULt :
wUTt = wUMt (θULt )−κU
,
where κU > 0. When unemployment is high, the probability of finding a job (at any
given wage) is low. Consequently, the higher the unemployment rate, the greater
the incentive for the union to moderate its wage demands in order to induce firms
to increase employment.
In turn, the minimum wage is linearly related to the level of per capita income:
wUMt = wU0 (YtN
), (2.37)
where wU0 > 0 is an indexation parameter.
15To ensure that wht > 0 requires ξh < 0.5, a condition that we impose in the parameterisation.
131
Substituting the above expressions into (2.36) therefore yields
wUt = wU0 (1− ξU
1− 2ξU)(YtN
)(θULt )−κU
. (2.38)
The target wage for specialised workers is negatively related as well to the un-
employment rate for that category of workers, θSLt , and linearly related once again
to the level of per capita income, Yt/N , so that wSY Tt = wSY0 (θSLt )−κSYt/N , where
wSY0 > 0 is an indexation parameter16. Inserting this result into (2.36) yields
wSYt = wSY0 (1− ξSY
1− 2ξSY)(θSLt )−κ
S
(YtN
). (2.39)
The equilibrium condition of the market for untrained labour is given by
NUt = NUL
t +NUYt ,
where NULt is the number of unemployed. Equivalently, in terms of ratios to popu-
lation,
θUt = θULt + θUYt , (2.40)
where θUt = NUt /N , which from (2.8) is equal to aCt . Thus, the probability of
employment for an untrained individual, ζUYt , and the probability of an untrained
individual becoming unemployed, ζULt , are given respectively by
ζUYt =θUYtθUt
, and ζULt = 1− ζUYt =θULtθUt
. (2.41)
The equilibrium condition of the market for (effective) specialised labour is given
by:
NSt = NSY
t +NSRt +NSL
t ,
16Alternatively, the target wage of the specialised union can be set at a rate that is multiplethat of the untrained workers’wage. For the purposes of the numerical experiments, this can easilybe accomplished by setting wSY0 > wU0 when parameterising the model.
132
or equivalently, in terms of ratios to population,
θSt = θSYt + θSRt + θSLt . (2.42)
The employment and unemployment probabilities for specialised workers are
given by
ζSYt =θSYtθSt
, ζSRt =θSRtθSt
, and ζSLt = 1− ζSYt − ζSRt =θSLtθSt. (2.43)
Figure 2.1 summarises the production structure and the sectorial distribution of
labour. Although it does not show (for clarity) how employment and unemployment
probabilities are determined, it illustrates fairly well how labour market rigidities
affect wage formation and unemployment, and the feedback effect of unemployment
(through its impact on compensation for the unemployed) on expected wages and
the decision to acquire advanced training.
2.2.7 Savings-Investment Balance
Given full depreciation, the saving-investment balance requires private capital in
t+ 1 to be equal to savings in period t by all individuals, employed or unemployed,
born in t− 1:
KPt+1 = (sUYt NUY
t + sULt NULt ) + (sSYt NSY
t + sSRt NSRt + sSLt NSL
t ). (2.44)
2.3 Balanced Growth Equilibrium
In this economy, an equilibrium with imperfect competition and unemployment is a
sequence of consumption and saving allocations ch,jt|t , ch,jt|t+1, s
h,jt ∞t=0, for h = U, SY, SR,
j = E,L, prices of production inputs wUt , wSYt , wSRt , rt+1∞t=0, private capital KPt ∞t=0,
public capital KGt ∞t=0, existing varieties Mt∞t=0, prices and quantities of in-
133
termediate inputs P st , xs,t∞t=0, ∀s ∈ (0,Mt), such that, given initial stocks KP
0 ,
KG0 ,M0 > 0,
a) all individuals, specialised or untrained, employed or unemployed, maximiseutility by choosing consumption subject to their intertemporal budget constraint,taking factor prices, the tax rate, and the unemployment benefit as given;b) firms in the final good sector maximise profits by choosing labour, private
capital, and intermediate inputs, taking factor prices as given;c) intermediate input producers set prices so as to maximise profits, while in-
ternalising the effect of their decisions on the perceived aggregate demand curve fortheir product;d) producers in the innovation sector maximise profits by choosing labour and
wages, taking patent prices and productivity as given;e) the price of each blueprint extracts all profits made by the corresponding
intermediate input producer;f ) the trade union in the manufacturing sector sets wages so as to maximise its
utility, subject to the demand for labour by firms in the final good sector;g) the final good market clears; andh) unemployment of both categories of workers prevails.
A balanced growth equilibrium is an equilibrium with imperfect competition and
unemployment in which
a) ch,jt|t , ch,jt|t+1, s
h,jt ∞t=0, for h = U, SY, SR, j = E,L, and KP
t , KGt , Yt, Mt, wUt ,
wSYt , wSRt , bht , h = U, S, grow at the constant, endogenous rate 1 + γ, implying thatthe knowledge-private capital ratio and the public-private capital ratio are constant;b) the rate of return on capital, 1 + rt+1, is constant;c) the price of intermediate goods, Pt, and the patent price, Qt, are constant;e) the threshold level of individuals who choose to remain untrained, aCt , is
constant;f ) the fractions of the specialised and untrained labour force employed in man-
ufacturing, θUYt and θSYt , and the fraction of specialised workers employed in theinnovation sector, θSRt , are constant;g) specialised and untrained unemployment rates, θULt and θSLt , are constant;
andh) employment and unemployment probabilities, ζUYt , ζSYt , ζSRt , and ζ
ULt , ζSLt
are constant.
2.4 Properties of the Equilibrium
A key step in deriving the equilibrium growth rate is to establish the restrictions
needed on the congestion parameters in (2.12). With mt = Mt/KPt denoting the
134
knowledge-private capital ratio, equation (2.12) yields
Yt = (1− ε)βS(θSYt )βS
(θUYt )βU
NβS+βU−ωζNt (2.45)
×(kGt )ω
Λ1m(1−η)/ηt (
YtKPt
)
γ(KP
t )α+γ/η+ω(1−ζK),
where Λ1 = γη. To ensure that production is linear in the private capital stock, ζK
and ζN must satisfy the conditions βS +βU −ωζN = 0 and α+γ/η+ω(1− ζK) = 1.
As a result, the level of output becomes:
Yt =(kGt )ω/(1−γ)Λ2
[(θSYt )βS(θUYt )β
U]−1/(1−γ)
mt
(1−η)/ηγ/(1−γ)
KPt , (2.46)
where Λ2 = (1− ε)βSΛγ/(1−γ)1 .
The Appendix for this chapter shows that the dynamic system that drives the
economy is characterised by two first-order dynamic equations in terms of the
knowledge-private capital ratio, mt, and the public-private capital ratio, kGt , as
well as 9 core static equations, in terms of the output-private capital ratio, Yt/KPt ,
the patent price, Qt, the threshold level of ability (or equivalently the share of un-
trained workers), aCt , the shares of specialised workers in the final good production
and innovation activities, θSYt and θSRt , the share of untrained workers in final good
production, θUYt , the shares of specialised and untrained workers in unemployment,
θSLt and θULt , and the payroll contribution rate, ς t . These are:
mt+1 =[1 + (eR)λ(1− ε)λ(kGt )φ
R1 (θSRt )λ
]mt (2.47)
⟨(
σ
1 + ς t)YtKPt
(1− τ)βU + κUθULt (1 + ς t)
+(1− τ)
1− µ(1− τ)
[0.5(1 + aCt )]χ(θSYt + θSRt
θSt)
[θSYt + θSRt
κR
θSt(θSYt + ΩS(θSLt )1+κS ]
βS
θSYt
+
ΩSθ
St (θSLt )κ
S − µ(1− τ)
[0.5(1 + aCt )]χ[θSYt + θSRt
κR
θSt(θSYt + ΩS(θSLt )1+κS ]
βSθSLtθSYt θSt
⟩−1
.
135
KGt+1
KPt
= (ϕυIτ
1 + ς t)YtKPt
βU +
βS
θSYt[θSYt + θSRt
κR
θSt(θSYt + ΩS(θSLt )1+κS ] , (2.48)
× [1− µ(1− τ)
[0.5(1 + aCt )]χ(θSYt + θSRt
θSt)]
KPt+1
KPt
= (σ
1 + ς t)YtKPt
(1− τ)βU + κUθULt (1 + ς t) (2.49)
+(1− τ)
1− µ(1− τ)
[0.5(1 + aCt )]χ(θSYt + θSRt
θSt)
[θSYt + θSRt
κR
θSt(θSYt + ΩS(θSLt )1+κS ]
βS
θSYt
+
ΩSθ
St (θSLt )κ
S − µ(1− τ)
[0.5(1 + aCt )]χ[θSYt + θSRt
κR
θSt(θSYt + ΩS(θSLt )1+κS ]
βSθSLtθSYt θSt
,
kGt =KGt
KPt
, (2.50)
YtKPt
=(kG)ω/(1−γ)Λ2
[(θSYt )βS(θUYt )β
U]−1/(1−γ)
mt
(1−η)/ηγ/(1−γ)
, (2.51)
Qt = (1− η)γ(YtKPt
)(mt)−1, (2.52)
θUt = µ1/χ
1− ζUYt (1− τ) + [ζULt κU − (1− ε)ζSLt κS]ΩU(θULt )κ
U
(1− τ)[ζSYt + κRζSRt [ζSYt + (θSt )−1ΩS(θSLt )1+κS ]]
βU
βS(θSYtθUYt
)
−1/χ
,
(2.53)
θSt =1− (θUt )2
2, (2.54)
θSRt =
λγ(1− η)(kGt )φ
R1 (eR)λ(1− ε)λθSt
κR[βS + (1 + ς t)κS(1− ε)θSLt ]
1/(1−λ)
, (2.55)
θSYt =βS
wSY0
(1− 2ξS
1− ξS)(
1
1− ε)(1
1 + ς t)(θSLt )(κS), (2.56)
θSLt = θSt − θSRt − θSYt , (2.57)
θUYt = θUt − θULt , (2.58)
θULt = (wU0 )(κU )−1(1− 2ξU
1− ξU)−(κU )−1(
βU
1 + ς t)−(κU )−1(θUYt )(κU )−1 , (2.59)
136
ς t =θULt κU + θSLt κS
βU + βS
θSYt[θSYt + ζSRt κR(θSYt + ΩS(θSLt )1+κS ]− (θULt κU + θSLt κS)
, (2.60)
ζUYt =θUYtθUt
, and ζULt = 1− ζUYt =θULtθUt
, (2.61)
ζSYt =θSYtθSt
, ζSRt =θSRtθSt
, and ζSLt = 1− ζSYt − ζSRt =θSLtθSt
, (2.62)
1 + rt = α(YtKPt
). (2.63)
The first-order difference equation of (2.47) and the dynamic form (2.48) drive
growth and the main dynamics of the solutions. The former contains the innovation
and physical capital accumulation dynamics, with the former being the main engine
of growth. The latter contains the dynamics of both public and private capital. The
resource allocation mechanisms are determined by all the labour shares (θSt , θUt ,
θUYt , θULt , θSYt , θSRt , θSLt ), which also produce the by-product of the group-specific
probabilities (ζUYt , ζULt , ζSYt , ζSRt , ζSLt ). Along with the determination of the payroll
contribution rate, these provide the multiple feedback mechanisms in this dynamic
general equilibrium framework.
The long-run growth rate, 1 + γ, is given by17
1 + γ = (eR)λ(1− ε)λ(kG)φR1 (θSR)
λ
. (2.64)
The stability of the economy cannot be studied analytically, given the complexity
of the system. However, it is established numerically (using the parameterisation
discussed next) by solving for an initial balanced growth equilibrium that satisfies
the properties defined earlier and verifying that following a shock, or combination
of shocks, the system converges to a new equilibrium.
2.5 Model Parameterisation
To study the impact of labour market reforms, we parameterise two versions of
the model, the first corresponding to a “typical” high-income economy, based on
17Given that all stock variables grow at the same rate in equilibrium, other equivalent forms forthe steady-state growth rate can of course be defined.
137
averages for five European economies (Belgium, France, Italy, Portugal, and Spain)
and the second to a “typical”middle-income economy, based on averages for five
upper-income Latin American economies (Argentina, Brazil, Chile, Colombia, and
Peru). These two versions allow us to explore the extent to which the effects of labour
market reforms depend on structural characteristics. Indeed, beyond the level of
income, the countries included in each group share a number of common economic
features; in particular, all the Latin American countries have a relatively small
innovation sector (both in terms of employment and capacity to create knowledge),
whereas all the European countries impose high income tax and payroll contribution
rates to finance large redistribution programmes. At the same time, countries in both
groups are characterised by significant labour market rigidities and high levels of
unemployment, caused largely by permanent, structural factors rather than cyclical
determinants. The main sources of data are the OECD for European economies
and the Inter-American Development Bank, the International Labour Offi ce (ILO),
and the World Bank for Latin American countries. For convenience, population is
normalised to unity in both cases.
First, consider the high-income economy. On the household side, the annual dis-
count rate is set at 0.04. Assuming that there is an implicit first period (childhood-
early adulthood) that is not accounted for, each period in the model is set to 25
years to match life expectancy data. This gives an intergenerational discount rate of
0.375; the same value is used for the middle-income economy. The household savings
rate, σ, is set at 0.1094, based on the average (net) household savings rate estimated
using OECD data for 2006-13. The relative cost of specialised training (or tertiary
education), µ, and the average time spent in such training, ε, are calibrated using
data from OECD Education at a Glance 2015. Specifically, for the five countries
considered, the expected number of years of full time schooling in tertiary education
is 2.86 years. Divided by 25, this gives ε = 0.115. Regarding education expenditure,
we use the estimated annual average tuition fees charged by educational institutions
in 2013-14. While the OECD publishes a range of values for each country and across
public and independent private institutions, we narrow them down to a single range
estimate for each country. Then, dividing by the reported average annual wage, the
average tuition fee is calculated to be about 6.1-7.7 percent of the average wage. We
set µ to a slightly higher value of 0.08 to account for other ancillary expenditure. To
138
account for a high degree of effi ciency of training in a developed-economy setting,
the parameter χ is set at a high value of 0.9.
In the final good sector, the elasticity of production with respect to the public-
private capital ratio, ω, is set at 0.17, in line with the meta-analysis of Bom and
Ligthart (2014) and the results of Calderón et al. (2015). The elasticities of output
with respect to private capital and labour are set at standard values of α = 0.3
and 0.6, respectively, consistent with the evidence (see for instance, Afonso and St.
Aubyn (2009) and Varga et al. (2014)). We then set βS = βU = 0.3, to reflect equal
importance of both types of labour in production. Given the assumption of constant
returns to scale, the elasticity of output with respect to intermediate inputs, γ, is
set at 0.1.
In the intermediate goods sector, the substitution parameter, η, is set at 0.61,
consistent with the value used by Iacopetta (2011) for instance. This yields an
elasticity of substitution between intermediate goods of 2.6, which corresponds to
the value estimated by Acemoglu and Ventura (2002).
In the innovation sector, the productivity parameter with respect to public in-
frastructure, φR1 , is set at 0.186, based on the estimates of Agénor and Neanidis
(2015). The elasticity of design production with respect to labour, λ, is set at 0.6,
the same value used by Varga et al. (2014) for Italy and Spain. It is also within the
range of 0.13-0.74 estimated by Pessoa (2005) for OECD countries. The elasticity of
effort with respect to relative wages, ψ, is set at 0.7, slightly higher than the value
used by Wauthy and Zenou (1997). To capture the idea that researchers in innova-
tion value wages more than leisure, we set δR = 0.9 for the elasticity parameter in
the second-stage utility function. This yields a probability of getting caught shirking
of π = 0.078. With a minimum research effort of eRm = 0.1, this yields a value of
1.46 for the composite parameter κR; consequently equilibrium effort is eR = 0.31.
For the government, the effective tax rate on wages, τ , is calculated in two steps,
based on OECD tax statistics. First, taxes on household factor income are estimated
by calculating total tax revenues net of taxes on property, goods and services, and
social security contributions. As a share of GDP, this gives an average of 11.9 percent
for the period 2006-13.18 Second, this number is divided by the total labour share
βS + βU = 0.6 to give τ = 0.198. To calculate the initial share of public investment
18Given the OECD’s revenue classification system, this is equivalent to calculating taxes onhousehold income by adding up income taxes and taxes on workforce and payroll.
139
on infrastructure in total (noninterest) spending, υI , we also proceed in two steps.
First, using combined OECD data on non-ICT infrastructure investment and ICT
investment for the years 2006-13, the average percentage of (total) infrastructure
investment to GDP across the sample economies is estimated at 0.0106. Second,
this estimate is divided by the average share of noninterest expenditure in GDP
for the same period, as estimated from OECD data, which is 0.4972. This yields
υI = 0.021, or equivalently 1.1 percent of GDP. Lastly, the effi ciency parameter
of government investment, ϕ, is calibrated using the “wastefulness of government
spending” indicator in the Global Competitiveness Report index compiled by the
World Economic Forum, which is consistent with the methodology used by the
European Commission. This yields ϕ = 0.5. This value is rather on the low side for
a high-income economy but is consistent with the informal evidence on comparative
public sector effi ciency in Afonso et al. (2003) for instance, who identified Italy,
Portugal, and Spain as among the most ineffi cient among the 23 developed economies
in their sample.
In the labour market, the benefit indexation parameters, κU and κS, are both set
equal to 0.4, in line with values used in models with unemployment insurance, such
as Heer and Morgenstern (2005). Given (2.35), this means that the initial values of
bS and bU are the same. For the union bargaining parameters, ξU and ξSY , we start
with the estimates of Blanchflower and Bryson (2002), which give an average union
wage mark-up of 1.069.19 Using this value, estimates for ξU and ξSY can be derived
by solving (2.36) backward; this gives ξU = ξSY = 0.06. In terms of the elasticity of
the union’s target wage with respect to unemployment, κh, h = U, S, Montuenga et
al. (2003) estimate the wage elasticity with respect to the unemployment rate for
four of the European economies in our sample (with the exception of Belgium); this
yields an average value of −0.12. In the absence of skills-specific estimates, we set
κU = κS = 0.12. The shift parameter wU0 is solved implicitly from the minimum
wage equation (2.37), based on OECD data on monthly minimum wages relative
to monthly average earnings (as a proxy for monthly income per capita); this gives
0.522. The shift parameter wSY0 in (2.39) is solved for in the same manner, using
data on monthly earnings for skilled workers, after accounting for the average gap
in earnings dispersion provided in the OECD’s Employment Database. This gives
19For France, the more recent results of Breda (2015) corroborate the Blanchflower-Brysonestimate.
140
wSY0 = 0.74.
These values are all summarised in Table 2.1. Initial steady-state values are
shown in Table 2.2 and are calibrated as follows.
The share of untrained workers in the adult population, θU , is set equal to 0.732,
which is calculated by subtracting the average share of workers with tertiary edu-
cation (obtained from OECD data) from unity. Hence, θS = 0.232. The share of
effective specialised workers in the innovation sector, θSR, is set equal to 0.0194,
based on the OECD’s consolidated data on (private and government) researchers.
The share of unemployed specialised workers in the population, θSL, is set at 0.068,
which corresponds to the value provided by the OECD’s World Indicators of Skills
for Employment data for skilled unemployment over the period 2006-13. By im-
plication, the share of effective specialised workers in the final good sector, θSY , is
equal to 0.145. Based on the same OECD data, the untrained unemployment rate,
θUL, is set equal at 0.126, corresponding to the average, group-specific unskilled
unemployment rate. By implication, the share of untrained workers in the final
good sector, θUY , is 0.606. The probabilities in (2.41) and (2.43) are then easily
calculated and are also reported in Table 2.2. The aggregate unemployment rate
can also be easily derived, given relative shares of untrained and specialised work-
ers in the workforce; this gives 0.1058. To estimate the misallocation of talent, we
use the average value over 2006-13 from OECD data on the proportion of workers
who are overqualified, which is equal to 0.189.20 Based on that value, the potential
supply of specialised labour to that segment of the market, θR, can be estimated
backward using the definition of the share of “overqualified”workers in the final
good sector, (θR − θSR)/θSY . Given that θSRt = 0.0194 and θSYt = 0.145, this yields
θR = 0.189 · 0.145 + 0.0194 = 0.0467. By implication, the threshold value of abil-
ity to work in the innovation sector is solved from (2.10) to give aR = 0.952. For
the firms’payroll contribution rate, ς, the average employers’contribution rate of
the five economies obtained from the OECD Social Security Dataset is used; this
gives ς = 0.126. Using the OECD’s relative earnings data by education gap for
2012 (low and medium-skilled workers on the one hand, and high-skilled workers on
20The data is based on OECD calculations using the EU Labour Force Survey. Based on OECDdefinition, the published figures reflect the “proportion of workers whose educational attainmentlevel is higher than the level required in their job (as measured based on the modal education levelfor all workers in the same occupation)”.
141
the other), the untrained-specialised wage ratio is calibrated at 0.55; the inverse of
this ratio gives a wage premium of 1.818. The public-private capital ratio, kG, is
set based on Kamps’(2006) estimates of public and private capital stocks, yielding
kG = 0.189. Using OECD data, the average final output-private capital ratio is cal-
culated as Y/KP = 0.286. An initial estimate of the knowledge-private capital ratio,
m, is diffi cult to construct, given that the two variables are in principle measured
in different units (the number of patents for instance, for the stock of knowledge,
and cumulated real investment spending, through an effi ciency-adjusted, perpetual
inventory method, for the capital stock). Given that this initial ratio is immaterial
to the results, we normalise it to 0.1 largely for computational convenience. The
growth rates of final output and physical capital in the initial steady state are 0.8
percent on an annual basis, based on the GDP-weighted average growth rates of the
five economies during 2006-13.
Consider now the typical middle-income country. To capture some relevant styl-
ised facts for these economies, its baseline parameterisation needs some distinctive
structural characteristics. Given the issues at stake, we highlight the following fea-
tures. First, it is more costly, and less effi cient, for a worker to train and become
specialised. Second, due to relative scarcity, the elasticity of final good production
with respect to specialised workers is higher, and there is less substitutability among
intermediate goods. Third, the share of public spending on infrastructure is higher
but investment (as a result of poor governance) is less effi cient. At the same time,
the elasticity of manufacturing output with respect to public capital is higher, to
reflect stronger marginal benefits due to a lower initial stock of infrastructure assets.
Fourth, the innovation sector (as measured by the number of researchers) is smaller
and workers are subject to less intense monitoring. Quantitatively, the differences
that these features lead to, as well as other differences in terms of initial values (as
discussed next), are shown in Tables 2.1 and 2.2 as well.
On the household side, estimates based on household surveys by Gandelman
(2015) are used to set the savings rate σ at 0.138. The average school life expectancy
at tertiary level for the five Latin American economies is 3.07 years, which gives
ε = 0.123. To account for more costly and less effi cient training, and in the absence
of data similar to those referred to earlier for the high-income economy, the training
cost µ is set at 0.12, and the effi ciency of training χ at 0.5. In the final good sector,
142
the elasticity of production with respect to the public-private capital ratio ω is set at
0.24, in line with the general equilibrium estimates of Agénor and Neanidis (2015).
The elasticity parameter with respect to private capital, α, is set equal to 0.35.
This is the average value for the five Latin American economies used for instance,
in the growth accounting exercises of Loayza et al. (2005). Following Agénor and
Alpaslan (2014), we set βU = 0.20 and βS = 0.35, so that γ = 0.1 again. The
implied private capital/labour share, 0.35/0.55, is consistent with a 0.4/0.6 ratio
used in some models without intermediate goods.21
In the intermediate goods sector, the substitution parameter, η, is set at 0.25,
which corresponds to the value used by Agénor and Neanidis (2015) to examine
innovation-driven growth in a developing-economy context. This value implies there-
fore a lower elasticity of substitution (about 1.33) between intermediate goods than
before. In the same vein, in the innovation sector φR1 is set at 0.3, which is consis-
tent with the initial parameterisation and the higher range of estimates obtained by
Agénor and Neanidis (2015). To capture lower research monitoring intensity, the
probability of being caught shirking is set 3 percentage points lower than in the high-
income economy, so that π = 0.048. This yields ψ = 0.43 and an equilibrium effort
level of eR = 0.143, which is about half the value calibrated for the high-income
economy.
For the government, a similar parameterisation strategy based on the same
sources (OECD tax revenue statistics for Latin America, and Global Competitive-
ness Index) is used to estimate the effective tax rate, τ , and the effi ciency of public
investment, ϕ. These calculations give averages of τ = 0.123 and ϕ = 0.4. This
estimate of ϕ is close to the median value obtained by Dabla-Norris et al. (2012) in
their study of the effi ciency of public investment in developing countries. The share
of public spending on infrastructure, υI , is estimated in two steps, based on the data
on total infrastructure investment as a proportion of GDP compiled by Calderon
and Servén (2010) and Carranza et al. (2014). The private component of total
investment, obtained from the World Bank’s Private Participation in Infrastructure
Database, is first subtracted to obtain the share of public infrastructure investment
as a proportion of GDP. This figure is then multiplied by the inverse of the ratio of
non-interest government expenditure to GDP to obtain an estimate of υI for each of
21See Agénor and Canuto (2015a) for Brazil, and Ferreira et al. (2013) for Latin America.
143
the five Latin American economies. The average value for the five economies for the
period 2006-13 gives υI = 0.069, or equivalently 2.0 percent of GDP. Regarding the
labour market, in the absence of reliable estimates, the same values of κU and κS as
given earlier are used. The minimum wage shift parameter, wU0 , is calibrated again
based on the average ratio of the gross monthly minimum wage over gross monthly
earnings, as provided in ILO Statistics. This gives wU0 = 0.546. For wSY0 , the median
wage differentials between secondary-primary and secondary-tertiary are used (see
Inter-American Development Bank (2004, Table 1.8)) to estimate an average value
for wage dispersion in the five Latin American economies. This yields 0.153, which
implies, solving again (2.39) implicitly, wSY0 = 0.699. This also means that the initial
wage gap for workers in the final good sector is smaller in the high-income economy.
In terms of unemployment benefits (which cover in reality a fairly limited number of
workers), estimates by Cortazar (2001) and Ferrer and Riddell (2009) suggest that
for the group of countries under consideration unemployment insurance represents
from 0.12 to 2.5 times the minimum wage. Multiplying by wU0 = 0.546 yields a range
of 0.06-0.82 for κU and κS. Mid-range values of κU = κS = 0.4 are used initially.
Lastly, for the union wage mark-up, the Inter-American Development Bank (2004)
documents that unions in South America increase their members’earnings by any-
where between 5 and 10 percent. Setting the wage mark-up to 1.1, and solving again
(2.36) backward yields ξU = ξSY = 0.08.
In terms of initial steady-state values, the labour shares are estimated using data
from ILO and theWorld Bank. The share of untrained workers in the population, θU ,
is set equal to 0.795, which yields θS = 0.184. The share of effective specialised work-
ers in innovation, θSR, is estimated by dividing the average number of researchers
over the total workforce for the five economies over 2006-13, yielding θSR = 0.004.
The share of unemployed specialised workers, θSL, is set equal to 0.071, based on ILO
data. By implication, θSY = 0.109. The unemployment rate for untrained workers,
θUL, is also obtained from ILO data and is set at 0.087. These data therefore imply
that θUY = 0.708, and the aggregate unemployment rate is now 0.0791. In the
absence of OECD-type data on the proportion of “overqualified”workers in Latin
America, we set the ability threshold aR (and therefore θR, as implied by (2.10)) at
the same value as in the high-income economy, 0.952. The initial degree of talent
misallocation can thus be solved backward from (θR− θSR)/θSY , to give 0.392. This
144
implies that there are a lot more overqualified workers in the final good sector of the
middle-income economy, consistent with recent theories of middle-income traps (see
Agénor (2016)). The firms’payroll contribution rate, ς, is set at 0.052.22 The initial
relative wage ratio is estimated at 0.75 based on ILO data, implying that the initial
expected wage premium is now lower, at 1.333. The public-private capital ratio
calculated for Brazil by Agénor and Canuto (2015a), kG = 0.147, is used as a proxy
for the group average. The final output-private capital ratio, Y/KP , is calibrated
using the private capital-GDP ratios for Argentina, Brazil and Chile estimated by
Tafunell and Ducoing (2016). This yields Y/KP = 0.429. The knowledge-private
capital ratio, m, is normalised again to 0.1. Lastly, the annual growth rates for final
output and capital in the initial steady state are equal to 3.9 percent, based on the
GDP-weighted average growth rate of the five economies during 2006-13.
Based on Tables 2.1 and 2.2, and consistent with our earlier discussions, the key
differences between the middle-income economy and the high-income economy can
be summarised as follows: a) higher effi ciency and lower cost of training in the high-
income economy; b) a lower degree of substitution between intermediate goods in the
middle-income economy; c) higher elasticities of final output and innovation activity
with respect to public capital in the middle-income economy; d) a higher share of
specialised workers in the population and in the innovation sector in the high-income
economy; e) a higher open unemployment rate for untrained (specialised) workers in
the high- (middle-) income economy; f ) a higher degree of misallocation of talent in
the middle-income economy; g) a higher payroll contribution rate in the high-income
economy; and h) higher public-private capital and final output-private capital ratios
in the high-income economy.23
22While payroll taxes represent on average of 31 percent of wages in Latin America (see Loraand Fajardo (2012)), only the portion that employers contribute to the unemployment/severancefund is accounted for here.
23Another important structural difference between the two types of economies is the share ofspending on R&D: Latin American countries spend much less than European countries in that area(see Inter-American Development Bank (2014)). Given the focus of this chapter, we did not explic-itly account for that component of public spending or other measures aimed at stimulating R&D(such as tax credits or “matching grants”subsidies). Note also that, consistent with the evidence,for the middle-income country, innovation is perhaps best understood as imitation (adaptation ofimported technologies) with the patent price being akin to a a license fee paid by intermediategoods producers.
145
2.6 Policy Experiments
Consider a series of individual labour market policies– a reduction in the minimum
wage, a cut in unemployment benefit rates, and a reduction in the union’s wage
mark-up. In addition, a policy aimed at promoting the accumulation of human cap-
ital (a cut in training cost) is also considered. These policies have been discussed
extensively in recent years, in both developed and developing countries.24 All shocks
are permanent and their impact is measured in terms of a few key variables– the
supply of untrained workers, the effective supply of specialised workers (both total
and in the innovation sector), the expected wage premium (which determines train-
ing decisions), unemployment rates (total and for both categories of workers), the
payroll contribution rate, and the growth rate of final output.
To measure the effi ciency gains of reforms in terms of factor allocation, the index
of misallocation of talent defined earlier is used. To measure welfare, discounted
utility across an infinite sequence of generations is used (see De la Croix and Michel,
2002, p. 91):
Wt = 0.2∞∑h=0
Λh(V U,Et+s + V U,L
t+h + V SY,Et+h + V SR,E
t+h + V S,Lt+h ), (2.65)
where Λ ∈ (0, 1) is the social discount factor and V h,jt is the indirect utility function
for agent j, h at t, where h = U, SY, SR and j = E,L. Thus, the utility of agents
in each generation in all five states– untrained workers employed or unemployed,
specialised workers employed in the final good sector and innovation activities or
unemployed– are equally weighted.25 For tractability, we restrict our analysis to
the balanced growth path; the Appendix provides an approximation to (2.65) along
that path, with Λ set to the same value used for households.
The simulation results (impact and steady-state effects) are summarised in Table
2.3 for the high-income economy and in Table 2.4 for the middle-income economy,
whereas Figure 2.2 shows the steady-state effects for all experiments. As noted
earlier, a period corresponds in principle to a generation in our OLG structure. This
24See Inter-American Development Bank (2004), World Bank (2012a, 2012b), Adascalitei andPignatti Morano (2015) and International Monetary Fund (2016) for instance.
25Alternatively, weights based on steady-state relative shares of each group of workers in thelabour force (which deviate from baseline values as a result of the labour reallocation effectsassociated with each experiment), could be used. Qualitatively the results are broady similar tothose reported.
146
is reflected, in particular, in the parameterisation of the discount factor, household
time allocation, and the assumption of full depreciation of physical capital. However,
all of the other parameters and variables (including the growth rate of output) either
do not have a time dimension or are parameterised on the basis of average annual
data; thus, for the numerical experiments, the intended length of a unit of time is
best understood as one year.
2.6.1 Reduction in Minimum Wage
Consider a reduction in the minimum wage, measured by a 5 percent drop in the
shift parameter wU0 . The reduction in the cost of untrained labour increases demand
not only for that category of workers but also (due to gross complementarity) for
specialised labour in manufacturing. At the initial level of wages, the unemployment
rate falls and the employment probability rises for both categories of workers. How-
ever, the expected wage for specialised workers increases by more than the expected
wage for the untrained workers, thereby creating incentives to invest in advanced
training. The proportion of untrained (specialised) workers therefore falls (increases)
on impact. The increase in specialised employment occurs in both the final good
and innovation sectors, though in the middle-income economy, not all specialised
labour from the expansion are absorbed, resulting in a slight increase in long-run
specialised unemployment rate. The long-run drop in unemployment is particularly
large for untrained workers, of the order of 2.8 percentage points for the high-income
economy and 2.0 percent for the middle-income economy.26
Higher employment for both types of workers translate into a reduction in the
payroll contribution rate, which magnifies the expansion of labour demand in manu-
facturing. Although the initial fall in unemployment tends to raise the union’s target
wages in the manufacturing sector– thereby mitigating the initial effect of a lower
minimum wage– the increased demand for both types of workers tends to promote
activity and economic growth, both on impact and in the long run. However, the
long-run effects are fairly small in both economies.
Higher wages for specialised workers in manufacturing imply higher wages in
26The reduction in unemployment is consistent with the evidence reviewed by Neumark andWascher (2006) although, as they point out, the wide range of estimates makes the precise identi-fication of the magnitude of this effect diffi cult.
147
the innovation sector as well, to maintain effort there. This helps to increase the
share of that type of labour engaged in innovation activity, thereby mitigating the
misallocation of talent, by a magnitude of 0.9 and 0.4 percentage points in the long-
run for the high- and middle-income economy, respectively. In addition, welfare
improves moderately in both cases. In terms of their magnitude, both results reflect a
small increase in employment in the innovation sector, a weak effect on the expansion
of varieties of intermediate goods, and therefore a small impact on growth in the long
run. Overall, lower minimum wages do not necessarily harm growth and welfare–
in contrast to the predictions of some small analytical models, such as Cahuc and
Michel (1996)– but their effects on these variables, given our parameterisation, are
not quantitatively large.
2.6.2 Reduction in Unemployment Benefit Rates
Three separate experiments with respect to a scaling down in unemployment benefit
indexation are considered: a) a reduction in the indexation parameter for only
untrained workers, b) a reduction for only specialised workers, and c) a reduction for
both type of workers. Specifically, we consider cuts in κU and κS by 10 percent (from
0.40 to 0.36) each, and a joint reduction in κU and κS of the same magnitude. These
experiments allow for the examination and comparison of the effects of asymmetric
adjustments in unemployment insurance schemes, as well as the case of an across-
the-board reform.
A reduction in the benefit rate for untrained workers lowers their expected wage
at the initial level of employment. It therefore raises the education premium and
incentives to undergo training. As a result, the share of untrained (specialised)
workers falls (increases). The opposite occurs for a reduction in the benefit rate for
specialised workers. However, in both cases, aggregate unemployment falls– more so
for the high-income economy– both on impact and in the long run. This stems from
the fact that the direct effect of a lower wage is (as a result of gross complementarity)
to stimulate the demand for both types of labour. This effect, which is magnified by a
reduction in the payroll contribution rate needed to ensure that the unemployment
fund’s budget is balanced, persists over time as well. However, unlike the more
effi cient high-income economy, for the middle-income economy, long-run specialised
(untrained) unemployment rate increases slightly when the indexation parameter
148
is reduced for the untrained (specialised) workers. This is due to a weaker gross
complementarity effect and a smaller expansion in the innovation sector, which
mitigates its capacity to absorb the increase in specialised labour.
On impact, the growth rate of final output falls in both types of economies.
The reason is that the drop in benefits for the unemployed has an adverse effect on
savings, which reduces investment and capital accumulation in the short run. Over
time, however, two offsetting general equilibrium effects kick in: lower benefits (for
untrained workers) improve incentives for individuals to acquire training, whereas a
lower contribution rate raises labour demand. In the long run, the net effect of the
policy is in fact positive– albeit fairly weak for both economies. Although talent
misallocation is mitigated, welfare falls in both cases (for either shock) essentially
because the unemployed are worse off. The joint reduction in unemployment benefit
indexation gives results that are qualitatively similar to those obtained in the indi-
vidual experiments, and in this instance, unemployment falls– both at the aggregate
level and its components– for both types of economies.
In the well-cited study of Bouis and Duval (2011), their regression-based analysis
of a cut in unemployment benefit rate for the 5 European economies examined here
(ranging from 52−62 percent cuts) would result in a decline in total unemployment
rate, ranging from −1.3 percent in Spain to −3.2 percent in Portugal. If we were to
simulate a uniform reduction in unemployment benefit rate to such a large magnitude
using this model (the benchmark here is a 10 percent cut), the decline in total
unemployment rate for the high-income economy will be approximately 2.7 percent,
which is within the range of their estimates. In a subsequent DSGE model-based
study of Cacciatore et al. (2012), their policy simulation results of a similar large cut
in unemployment benefit rate (from average rate of “non-rigid”to those of “flexible”
OECD countries) find a −4.0 percent steady-state reduction in total unemployment
rate and a 2.9 percent increase in output growth. The direction for the steady-state
effects of the two variables are consistent with the simulation results here, though
their expansionary growth effects are too optimistic compared to those found here
with the endogenous growth framework. However, the conflicting effect on long-run
growth and welfare has not been documented in previous contributions. It suggests
that a reduction in unemployment benefit indexation, while effective in terms of
reducing unemployment for both types of labour, may need to be accompanied by
149
other measures aimed at mitigating their potential adverse impact on household
well-being.
2.6.3 Reduction in the Union’s Wage Mark-Up
Consider a large reduction in the mark-up over the target wage for both untrained
and specialised workers, as measured by the parameters ξU and ξSY , respectively (see
(2.36)). This experiment involves a uniform 37.5 percent cut in these parameters,
from 0.06 to 0.0375 for the high-income economy and from 0.08 to 0.05 for the
middle-income economy. By implication, the union wage mark-up over the target
wage (for both untrained and specialised workers) drops by 2.6 percent in the former
and by 3.6 percent in the latter.
In both cases, unemployment rates for the two types of workers are lower in
the short run. However, similar to the previous experiments, for the middle-income
economy, this labour market policy targeted at untrained (specialised) workers is in-
effective again in reducing unemployment of specialised (untrained) workers due to a
weaker gross complementarity between the two types of labour. In both economies,
the benefits in terms of short-term growth are substantially higher for the mark-
up reduction for specialised workers, but in the long run, the unemployment and
growth effects (although qualitatively similar to the short-run effects) are fairly
small. Again, the magnitudes found are largely consistent with reform studies that
explicitly examine trade union reform. For instance, Lusinyan and Muir (2013), who
study the macroeconomic impact of reforms in Italy, document that a 5 percent re-
duction in the economy wide wage mark-up would lead to about 1.0 percent increase
in output growth. However, they did not examine the effects on employment. In
comparison to the policy experiments here (which simulated a 2.6 and 3.6 percent
decline in union wage mark-up for the untrained and specialised respectively), this
is consistent with the range of impact effects observed (for instance, in the high-
income economy, a 0.4 and 0.9 percent increase in growth is observed for untrained
and specialised respectively).
For both types of economies, welfare deteriorates when the mark-up for spe-
cialised workers is reduced, but improves slightly when the mark-up for untrained
workers is lowered. Again, these results suggest that, taken in isolation, these poli-
cies do not have substantial effects on growth and unemployment in the long run,
150
and maybe detrimental to welfare.
2.6.4 Reduction in Training Cost
Finally, consider a policy designed to reduce across the board the cost of specialised
training for individuals, paid for by a reallocation of outlays within the unproductive
component of public spending. The policy once again has no direct fiscal effects and
is measured by a reduction in µ by 5 percent, from 0.080 to 0.076 for the high-income
economy and from 0.120 to 0.114 for the middle-income economy. The size of this
shock is suffi cient to illustrate the issues at stake.
A reduction in training costs generates a large increase in the supply of specialised
workers (by 2.1 and 3.8 percentage points in the long run, respectively, for the high-
and middle-income economies), a fraction of which being absorbed in the innovation
sector. This increase in supply occurs despite the mitigating effect on wages for that
category of workers and a drop in the expected wage premium. The reduction in the
share of untrained workers has a sizable effect on their unemployment rate; however,
the large increase in the supply of specialised workers leads over time to a higher
unemployment rate for them (by 1.3 and 3.2 percentage points in the long run for
the high- and middle-income economies, respectively). The thrust of these results
is that, in both types of economies, promoting human capital accumulation without
adequate measures aimed at encouraging simultaneously a sustained expansion in
labour demand may create an absorption problem or oversupply of specialised labour
in the long run.
In addition, the positive effect on the rate of economic growth is small on impact
in both types of economies and, in the case of the middle-income economy, also
weaker in the long run. The reason, as noted earlier, is that the net benefit of an
increase in the supply of specialised workers is muted, due to a smaller expansion in
labour demand in the innovation sector. The larger increase in the specialised un-
employment rate in the middle-income economy also results in a higher payroll con-
tribution rate, which mitigates the increase in labour demand and dampens steady-
state growth. Nevertheless, and despite the increase in specialised unemployment,
welfare improves for both types of economies because employed untrained workers
and both types of unemployed workers gain from this policy. For the former, this
is because wages are ultimately higher than initially. For the unemployed, this is
151
because unemployment benefits are higher along the equilibrium path, due to higher
steady-state growth.
The negative correlation between the incentive to acquire skills and the supply of
specialised workers induced by a reduction in the cost of training, as predicted here, is
consistent with the evidence on the inverse association between increases in the num-
ber of university graduates and the wage premium provided by Machin and McNally
(2007) for Spain– one of the countries in our sample of high-income economies– and
New Zealand. Although they do not link it explicitly with a government-induced,
sustained reduction in the real effective cost of higher education (a broader interpre-
tation of a lower µ in the experiment), the evidence for both countries is consistent
with it.27
Evidence supportive of the possibility that more university graduates may lead
to higher open unemployment, as also predicted here, is more diffi cult to come by
for at least three reasons—which are equally relevant for high- and middle-income
countries. First, higher unemployment rates for new university graduates often re-
sult from mismatches between supply and demand for particular skills (for instance,
liberal arts), or low quality standards– an important problem in Latin America,
as noted by Yamada (2015)– rather than an across-the-board lack of demand for
labour, as predicted by our experiment. Second, rather than open unemployment,
in practice, university graduates may choose to be employed in occupations that
do not fully exploit their skill levels, which therefore translates into underemploy-
ment or disguised unemployment.28 Finally, graduates may also choose to migrate
abroad, a form of brain drain. Although the model does not explicitly capture any of
these possibilities it does nevertheless draw attention to the adverse labour market
27Although we were unable to find publicly available statistics on real effective cost of highereducation and its evolution over time, in the case of Spain for instance, two specific educationalpolicies– Ley Orgánica, de Reforma Universitaria in 1983 and "Informe sobre la financiación delas universidades" in 1994– led directly to the establishment of student financial aid system andthe reduction of tuition fees. These, coupled with the large subsequent increase in the number ofpublic universities (the total number of universities increased from 35 in 1985 to 78 in 2010, and themajority of these are public universities) would almost certainly result in a significantly decrease inthe real effective cost of tertiary education– consistent with the experiment. In practice, however,an increase in the number of university graduates may also result from improving high schoolenrollment and completion rates (especially for middle-income countries) or sustained increases inper capita income, which translates into a higher demand for education.
28The possibility that underemployment may result from overeducation is the subject of an ex-tensive microeconomic literature reviewed by Leuven and Oosterbeek (2011), who also documentedits incidence in Europe and Latin America.
152
effects of an oversupply of skills, due to a low effective cost of education promoted
by government subsidies. Social demands to expand access to higher education may
ultimately prove counterproductive.
2.7 Composite Reform Programmes
The foregoing analysis suggests that reforms may entail dynamic tradeoffs: they
can have adverse effects on the labour market and growth in the short run, despite
improving these outcomes in the long run. This tradeoff could induce a government
motivated by short-term electoral considerations to postpone, or abandon altogether,
the implementation of structural reforms. In addition, growth and welfare may move
in opposite directions in the long run, as illustrated in the case of a reduction in the
degree of indexation of unemployment benefits and a cut in the trade union’s mark-
up on specialised workers’wage target. A natural issue to address therefore is to
what extent a combination of measures—assuming that it is politically feasible—can,
by exploiting policy externalities, mitigate the contrasting effects associated with
individual reforms.
Accordingly, we now consider alternative composite reform programmes involv-
ing a combination of the individual policies discussed earlier. In addition, we exam-
ine the extent to which composite programmes designed to reduce unemployment
and promote growth would benefit from an increase in public infrastructure in-
vestment. This issue has been much discussed in recent years, in the context of
persistent, ultra-low interest rates in the global economy.29
2.7.1 Core Programmes
Two core composite reform programmes are considered first. In both of them we
assume that the key objectives of policymakers are to reduce unemployment and
to promote skills acquisition to support innovation-driven growth. Given that the
29The European Commission for instance, has ambitious deployment targets for high-speed,fiber-based broadband networks in its 2020 strategy. Many observers have argued that publicfunding is necessary to achieve ubiquitous coverage in remote and unprofitable regions, as opposedto densely populated areas; see Briglauer et al. (2016) for a discussion. In Latin America ba-sic infrastructure needs (including core internet access) remain large and calls for higher publicinvestment have also been vocal; see Serebrisky et al. (2015) for instance.
153
distribution of high-ability individuals in the population is fixed, the latter objective
can be achieved only by raising the productivity of those currently employed in the
innovation sector, in order to induce higher wages and reduce the misallocation of
talent. The combination of policies considered, although fairly targeted (given the
focus on structural, rather than cyclical, unemployment), is consistent with long-
standing calls for comprehensive programmes of labour market reforms, as noted
earlier.
The first programme, denoted Programme A, consists of pure labour market
reform measures, which are the same in both countries in relative terms. It involves
a cut in the minimum wage, as measured by a 10 percent decrease in the shift
parameter wU0 , a reduction in the unemployment benefit indexation parameters, κU
and κS, by 6.25 percent (from 0.4 to 0.375), and a 37.5 percent cut in the union’s
untrained wage preference parameter ξU (a drop from from 0.16 to 0.10 for the
high-income economy and from 0.08 to 0.05 for the middle-income economy).30
The second programme, Programme B, adds human capital-promoting policies
to these measures, to exploit potential gains associated with a skills expansion.
Specifically, in addition to the measures in Programme A, Programme B adds an
increase in specialised training time, as measured by ε, and a reduction in specialised
training cost, µ.31
The impact and steady-state effects of both programmes are shown in Table
2.5 whereas the transitional dynamics for both types of economies are illustrated
in Figures 2.3 and 2.4. The transmission mechanism of the combined shocks is,
naturally enough, a composite of the features outlined earlier. The effects of Pro-
gramme A, which consists of pure labour market reforms, are clear: reductions in
both untrained and specialised unemployment rates in both the short and the long
run—in the steady state the former (latter) drops by 6.5 percent (0.5 percent) for the
high-income economy and 4.9 percent (0.1 percent) for the middle-income economy—
reduced misallocation of talent, small gains in both overall specialised workers and
the proportion employed in the innovation sector (despite the increase in the wage
30We consider an across-the-board cut in unemployment benefit indexation, even though weassume that reforms mainly target untrained unemployment, because this is the way these policiesare implemented in practice.
31For the high-income economy, this translates into a rise in ε from 0.1145 to 0.14, and a 5percent reduction in µ from 0.08 to 0.076. For the middle-income economy, ε rises from 0.123 to0.15 and µ falls from 0.120 to 0.114.
154
premium), weak growth effects, and a deterioration in social welfare. This last result
is largely due to the unemployed being worse off from the benefits cut, given the
small gain in long-run growth in output and income.
As expected, the results for Programme B show a fairly significant increase (re-
duction) in the supply of specialised (untrained) workers—of the order of 2.3 (−3.1)
percentage points for the high income economy in the long-run, and 3.8 (−4.9) in the
middle-income economy– and reduced misallocation of talent. The middle-income
economy registers greater gains in these indicators largely due to a higher initial µ
value, and a lower initial base in terms of specialised labour. By contrast, the high-
income economy, with a relatively more effi cient production structure, benefits from
higher gains in terms of the share of specialised labour employed in the innovation
sector and the growth rate of final output, which increases by 0.5 percentage points.
Nevertheless, the change in welfare remains negative in both cases, and in the long
run both types of economies suffer from a higher unemployment rate for specialised
labour– the oversupply problem discussed earlier.
In this setting, the response to this issue is to either a) lower supply, by reducing
incentives to accumulate human capital, or b) expand demand, by implementing
additional policies. Regarding a), making the reduction in the cost of training in
Programme B smaller obviously leads to lower specialised unemployment in the
long run.32 More interesting in the current economic context is to focus on b), by
considering next whether a concomitant increase in public investment may provide
the required stimulus.
2.7.2 Infrastructure Investment
We now consider whether comprehensive labour market reform programmes perform
better when accompanied by an increase in public infrastructure investment. The
important point about this type of spending is that it has both demand-side effects
(in the short run) and supply-side effects (in the longer run) by boosting directly
the economy’s capacity to produce and by stimulating private investment through a
higher marginal product of capital. In addition, in our setting, improved access to
32Specifically, reducing µ by 2 percent (instead of 5 percent), from 0.120 to 0.118, shows thatthe specialised unemployment rate for the middle-income economy increases by only 0.6 percentagepoints (compared to 2.2 percentage points in Table 2.5).
155
infrastructure helps to promote innovation activity, especially through its impact on
knowledge networks, as stressed in the recent literature.33 In that sense, therefore,
the provision of public capital is also a productivity-enhancing measure for research
activities.
To examine this issue, an additional reform programme is considered: Programme
C, which adds to Programme B a 20 percent increase in the share of public spending
on infrastructure, υI , from 0.05 to 0.06 for the high-income economy and from 0.069
to 0.083 for the middle-income economy. The impact and long-run effects of this
programme are shown also in Table 2.5 and the transitional dynamics are displayed
in Figure 2.5. The results show that for the high-income economy, the specialised
unemployment rate now falls, both in the short and the long run. However, for the
middle-income economy, the absorption problem is only slightly mitigated. More-
over, the change in welfare remains negative and of the same order as in Programme
B, for both types of economies.
As noted earlier, addressing the labour absorption issue could be achieved by
mitigating incentives to acquire skills (namely, by keeping the cost of training high).
The question here is whether more ambitious policies aimed at increasing labour
demand in both the innovation and final good sectors can prevent a rise in specialised
unemployment—even when training costs are lowered by as much as before. Indeed,
consider Programme C and suppose that public investment in infrastructure is now
increased from 2.0 percent of GDP to 6.2 percent—which translates into an increase
in υI from 6.9 percent of noninterest public expenditure to 21.0 percent. This value
is consistent with the upper range of estimates reported by Serebrisky et al. (2015, p.
7) and deemed necessary in a number of policy reports to eliminate Latin America’s
infrastructure gap with respect to East Asia. In addition, suppose that through
governance reforms, public investment effi ciency, as measured by ϕ, is increased in
all countries from 0.4 to the level of Brazil’s, as estimated by Dabla-Norris et al.
(2012, Table 1), that is, 0.78. The higher stock of public capital contributes to
higher productivity in both the final good and innovation sectors, which improves
the middle-income economy’s ability to absorb specialised labour. In addition to
significantly higher long-run growth (from 0.3 percentage points in Table 2.5 to 2.4
33See Agénor (2016) and the references therein. The effects of an increase in public investment,considered in isolation, are shown in Tables 2.3 and 2.4; these effects are faily muted in the caseof the middle-income economy and show again conflicting effects on growth and welfare.
156
points), this programme leads to an increase in specialised unemployment of only
1.1 percentage points (compared to 2.6 in Table 2.5). However, this combination
of policies does not solve the absorption problem.34 The broader lesson from this
experiment is therefore that, although investing in infrastructure and improving
effi ciency in public spending are important to promote labour demand and growth
in middle-income economies, caution is also needed in promoting higher education
through reductions in training costs, to avoid creating an oversupply of specialised
workers. In many of these countries, improving the quality of education may prove
more effective.
It is worth noting also from Figures 2.3, 2.4, and 2.5 that the transitional dynam-
ics associated with the composite programmes, with or without public investment
in infrastructure, are largely monotonic, except for the growth rate of output which
follows an inverted U-shape– growth accelerates during the first phase of the tran-
sition, but slows down gradually in the second phase. In addition, the adjustment
path is very similar for all the variables shown– except for the wage premium and
the specialised unemployment rate for the middle-income economy when public in-
vestment is added to the composite programmes.
The inverted U-shape pattern of output growth largely reflects the composition
of the reform programmes. During the first phase of the transition, the effects of
policy reforms on skills expansion and employment tend to dominate. The easing
of labour market rigidities (reductions in the minimum wage and union bargaining
power) and active labour market policies (cut in training cost) raise incentives to
acquire advanced skills. At the same time, the drop in the marginal cost of hiring
specialised labour leads to the hiring of more of that type of workers in the final
good and innovation sectors. In addition, under Programme C, improved access
to infrastructure raises labour productivity in both sectors. The combination of
these effects translates into a sharp growth acceleration. During the second phase
of the transition, however, these effects are mitigated. The labour market reforms
lead to an overshooting in specialised wages and therefore to too much specialised
workers in the economy, outpacing the expansion in demand and thereby putting
downward pressure on specialised wages. At the same time, the marginal product
of untrained labour in the final good sector improves, thereby raising the effective
34Moreover, it is an open question as to whether, in practice, a programme involving a permanentincrease in the ratio of investment to GDP to more than 6 percent is sustainable politically.
157
wage of that category of workers. This leads to a reduction in incentives to acquire
skills, and a reduced supply of specialised labour—which in turn rekindles upward
pressures on specialised wages and translates into reduced labour demand in the
innovation sector. The expansion of intermediate varieties therefore decelerates over
time, resulting in a gradual slowdown in output growth.
2.7.3 Policy Externalities
Finally, a question worth asking is to what extent composite reform programmes
generate long-run gains that exceed those generated by independent policies? This
issue can be addressed in a simple manner by adding up the steady-state results
for each individual policy in a composite programme with respect to a particular
set of variables, and comparing the aggregate numbers with those reported in Table
2.5 for the relevant programme. The difference between the latter and the sum of
individual effects gives a measure of interactions between reforms and (depending
on its sign) whether they complement or offset each other, that is, whether policy
externalities are positive or negative.
For instance, for Programme C, for the high-income economy the sum of partial
effects gives a total of 0.0094 for the growth rate (compared to 0.0095 in Table 2.5),
−0.0645 for the aggregate unemployment rate (compared to −0.0522) and −0.1700
for social welfare (compared to −0.1801). For the middle-income economy and for
the same programme, the sum of partial effects gives 0.0029 for the growth rate
(compared to 0.0031), −0.0509 for the aggregate unemployment rate (compared to
−0.0373), and −0.1259 for social welfare (compared to −0.1207). Whether exter-
nalities are positive or negative, the benefits of comprehensive programmes depends
therefore on which outcomes one chooses to focus on; in terms of growth, inte-
grated programmes perform better because they generate positive externalities. In
terms of unemployment or welfare, however, integrated programmes perform worse.
Intuitively, policies aimed at cutting unemployment benefits and diluting union bar-
gaining power for untrained workers tend to be associated with drops in wages and
consumption for the unemployed and untrained groups– —despite the fact that they
are complementary to other policies in promoting innovation and specialised em-
ployment. Similarly, while combining skills expansion policies (cuts in training cost)
with conventional labour market policies tends to create positive externalities in
158
terms of growth and talent allocation, these policies also produce counteracting ef-
fects on the specialised wage premium. Consequently, instead of a complementarity
effect, they generate a negative externality which contributes to weaker outcomes
for the composite programme in terms of its impact on (untrained) unemployment
and social welfare.
2.8 Sensitivity Analysis
The validity of the derived policy implications based on the composite reform pro-
grammes hinges largely on the robustness of the results from the individual policy
experiments. As such, this section presents further discussions on sensitivity analy-
sis of the policies analysed. For each policy discussed, two sets of the most relevant
sensitivity analysis results are presented alongside the benchmark results, for both
types of the economies in Tables 2.6-2.10, with the transitional dynamics presented
in Figures 2.6-2.12. Furthermore, a scenario where the benchmark results are com-
pared against results from a slightly modified model (in which the unemployment
benefits are not factored into expected income for the choice of specialised training)
is also examined, with the relevant results presented in Tables 2.11-2.12. To pre-
view, these additional investigations reaffi rm the robustness of all the benchmark
results, as the signs of the deviations of key variables remain largely similar as in
the benchmark case despite parameter changes.
2.8.1 Reduction in Minimum Wage
For the individual policy of minimum wage reduction (5 percent cut in wU0 ), the key
results documented in the benchmark analysis are that growth and welfare effects
can be positive due to the net positive skills acquisition incentive. In both cases,
these effects are nonetheless small. These remain largely similar in the sensitivity
analysis scenarios presented in Table 2.6.
First, sensitivity test is implemented on the elasticity of untrained wage with
respect to untrained unemployment, κU , where a higher elasticity (compared to
baseline) of 0.24 is examined. For both the high- and middle-income economy, the
primary wage effects of the transmission mechanism from the drop in the shift para-
159
meter, wU0 , are similar to the benchmark case, with the untrained unemployment rate
falls and the expected wage premium increases. Nevertheless, the greater respon-
siveness of untrained wage to untrained unemployment level means the secondary
feedback of an increase in the expected untrained wage is relatively more than the
benchmark case (see (2.38)), therefore leading to a smaller increase in expected
wage premium for the specialised over the untrained. Relative to the benchmark,
the incentive to acquire skills is therefore lower. Consequently, the magnitudes of
the policy effects for all key variables (unemployment rates, proportion of specialised
workers employed in innovation, payroll contribution rate, and the misallocation of
talent) are smaller than in the benchmark case. These smaller job creation and in-
novation expansion effects apply to both impact and the steady state. To illustrate,
the untrained unemployment rate falls by only 1.8 percentage points for the high-
income economy and 1.3 percent for the middle-income economy, both being smaller
than the 2.8 and 2.0 percentage points observed respectively for the benchmark case.
On growth and welfare, the fairly small positive effects remain.
Second, sensitivity analysis is also conducted for the standard production para-
meters (with respect to labour inputs) in the final good sector. Specifically, con-
ditioned on the same labour shares, the elasticity of final good production with
respect to untrained labour, βU , is set at a higher level relative to the elasticity
of specialised workers. This is to reflect a scenario where untrained labour is the
relatively scarce labour input for final good production. In this instance, for both
economies, while the increase in the specialised-untrained wage premium is smaller
than the benchmark case, the expansionary effects associated with the untrained
labour, in terms of the steady-state increase in effective share of specialised workers,
those employed in innovation, and final output growth are marginally better than
in the benchmark case, at the cost of higher (lower reduction in) specialised un-
employment rate for the middle-(high-)income economy. This is mainly due to the
different production elasticity specification for the final good sector that is used now
(favouring untrained labour). In the long-run, while the difference in the change in
final output growth rate is negligible, welfare is worse off for both economies in this
specific scenario due to the comparatively smaller increase in relative specialised
wage yet larger specialised unemployment rate.
160
2.8.2 Reduction in Unemployment Benefit Rates
For the three separate experiments considered with respect to reduction in unem-
ployment benefit rates, the first (a cut in κU only) produces similar general equilib-
rium effects to the previous experiment (a cut in wU0 ). This is despite of the different
transmission mechanism in action for the two shocks (see previous explanations for
the benchmark cases). Not surprisingly, as can be seen in Table 2.7, when sensitivity
analysis is implemented with respect to the two parameters of κU and βU , the broad
direction of difference in magnitudes (both impact and steady-state effects) for all
key variables in both sensitivity exercises is largely similar to the minimum wage
shock.
For the second experiment with a cut in κS only, the corresponding parameters
examined are a higher elasticity of specialised wage with respect to specialised un-
employment, κS (κS = 0.24), and a final good production structure with a higher
elasticity with respect to specialised labour, βS. These two parameters concern-
ing specialised labour are the direct counterparts to those examined for untrained
labour. As expected, the opposing impact and steady-state effects to the κU shock
are observed. For instance, the effective share of specialised workers in the econ-
omy declines due to the overall reduction in expected wage premium– —hence lower
incentive to acquire skills– for the specialised labour. Nevertheless, the cut in spe-
cialised unemployment benefit rate does create a ‘shirking prevention’disciplinary
effect specifically to the specialised labour employed in the innovation sector, and
this boost in labour productivity in the sector eventually results in an increase in
labour demand in that sector, henceby an expansion in the intermediate varieties.
Overall, the general equilibrium effects from these lead to a larger reduction in
specialised unemployment rate for both the high- and middle-income economy.
When the specialised wage-setting is relatively more elastic to change in the spe-
cialised unemployment rate (higher κS), the magnitudes of all the general equilib-
rium effects discussed are smaller for both economies, as in the previous case of a κU
cut. Given that the expected wage premium for specialised labour actually increases
for the high-income economy here, the skills acquisition disincentive is smaller too,
therefore smaller decline (increase) in proportion of specialised (untrained) labour.
The disciplinary effect in the innovation sector, as well as the subsequent general
equilibrium effects leading to the reduction in specialised unemployment rate are
161
also smaller, hence less expansion (reduction) in the share of specialised workers in
innovation (unemployed). The differences in magnitude are nonetheless very small.
For the scenario where the elasticity of final good production with respect to
specialised labour, βS, is set at a larger value, the skills acquisition disincentive is
even smaller than the benchmark case, with smaller increase (decrease) in the pro-
portion of untrained (specialised) labour observed. However, given that βS directly
influences the relative wage ratio in the final good production sector, the change
in untrained labour employed in that sector is actually more sensitive, therefore a
greater reduction in untrained unemployment rate relative to the benchmark case
(unlike the case with specialised unemployment rate). Overall, in this sensitivity
scenario, the welfare effect is predictably worse than the benchmark case for both
economies, given that specialised labour plays a more important production role in
the economy yet is the group that faces the direct hit with respect to their wages.
Lastly, for the experiment of a cut in both types of unemployment benefits,
the same sensitivity analysis scenario with respect to the different specification of
labour elasticity in final good production considered are implemented again. For
the high-income economy, the steady-state effect of expected specialised-untrained
wage premium is positive, whereas the expected wage premium is negative for the
middle-income economy. In other words, while the combined reduction in κS and
κU produces a net skills acquisition disincentive (the general equilibrium effect of
κS cut dominates that of κU) and therefore an increase (decrease) in the share of
untrained (specialised) labour supply, the productivity-enhancing ‘shirking preven-
tion’disciplinary effect in the innovation sector is much stronger for the high-income
economy. This eventually translates to an increase in expected specialised wage pre-
mium (whereas the expected wage premium drops for the middle-income economy).
If we accounted for this difference in results between the high- and middle-income
economy that is observed even in the benchmark case, then the sensitivity analy-
sis produces largely similar results across the two economies. Specifically, if the
final good production is relatively specialised labour-intensive, this specific policy
considered is marginally more effective in reducing untrained unemployment rate
than in the benchmark case. Nevertheless, in such a case where specialised labour
takes up a higher share in final good production, this uniform cut in unemployment
benefit rates is less effective in reducing specialised unemployment, and therefore
162
also produces much larger negative welfare effects than in the benchmark case. The
long-run tradeoff between growth and welfare for this specific policy is consistently
observed.
2.8.3 Reduction in the Union’s Wage Mark-Up
With the benchmark model, we consider a reduction in the mark-up over the target
wage for both untrained and specialised workers, as measured by a 37.5 percent cut
in the parameters, ξU and ξSY , respectively. As seen in (2.36), this influences the
respective actual wage, wUt and wSYt on impact and subsequently, the respective
expected wages in the long-run. In the benchmark analysis, for the high-income
economy we see that unemployment rates for the two types of workers fall in both
cases, whereas for the middle-income economy ξU (ξSY ) cut is effective only on
untrained (specialised) unemployment reduction, but has negligible effect on spe-
cialised (untrained) unemployment rate due to smaller gross complementarity in
production. In both economies, the growth effect is positive (stronger on impact,
small in the long-run), while the welfare effect is positive for the ξU cut but negative
for the ξSY cut. These observations remain robust to the sensitivity analysis scenar-
ios considered in Table 2.8, save for the scenario when the ξU cut is implemented in
a setting where untrained labour is the relatively scarce labour input for final good
production (in which case welfare effect becomes negative).
For the experiment with respect to ξU , the sensitivity analyses considered are a
higher elasticity of untrained wage with respect to untrained unemployment (κU =
0.24) and a relatively higher final good production elasticity with respect to un-
trained labour (for example, βU = 0.4, βS = 0.2 is set for high-income economy);
whereas for the ξSY experiment, the sensitivity analyses considered are a higher
elasticity of specialised wage with respect to specialised unemployment (κS = 0.24)
and a higher final good production elasticity with respect to specialised labour (for
the same high-income economy example, βU = 0.2, βS = 0.4 is set).
For the ξU experiments, the sensitivity results are similar to those observed for
the untrained minimum wage shock. This is expected since both shocks do result
in the reduction of the cost of untrained labour, the expansion of specialised labour
supply, as well as the increase in untrained and specialised labour demand across all
sectors. When the untrained wage elasticity to untrained unemployment is higher at
163
κU = 0.24, even though on impact the untrained wage falls due to lower mark-up, the
steady-state effect from the feedback of a lower unemployment level to the untrained
wage is higher than in the benchmark case. As such, the corresponding change in
untrained wage is also more elastic. This in turn means that the decline in untrained
wages is smaller, which translates to a smaller increase in specialised wage premium
relative to the benchmark case. The steady-state effects on reducing both specialised
and untrained unemployment rate are therefore also relatively smaller, especially on
the latter (for high-income economy, θULt has declined only by 0.9 percent, whereas
θULt declined by 1.5 percent in the benchmark case; for middle-income economy, it is
a 1.0 percent drop against a 1.5 percent drop in the benchmark case). The welfare
effect is therefore also smaller in both economies.
When a relatively higher final good production elasticity with respect to un-
trained labour is set, the policy effect on reducing untrained unemployment rate is
marginally more significant than in the benchmark case. However, with the new
final good production elasticities, the effectiveness of this policy in reducing spe-
cialised unemployment rate is less significant than in the benchmark case. Indeed,
for the middle-income economy, the slight increase in specialised unemployment rate
observed in the benchmark case actually becomes larger in this setting, as the ex-
pansion in specialised labour supply is not fully absorbed by demand. Similar to the
case with minimum wage experiment earlier, the welfare effect is actually negative
for this specific setting.
For the ξSY experiments, as seen in the benchmark analysis, the opposite results
to the untrained union wage mark-up reduction are observed, and there is a long-run
growth-welfare tradeoff (with the latter being negative). For the sensitivity analysis
where the specialised wage elasticity to specialised unemployment is set higher at
κS = 0.24, results with similarly less significant magnitude are observed. The expla-
nations are similar to the case of κU = 0.24 for untrained union wage mark-up, with
the policy effects on reducing specialised unemployment rate being relatively smaller
than in the benchmark case (since specialised wage is more responsive to a change in
its unemployment rate). Lastly, for the sensitivity analysis with a relatively higher
final good production elasticity with respect to specialised labour, the reduction
(increase) in specialised (untrained) labour supply is predictably smaller. However,
given that this specific policy directly reduces wage of specialised labour employed
164
in the final good sector but only indirectly on those employed in the innovation
sector, its specialised unemployment-reducing effect is also slightly lower than the
benchmark case. In terms of growth and welfare, in this setting where specialised
labour is valued more in final good production, the overall effects on growth and
welfare are marginally better than the benchmark case. The growth-welfare tradeoff
is observed again.
2.8.4 Reduction in Training Cost
In the benchmark analysis for the training cost reduction experiment, the key find-
ings derived are that: (i) it is most effective in generating a large increase in the sup-
ply of specialised workers; (ii) as much as it is a powerful tool in reducing untrained
unemployment and talent misallocation, it also leads to an increase in specialised
unemployment rate, hence creating an absorption problem; and (iii) moderately
positive growth and welfare effects for both types of economies. The various bench-
mark findings outlined are again, robust to the sensitivity analyses considered. Two
parameters are examined: (i) a higher training parameter value, χ = 0.95; and (ii)
a higher elasticity of specialised wage with respect to specialised unemployment, κS
(κS = 0.24). The former examines the responsiveness with respect to the train-
ing process, while the latter concerns the sensitivity of the specialised wage-setting
process. The results are presented in Table 2.9.
In comparison to the benchmark, when χ is higher at χ = 0.95, any individual
with a given ability level is able to become specialised more easily. As such, this
makes the specific policy of training cost cut less effective in promoting skills ac-
quisition relative to the benchmark case. This is reflected by a marginally smaller
reduction in the share of untrained workers in the economy, as well as smaller effects
on unemployment reduction, talent misallocation improvement, and output growth,
both on impact and in the steady state for both economies. However, as the decline
in expected wage premium is now smaller (when χ = 0.95), the welfare effect is
more positive in this specific scenario.
165
2.8.5 Composite Reform Programmes
For all three Composite Programmes discussed previously, sensitivity analysis is
conducted with respect to two key parameters: (i) a higher elasticity of untrained
wage with respect to untrained unemployment, at κU = 0.24; and (ii) a higher
elasticity of production with respect to labour input in the innovation sector (λ =
0.7). Given that untrained unemployment and innovation-driven growth are the
salient features of the model, these two parameters therefore provide good means to
assess the robustness of the results.
For the core program, Programme A, a larger κU means that untrained wage-
setting is more responsive to change in untrained unemployment rate than in the
benchmark case. As seen in the benchmark analysis, untrained unemployment rate
falls on impact and in steady state for both economies. In this setting, the up-
ward adjustment to untrained wage is more responsive than in the benchmark case,
and the increase in expected wage premium is therefore smaller. This leads to
relatively lower incentives to acquire skills, and therefore less effective in reducing
untrained labour supply. As such, the subsequent policy effects in reducing unem-
ployment are also comparatively lower than in the benchmark case. For instance,
in the high-income economy, the untrained unemployment rate falls by 4.4 percent-
age points (falls by 6.5 points in benchmark case); for the middle-income economy,
the untrained unemployment rate falls by 3.3 percentage points (falls by 4.9 points
in benchmark). The positive effect on growth is also marginally lower for both
economies. The negative welfare effects, and therefore the long-run growth-welfare
tradeoff, stay robust for both economies.
For the case with λ = 0.7, compared to the benchmark case of Program A, the
policy results are similar for the untrained unemployment rate, growth, and wel-
fare. This is unsurprising as the program consists of pure labour market reform
policies. However, in an environment with a higher λ, the secondary effects onto the
specialised labour-related indicators are relatively stronger than in the benchmark
case. This is reflected in a slightly larger reduction in long-run specialised unem-
ployment rate, and a greater reduction in misallocation of talent for both economies.
Programme B, which adds skills expansion policies to pure labour market reforms,
is also examined using the two parameters discussed. For κU = 0.24, the sensitivity
analysis results are largely consistent with those observed for Programme A, in
166
that for both impact and steady-state, the untrained unemployment rate reduction,
the specialised labour supply expansion, the decline in payroll contribution rate
and talent misallocation, and the positive output growth effect are all in smaller
magnitudes compared to the benchmark for both economies. When the elasticity
of production with respect to labour input in the innovation sector is set higher at
λ = 0.7, for Programme B, the steady-state increase in specialised unemployment
rate is smaller compared to the benchmark case for both economies. Lastly, for both
Programme A and B, it is worth noting again that the long-run growth-welfare
tradeoff associated with labour market reforms is persistent across the different
sensitivity scenarios.
For the programme with public infrastructure investment, Programme C, sensi-
tivity analysis is conducted with respect to: (i) a higher elasticity of untrained wage
with respect to untrained unemployment, at κU = 0.24; and (ii) a higher elasticity
of production with respect to labour input in the innovation sector (λ = 0.7). Again,
consistent results are observed, hence reaffi rming the robustness of the benchmark
results. As seen in Table 2.10, in the case of κU = 0.24, the effectiveness of the
composite programs is marginally lower for the high-income economy relative to the
benchmark case (in terms of reducing unemployment and raising growth). However,
for the middle-income economy, the positive deviation in specialised unemployment
rate is actually lower compared to the benchmark case.
In the setting of λ = 0.7, the overall general equilibrium effects associated with
key labour market variables in both economies, such as the untrained and spe-
cialised unemployment rates, the index of talent misallocation, the expansion of
effective specialised labour employed in the innovation sector, represent an improve-
ment compared to the benchmark case. However, for the middle-income economy,
the change in specialised unemployment rate remains positive. Payroll contribution
is also lower, which translates to a higher final output growth rate for the high in-
come economy (for the less effi cient middle-income economy, the difference from the
benchmark case is insignificant). In terms of social welfare, again, in this specific set-
ting (λ = 0.7) the huge increase in actual specialised wages (primarily wSRt ) exudes
a negative welfare effect on the specialised unemployed individuals. As specialised
wages increase significantly, the training cost incurred upon specialised individuals
who couldn’t find a job will increase significantly as well. This, couple with the ulti-
167
mately lower wages for the untrained workers, results in an overall negative welfare
effect.
2.8.6 Model Without Unemployment Benefit Consideration
for Training
An additional sensitivity analysis is also implemented, where unemployment benefits
are not factored into an individual’s expected earnings when deciding on whether
to undergo advanced training. Specifically, (2.5) is merely specified as
(1− ε)(ζSYt wSYt + ζSRt wSRt )− tct ≥ (1− ζULt )wUt ,
with training cost, tct, being proportionate to the expected specialised wage.
In this model, the threshold level of ability aCt for specialised training decision is
given by the simplified form of
aCt = µ1/χ
1− (1− ε)−1 (1− ζULt )wUt
ζSYt wSYt + ζSRt wSRt
−1/χ
.
Without the additional feedback from unemployment benefits, for both the high-
and middle-income economy, it can be seen from Table 2.11 and 2.12 that conven-
tional labour market policies, such as the reduction in minimum wage, the cut in
untrained unemployment benefit rate, and the reduction in untrained union wage
mark-up are less effective (relative to benchmark model) in promoting specialised
labour supply expansion, the increase of the effective share of specialised workers
in innovation, the reduction in misallocation of talent, and the cut in untrained un-
employment rate. However, for specialised unemployment rate, implementing these
policies in this model would result in greater steady-state reduction in specialised
unemployment rate. On the other hand, for the policies of a reduction in specialised
workers’unemployment benefit rate and the cut in specialised union wage mark-up,
the opposite is observed.
For the training cost cut, for the high-income economy where the training process
is more effi cient, the specific policy is clearly more effective in raising specialised
labour supply in the benchmark model when there is additional feedback from the
168
unemployment benefits. However, the oversupply problem is also more severe in
both economies. In terms of output growth, the positive output growth effect for
this modified model is also lower compared to the benchmark model.
For the Composite Programmes, the overall policy effects are generally milder
without unemployment benefit consideration. The only exceptions being the spe-
cialised unemployment rate and welfare, which show improvements instead when
unemployment benefits are removed from specialised skills acquisition consideration.
Indeed, these remain true for the sensitivity scenarios of the composite programme
with public infrastructure investment, Programme C. Thus, this reaffi rms the impor-
tance in accounting for unemployment benefits consideration when studying skills
acquisition decision– and therefore labour supply– in a general equilibrium model
with various labour market rigidities. What would appear to a good enough policy
in improving household well-being may be inadequate if there were much richer feed-
back (due to the unemployment benefit consideration) into the individuals’labour
supply decisions.
2.9 Concluding Remarks
The main implications of this chapter have been summarised in the introduction and
need not to be repeated again. It is therefore concluded that, for future research
extension, the model could be extended to account for other types of labour market
distortions, such as state-contingent firing costs and severance payments, deskilling
of the labour force associated with unemployment, as well as a positive effect of
a higher share of more educated workers on life expectancy and savings (and thus
on economic growth), and various other forms of active labour market policies (see
Almeida et al. (2012)). In particular, hiring and firing regulations, and hiring
costs, have been shown to have an adverse effect on unemployment, especially when
search and matching considerations are important;35 their implications for growth
and welfare, however, are less well understood.
A more systematic effort to integrate political economy considerations in as-
sessing the performance of labour market reforms in growth models would also be
warranted. Observers have often argued that the costs of these reforms are incurred
35See Bernal-Verdugo et al. (2012) and Millán et al. (2014) for some supportive evidence.
169
up front and concentrated on specific groups, whereas their benefits materialise
much later and are both more diffuse and less predictably allocated among workers
and households. In addition, conflicting growth and welfare effects may well lead
to organised resistance to reform. A key challenge then is to create the political
consensus needed to confront powerful vested interests and mitigate the dynamic
trade-offs between (short-term) costs and (longer-term) gains.
At the same time, if specific labour market reforms do not produce substantial
economic benefits– as suggested by our numerical experiments– political viability
may well require reform programs to eschew them and focus instead on upfront
measures that matter more for productivity, examples of which include investment
in infrastructure. Put differently, with limited political capital and little capacity to
compensate losers in the short run, pursuing a wide array of labour market reforms
at once may prove costly and ineffective, especially in middle-income economies
where administrative and governance capacity are weak.
170
2.10 Appendix
2.10.1 Dynamic Form
From (2.2) and (2.4), the household’s consolidated budget constraint is, for untrained
individuals who can be either employed (j = E) or unemployed (j = L),
cU,jt|t +cU,jt|t+1
1 + rt+1
=
(1− τ)wUt
bUt
if j = E
if j = L. (2.66)
From (2.3) and (2.4), for those specialised individuals who can work either in
the final good sector or in the innovation sector, or unemployed,
ch,jt|t +ch,jt|t+1
1 + rt+1
=
(1− τ)(1− ε)wht − tct
(1− ε)bSt − tct
if j = E, h = SY, SR
if j = L.
Equivalently, using (2.6),
ch,jt|t +ch,jt|t+1
1 + rt+1
=
(1− τ)(1− ε)wht − tct
(1− ε)bSt − tct
if j = E
if j = L. (2.67)
Each individual maximises (2.1) subject to the intertemporal budget constraint
(2.66) or (2.67). The first-order conditions give the standard Euler equation
ch,jt|t+1
ch,jt|t=
1 + rt+1
ηC(1 + ρ). h = U, SY, SR, j = E,L (2.68)
Substituting this result in (2.66) and (2.67) yields
cU,jt|t = [ηC(1 + ρ)
1 + ηC(1 + ρ)]
(1− τ)wUt
bUt
if j = E
if j = L,
171
cSY,jt|t = [ηC(1 + ρ)
1 + ηC(1 + ρ)]
(1− τ)(1− ε)wSYt − tct
(1− ε)bSt − tct
if j = E
if j = L,
cSR,,jt|t = [ηC(1 + ρ)
1 + ηC(1 + ρ)]
(1− τ)(1− ε)wSRt − tct
(1− ε)bSt − tct
if j = E
if j = L,
or equivalently,
cU,jt|t = (1− σ)
(1− τ)wUt
bUt
if j = E
if j = L,
cSY,jt|t = (1− σ)
(1− τ)(1− ε)wSYt − tct
(1− ε)bSt − tct
if j = E
if j = L,
cSR,jt|t = (1− σ)
(1− τ)(1− ε)wSRt − tct
(1− ε)bSt − tct
if j = E
if j = L,
so that,
sU,jt|t = σ
(1− τ)wUt
bUt
if j = E
if j = L, (2.69)
sSY,jt|t = σ
(1− τ)(1− ε)wSYt − tct
(1− ε)bSt − tct
if j = E
if j = L, (2.70)
sSR,jt|t = σ
(1− τ)(1− ε)wSRt − tct
(1− ε)bSt − tct
if j = E
if j = L, (2.71)
where σ = 1/ [1 + ηC(1 + ρ)] < 1.
Given (2.4), (2.69), and (2.70), we can also write
cU,jt|t+1 = σ(1 + rt+1)
(1− τ)wUt
bUt
if j = E
if j = L, (2.72)
172
cSY,jt|t+1 = σ(1 + rt+1)
(1− τ)(1− ε)wSYt − tct
(1− ε)bSt − tct
if j = E
if j = L, (2.73)
cSR,jt|t+1 = σ(1 + rt+1)
(1− τ)(1− ε)wSRt − tct
(1− ε)bSt − tct
if j = E
if j = L. (2.74)
With the consumptions derived for both time t and t+ 1, substituting them into
(2.1) yields the indirect utility function for the four states:
V U,Et = ηC ln(1− σ)(1− τ)wUt +
lnσ(1 + rt+1)(1− τ)wUt1 + ρ
, (2.75)
V U,Lt = ηC ln(1− σ)bUt +
lnσ(1 + rt+1)bUt1 + ρ
, (2.76)
V SY,Et = ηC ln(1− σ)[(1− τ)(1− ε)wSYt − tct] (2.77)
+lnσ(1 + rt+1)[(1− τ)(1− ε)wSYt − tct]
1 + ρ,
V SR,Et = ηC ln(1− σ)[(1− τ)(1− ε)wSRt − tct] (2.78)
+lnσ(1 + rt+1)[(1− τ)(1− ε)wSRt − tct]
1 + ρ,
V S,Lt = ηC ln(1− σ)[(1− ε)bSt − tct] (2.79)
+lnσ(1 + rt+1)[(1− ε)bSt − tct]
1 + ρ.
To determine the decision to acquire training, note that in equilibrium the indi-
vidual is indifferent between his two options, so that
(1−ε)[ζSYt (1− τ)wSYt + ζSRt (1− τ)wSRt + ζSLt bSt ]− tct = (1− ζULt )(1− τ)wUt + ζULt bUt .
Substituting (2.6) into this equation and rearranging gives
(1−ε)(1− µ
aχ)[ζSYt (1−τ)wSYt +ζSRt (1−τ)wSRt ] = (1−ζULt )(1−τ)wUt +ζULt bUt −(1−ε)ζSLt bSt ,
which after further algebraic manipulation, allows us to derive an expression for
173
the threshold ability value, aCt , such that all individuals with ability lower than aCt
would opt to remain untrained:
aCt = µ1/χ
1− (1− ε)−1 (1− ζULt )(1− τ)wUt + ζULt bUt − (1− ε)ζSLt bSt
(1− τ)[ζSYt wSYt + ζSRt wSRt ]
−1/χ
.
(2.80)
Using equation (2.29) to substitute out wSRt , and subsequently (2.35) to substi-
tute out both unemployment benefits, bUt and bSt , (2.80) can be rewritten as
aCt = µ1/χ
1− (1− ε)−1 (1− ζULt )(1− τ)wUt + [ζULt κU − (1− ε)ζSLt κS](Yt/N)
(1− τ)[ζSYt wSYt + κRζSRt ζSYt wSYt + κRζSRt ζSLt κS(Yt/N)]
−1/χ
.
Then, using (2.38) and (2.39), and knowing that ζUYt = 1−ζULt , we can substitute
Yt/N out and rewrite aCt as
aCt = µ1/χ
1− (1− ε)−1 ζ
UYt (1− τ) + [ζULt κU − (1− ε)ζSLt κS]ΩU(θULt )κ
U
(1− τ)[ζSYt + κRζSRt [ζSYt + (θSt )−1ΩS(θSLt )1+κS ]]
wUtwSYt
−1/χ
,
(2.81)
where ΩU = (1− 2ξU)(1− ξU)−1(wU0 )−1 and ΩS = κS(wSY0 )−1(1− 2ξSY )(1− ξSY )−1.
From the first-order conditions (2.13), we have
wSYt = (βS
1 + ς t)
Yt(1− ε)NSY
t
, (2.82)
and
wUt = (βU
1 + ς t)YtNUYt
. (2.83)
Combining these two equations yields the relative wage ratio as
wUtwSYt
= [(βU
1 + ς t)YtNUYt
][(βS
1 + ς t)
Yt(1− ε)NSY
t
]−1,
or equivalently,wUtwSYt
= [(1− ε)βU
βS](NSYt
NUYt
) = β(θSYtθUYt
), (2.84)
where β = (1− ε)βU/βS.Using (2.84) to substitute out the relative wage ratio in (2.81), with aCt = θUt ,
174
we can express the share of untrained labour as
θUt = aCt = µ1/χ
1− (1− τ)ζUYt + [ζULt κU − (1− ε)ζSLt κS]ΩU(θULt )κ
U
(1− τ)[ζSYt + κRζSRt [ζSYt + (θSt )−1ΩS(θSLt )1+κS ]]
βU
βS(θSYtθUYt
)
−1/χ
.
(2.85)
From (2.11), and with aCt = θUt ,
θSt =1− (θUt )2
2. (2.86)
Next, consider the innovation sector. From (2.22), we have
ARt = (kGt )φR1Mt. (2.87)
Substituting this result into equation (2.21) and dividing by Mt yield
Mt+1
Mt
= 1 + (eR)λ(1− ε)λ(kGt )φR1 (θSRt )λ, (2.88)
where eR = 1− (1− eRm)(κR)−ψ > 0.
From the first-order condition of (2.27), we derive an equilibrium expression
for θSRt by first substituting in (2.29) for wSRt , followed by using (2.35) and Qt =
(1 − η)γ(Yt/Mt) (from (2.18)-(2.20)) to substitute out bSt and Yt/Mt respectively,
yielding:
θSRt =
λγ(1− η)(kGt )φ
R1 (eR)λ(1− ε)λθSt
κR[βS + (1 + ς t)κS(1− ε)θSLt ]
1/(1−λ)
. (2.89)
From the same first-order condition, we can also derive an alternative expression
for wSYt , given by
wSYt =λQt(k
Gt )φ
R1Mt(θ
SRt )λ−1(eR)λ(1− ε)λ−1θSt
κR(1 + ς t)θSYt N
− κS θSLt
θSYt
YtN. (2.90)
Next, equating (2.82) with (2.39), we derive an expression for the share of em-
ployment of specialised labour in the final good sector, θSYt , as
θSYt =βS
wSY0
(1− 2ξS
1− ξS)(
1
1− ε)(1
1 + ς t)(θSLt )κ
S
. (2.91)
175
Using (2.42), the unemployment rate of specialised labour, θSLt , is thus
θSLt = θSt − θSRt − θSYt . (2.92)
Next, consider the untrained labour market. Equating the first-order condition
for the untrained wage, wUt , from (2.83) towUt = wU0 [(1−ξU)/(1−2ξU)](Yt/N)(θULt )−κ
U.
Equating this to (2.38) and subsequent rearrangements allow us to derive an
expression for the untrained unemployment rate:
θULt = (wU0 )(κU )−1(1− 2ξU
1− ξU)−(κU )−1(
βU
1 + ς t)−(κU )−1(θUYt )(κU )−1 . (2.93)
To derive the share of untrained employment in the final good sector, given (2.85)
and (2.93), we derive θUYt residually using
θUYt = θUt − θULt . (2.94)
From (2.6), (2.31), and (2.34), we have
KGt+1 = ϕυIτ
wUt θ
UYt N + (1− ε)θSYt
wSYt −
µ(1− τ)
[0.5(1 + aCt )]χ(ζSYt wSYt + ζSRt wSRt )
N
+(1− ε)θSRtwSRt −
µ(1− τ)
[0.5(1 + aCt )]χ(ζSYt wSYt + ζSRt wSRt )
N
,
where average ability of specialised workers is used.
From (2.29), we know that wSRt = κR(ζSYt wSYt + ζSLt bSt ), which by substituting
in (2.35), is also
wSRt = κRζSYt wSYt + κRκSζSLt (Yt/N).
Then, using again (2.39), as well as (2.43), we get
wSRt =κR
θSt[θSYt + ΩS(θSLt )1+κS ]wSYt , (2.95)
where ΩS = κS(wSY0 )−1(1− 2ξSY )(1− ξSY )−1, which is in turn substituted into the
176
expression for KGt+1, yielding
KGt+1 = ϕυIτ
wUt θUYt N + (1− ε)θSYt
1− µ(1−τ)
[0.5(1+aCt )]χ
× 1θSt
(θSYt + θSRtθStκR[θSYt + ΩS(θSLt )1+κS ])
wSYt N
+(1− ε)θSRt
κR
θSt[θSYt + ΩS(θSLt )1+κS ]− µ(1−τ)
[0.5(1+aCt )]χ
× 1θSt
(θSYt + θSRtθStκR[θSYt + ΩS(θSLt )1+κS ])
wSYt N
,
or equivalently,
KGt+1 = ϕυIτ
wUt θ
UYt N + (1− ε)
θSYt + θSRt
κR
θSt[θSYt + ΩS(θSLt )1+κS ]
− µ(1−τ)
[0.5(1+aCt )]χ[(θSYt + θSRt )/θSt ]
× θSYt + θSRtθStκR[θSYt + ΩS(θSLt )1+κS ]
wSYt N
.
Substituting (2.82) and (2.83) into this equation, yields
KGt+1 = (
ϕυIτ
1 + ς t)Yt
βU +
βS
θSYt[θSYt + θSRt
κR
θSt(θSYt + ΩS(θSLt )1+κS ]
× [1− µ(1−τ)
[0.5(1+aCt )]χ( θ
SYt +θSRtθSt
)]
. (2.96)
Dividing equation (2.96) by KPt yields
KGt+1
KPt
= (ϕυIτ
1 + ς t)YtKPt
βU +
βS
θSYt[θSYt + θSRt
κR
θSt(θSYt + ΩS(θSLt )1+κS ]
× [1− µ(1−τ)
[0.5(1+aCt )]χ( θ
SYt +θSRtθSt
)]
. (2.97)
Now, consider the dynamics of the private capital stock. Equation (2.44) can be
repeated here for convenience:
KPt+1 = (sU,Et NUY
t + sU,Lt NULt ) + (sSY,Et NSY
t + sSR,Et NSRt + sS,Lt NSL
t ).
First, substituting in (2.69), (2.70), (2.71), and then using (2.6), (2.29), (2.35),
and (2.95) in repeated substitutions, we can rewrite the KPt+1 expression as follows:
KPt+1 = σ(1− τ)wUt N
UYt + κU(Yt/N)NUL
t
177
+(1− τ)(1− ε)wSYt NSYt
1− µ(1− τ)
[0.5(1 + aCt )]χ(ζSYt +
ζSRt κR
θSt[θSYt + ΩS(θSLt )1+κS ])
+(1− τ)(1− ε)wSYt NSRt
κR
θSt[θSYt + ΩS(θSLt )1+κS ]
− µ(1−τ)
[0.5(1+aCt )]χ(ζSYt + ζSRt κR
θSt[θSYt + ΩS(θSLt )1+κS ])
+(1−ε)NSL
t
κS(Yt/N)− µ(1− τ)
[0.5(1 + aCt )]χ[ζSYt +
ζSRt κR
θSt[θSYt + ΩS(θSLt )1+κS ]wSYt
.
Then, knowing that Yt/N = κS(wSY0 )−1(1 − 2ξSY )(1 − ξSY )−1, and using also
(2.43), (2.82), and (2.83), we can further simplify the expression to
KPt+1 = (
σ
1 + ς t)Yt(1− τ)βU + κUθULt (1 + ς t)
+(1− τ)βS
θSYt
θSYt −
µ(1− τ)
[0.5(1 + aCt )]χθSYtθSt
(θSYt +θSRt κR
θSt[θSYt + ΩS(θSLt )1+κS ])
+(1− τ)βS
θSYt
κRθSRtθSt
[θSYt + ΩS(θSLt )1+κS ]
− µ(1−τ)
[0.5(1+aCt )]χθSRtθSt
(θSYt + θSRt κR
θSt[θSYt + ΩS(θSLt )1+κS ])
+
ΩSθ
St (θSLt )κ
S − µ(1− τ)
[0.5(1 + aCt )]χ[θSYt + θSRt
κR
θSt(θSYt + ΩS(θSLt )1+κS ]
βSθSLtθSYt θSt
,
or equivalently, when dividing by KPt , gives
KPt+1
KPt
= (σ
1 + ς t)YtKPt
(1− τ)βU + κUθULt (1 + ς t) (2.98)
+(1− τ)
1− µ(1− τ)
[0.5(1 + aCt )]χ(θSYt + θSRt
θSt)
[θSYt + θSRt
κR
θSt(θSYt + ΩS(θSLt )1+κS ]
βS
θSYt
+
ΩSθ
St (θSLt )κ
S − µ(1− τ)
[0.5(1 + aCt )]χ[θSYt + θSRt
κR
θSt(θSYt + ΩS(θSLt )1+κS ]
βSθSLtθSYt θSt
.
Combining (2.97) and (2.98) yields the public-private capital ratio:
kGt =KGt
KPt
. (2.99)
178
Dividing equation (2.88) by (2.98) yields
mt+1 =[1 + (eR)λ(1− ε)λ(kGt )φ
R1 (θSRt )λ
]mt (2.100)
⟨(
σ
1 + ς t)YtKPt
(1− τ)βU + κUθULt (1 + ς t)
+(1− τ)
[1− µ(1− τ)
[0.5(1 + aCt )]χ(θSYt + θSRt
θSt)
][θSYt + θSRt
κR
θSt(θSYt + ΩS(θSLt )1+κS ]
βS
θSYt
+
ΩSθ
St (θSLt )κ
S − µ(1− τ)
[0.5(1 + aCt )]χ[θSYt + θSRt
κR
θSt(θSYt + ΩS(θSLt )1+κS ]
βSθSLtθSYt θSt
⟩−1
.
From (2.33), the unemployment insurance scheme’s budget can be repeated here
for convenience:
ς t =bUt θ
ULt + bSt θ
SLt
wUt θUYt + (1− ε)(wSYt θSYt + wSRt θSRt )
. (2.101)
Substituting (2.95) and (2.35) into (2.101), as well as subsequent algebraic ma-
nipulations where the relationships in (2.43), (2.82), and (2.83) are used, expression
(2.101) can be rewritten into
ς t1 + ς t
=θULt κU + θSLt κS
βU + βS
θSYt[θSYt + ζSRt κR(θSYt + ΩS(θSLt )1+κS ]
,
or equivalently,
ς t =θULt κU + θSLt κS
βU + βS
θSYt[θSYt + ζSRt κR(θSYt + ΩS(θSLt )1+κS ]− (θULt κU + θSLt κS)
, (2.102)
From (2.19) and (2.20),
Qt = (1− η)γ(YtMt
),
or equivalently, noting that Yt/Mt = (Yt/KPt )(mt)
−1,
Qt = (1− η)γ(YtKPt
)(mt)−1, (2.103)
179
where Yt/KPt is derived from equation (2.46):
YtKPt
=(kGt )ω/(1−γ)Λ2
[(θSYt )βS(θUYt )β
U]−1/(1−γ)
mt
(1−η)/ηγ/(1−γ)
, (2.104)
where Λ2 = (1− ε)βSΛγ/(1−γ)1 .
The steady-state growth rate can be calculated using (2.88), as in
1 + γ = (eR)λ(1− ε)λ(kG)φR1 (θSR)
λ
. (2.105)
To determine the level of final output and its growth rate during the transition,
note that equation (2.104) relates the path of Yt to that of KPt . In turn, to derive
the path of KPt , equation (2.98) can be written for period t as
KPt+1
KPt−1
= (σ
1 + ς t−1
)Yt−1
KPt−1
(1− τ)βU + κUθULt−1(1 + ς t−1) (2.106)
+(1−τ)
1− µ(1− τ)
[0.5(1 + aCt−1)]χ(θSYt−1 + θSRt−1
θSt−1
)
[θSYt−1+θSRt−1
κR
θSt−1
(θSYt−1+ΩS(θSLt−1)1+κS ]βS
θSYt−1
+
ΩSθ
St−1(θSLt−1)κ
S − µ(1− τ)
[0.5(1 + aCt−1)]χ[θSYt−1 + θSRt−1
κR
θSt−1
(θSYt−1 + ΩS(θSLt−1)1+κS ]
βSθSLt−1
θSYt−1θSt−1
.
which gives the growth rate of KPt . For any given starting value K
P0 , the path of K
Pt
can be derived from (2.106). Substituting this result in (2.104) gives the solution
for Yt from which its growth rate can be derived.
In summary, the dynamic system that drives the economy is given by:
mt+1 =[1 + (eR)λ(1− ε)λ(kGt )φ
R1 (θSRt )λ
]mt (2.107)
⟨(
σ
1 + ς t)YtKPt
(1− τ)βU + κUθULt (1 + ς t)
+(1− τ)
1− µ(1− τ)
[0.5(1 + aCt )]χ(θSYt + θSRt
θSt)
[θSYt + θSRt
κR
θSt(θSYt + ΩS(θSLt )1+κS ]
βS
θSYt
+
ΩSθ
St (θSLt )κ
S − µ(1− τ)
[0.5(1 + aCt )]χ[θSYt + θSRt
κR
θSt(θSYt + ΩS(θSLt )1+κS ]
βSθSLtθSYt θSt
⟩−1
.
180
KGt+1
KPt
= (ϕυIτ
1 + ς t)YtKPt
βU +
βS
θSYt[θSYt + θSRt
κR
θSt(θSYt + ΩS(θSLt )1+κS ] , (2.108)
× [1− µ(1− τ)
[0.5(1 + aCt )]χ(θSYt + θSRt
θSt)]
KPt+1
KPt
= (σ
1 + ς t)YtKPt
(1− τ)βU + κUθULt (1 + ς t) (2.109)
+(1− τ)
1− µ(1− τ)
[0.5(1 + aCt )]χ(θSYt + θSRt
θSt)
[θSYt + θSRt
κR
θSt(θSYt + ΩS(θSLt )1+κS ]
βS
θSYt
+
ΩSθ
St (θSLt )κ
S − µ(1− τ)
[0.5(1 + aCt )]χ[θSYt + θSRt
κR
θSt(θSYt + ΩS(θSLt )1+κS ]
βSθSLtθSYt θSt
,
together with the static equations
kGt =KGt
KPt
, (2.110)
YtKPt
=(kG)ω/(1−γ)Λ2
[(θSYt )βS(θUYt )β
U]−1/(1−γ)
mt
(1−η)/ηγ/(1−γ)
, (2.111)
Qt = (1− η)γ(YtKPt
)(mt)−1, (2.112)
θUt = µ1/χ
1− ζUYt (1− τ) + [ζULt κU − (1− ε)ζSLt κS]ΩU(θULt )κ
U
(1− τ)[ζSYt + κRζSRt [ζSYt + (θSt )−1ΩS(θSLt )1+κS ]]
βU
βS(θSYtθUYt
)
−1/χ
,
(2.113)
θSt =1− (θUt )2
2, (2.114)
θSRt =
λγ(1− η)(kGt )φ
R1 (eR)λ(1− ε)λθSt
κR[βS + (1 + ς t)κS(1− ε)θSLt ]
1/(1−λ)
, (2.115)
θSYt =βS
wSY0
(1− 2ξS
1− ξS)(
1
1− ε)(1
1 + ς t)(θSLt )(κS), (2.116)
θSLt = θSt − θSRt − θSYt , (2.117)
θUYt = θUt − θULt , (2.118)
181
θULt = (wU0 )(κU )−1(1− 2ξU
1− ξU)−(κU )−1(
βU
1 + ς t)−(κU )−1(θUYt )(κU )−1 , (2.119)
ς t =θULt κU + θSLt κS
βU + βS
θSYt[θSYt + ζSRt κR(θSYt + ΩS(θSLt )1+κS ]− (θULt κU + θSLt κS)
, (2.120)
ζUYt =θUYtθUt
, and ζULt = 1− ζUYt =θULtθUt
, (2.121)
ζSYt =θSYtθSt
, ζSRt =θSRtθSt
, and ζSLt = 1− ζSYt − ζSRt =θSLtθSt
, (2.122)
1 + rt = α(YtKPt
). (2.123)
The properties of this system are discussed in the text.
2.10.2 Welfare Evaluation
The indirect utility functions for the five states are given by
V U,Et = ηC ln(1− σ)(1− τ)wUt +
lnσ(1 + rt+1)(1− τ)wUt1 + ρ
, (2.124)
V U,Lt = ηC ln(1− σ)bUt +
lnσ(1 + rt+1)bUt1 + ρ
, (2.125)
V SY,Et = ηC ln(1− σ)[(1− τ)(1− ε)wSYt − tct] (2.126)
+lnσ(1 + rt+1)[(1− τ)(1− ε)wSYt − tct]
1 + ρ,
V SR,Et = ηC ln(1− σ)[(1− τ)(1− ε)wSRt − tct] (2.127)
+lnσ(1 + rt+1)[(1− τ)(1− ε)wSRt − tct]
1 + ρ,
V S,Lt = ηC ln(1− σ)[(1− ε)bSt − tct] (2.128)
+lnσ(1 + rt+1)[(1− ε)bSt − tct]
1 + ρ.
Using the expressions for tct, wSRt , wSYt , wUt , bSt , and b
Ut from (2.6), (2.29), (2.82),
(2.83), (2.35), and with subsequent algebraic manipulations, we can rewrite (2.124)-
182
(2.128) as
V U,Et = (ηC +
1
1 + ρ) ln
(1− τ)βU
(1 + ς t)θUYt
YtN
+ ΛV , (2.129)
V U,Lt = (ηC +
1
1 + ρ) lnκU
YtN
+ ΛV , (2.130)
V SY,Et = (ηC+
1
1 + ρ) ln
(1− τ)(1− ε)
βS
(1+ςt)(1−ε)θSYt
− µ
(θUt )χ
(1+ζSRt κR)ζSYt βS
(1+ςt)(1−ε)θSYt
+κRκSζSRt ζSLt
YtN
+ΛV ,
(2.131)
V SR,Et = (ηC+
1
1 + ρ) ln
(1− τ)(1− ε)
(
κRβSζSYt(1+ςt)(1−ε)θSYt
+ κRκSζSLt
)− µ
(θUt )χ
(1+ζSRt κR)ζSYt βS
(1+ςt)(1−ε)θSYt
+κRκSζSRt ζSLt
YtN
+ΛV ,
(2.132)
V S,Lt = (ηC +
1
1 + ρ) ln
(1− ε)
κS − µ(1− τ)
(θUt )χ
(1+ζSRt κR)ζSYt βS
(1+ςt)(1−ε)θSYt
+κRκSζSRt ζSLt
YtN
+ ΛV ,
(2.133)
respectively, where ΛV = ηC ln(1− σ) + (1 + ρ)−1 lnσ(1 + rt+1).
Define now a social welfare function where with perfect foresight, the welfare of all
future generations of individuals are accounted for. The welfare of individuals in the
five states in each generation are expressed by the indirect utility functions, (2.129)-
(2.133). Assuming that the welfare criterion is equally weighted across the five states
within each generation, the social welfare function is given by the discounted sum of
utility across an infinite sequence of generations (see De la Croix and Michel, 2002,
p. 91):
Wt =
∞∑h=0
Λh
0.2V U,Et+h + 0.2V U,L
t+h + 0.2V SY,Et+h + 0.2V SR,E
t+h + 0.2V S,Lt+h
, (2.134)
where Λ ε (0, 1) is the social discount factor. Along the balanced growth path,
θUt+h = θU , θUYt+h = θUY , θSYt+h = θSY , ζSRt+h = ζSR, ζSYt+h = ζSY , ζSLt+h = ζSL, and
ς t+h = ς in the indirect utility functions in (2.134). With N = 1, social welfare is
therefore driven primarily by final output, Yt+h, which grows at the rate of 1 + g
183
along the balanced growth path: Yt+h = (1 + g)t+hY0, where Y0 can be normalised
to one.
The social welfare function can thus be expressed as
Wt =∞∑h=0
Λh
0.2(ηC +
1
1 + ρ)ln (1− τ)βU
(1 + ς)θUY+ lnκU (2.135)
+ ln(1− τ)(1− ε)
βS
(1+ς)(1−ε)θSY
− µ
(θU )χ
((1+ζSRκR)ζSY βS
(1+ς)(1−ε)θSY + κRκS ζSRζSL)
+ ln(1− τ)(1− ε)
(
κRβS ζSY
(1+ς)(1−ε)θSY + κRκS ζSL)
− µ
(θU )χ
((1+ζSRκR)ζSY βS
(1+ς)(1−ε)θSY + κRκS ζSRζSL)
+ ln(1− ε)[κS − µ(1− τ)
(θU)χ
((1 + ζSRκR)ζSY βS
(1 + ς)(1− ε)θSY+ κRκS ζSRζSL
)]
+ΛV + (t+ h) ln(1 + g).
Given that Λ < 1, Wt is strictly concave and bounded, subject to the usual
convex and compact choice set.
Based on (2.135), an optimal social welfare value can be approximated by36
Wt '0.2
1− Λ(ηC +
1
1 + ρ)
ln
(1− τ)βU
(1 + ς)θUY+ lnκU (2.136)
+ ln(1− τ)(1− ε)
βS
(1+ς)(1−ε)θSY
− µ
(θU )χ
((1+ζSRκR)ζSY βS
(1+ς)(1−ε)θSY + κRκS ζSRζSL)
+ ln(1− τ)(1− ε)
(
κRβS ζSY
(1+ς)(1−ε)θSY + κRκS ζSL)
− µ
(θU )χ
((1+ζSRκR)ζSY βS
(1+ς)(1−ε)θSY + κRκS ζSRζSL)
+ ln(1− ε)[κS − µ(1− τ)
(θU)χ
((1 + ζSRκR)ζSY βS
(1 + ς)(1− ε)θSY+ κRκS ζSRζSL
)]36The derivation involves using a standard approximation result,
∑∞h=0 hx
h ' x/(x− 1)2.
184
+ΛV
1− Λ+
Λ
(Λ− 1)2ln(1 + g),
which is the form reported in the simulation results.
2.11 Tables and Figures
Table 2.1: Calibrated Parameter Values: Benchmark Case
Parameter Description High Income Middle Income
Householdsρ Intergenerational discount rate 0.375 0.375σ Household savings rate 0.109 0.138χ Productivity parameter (effi ciency of training) 0.9 0.5µ Advanced education cost 0.08 0.12ε Time allocated to schooling activity 0.115 0.123
Productionω Elasticity wrt public-private capital ratio 0.17 0.24βS Elasticity wrt specialised workers 0.3 0.35βU Elasticity wrt untrained workers 0.3 0.2α Elasticity wrt private capital 0.3 0.35γ Elasticity wrt intermediate input 0.1 0.1η Substitution parameter, intermediate goods 0.61 0.25φR1 Elasticity wrt public infrastructure 0.186 0.300π Probability of being caught shirking 0.078 0.048δR Elasticity wrt wage for innovation 0.9 0.9λ Elasticity of production wrt labor input 0.6 0.6ψ Elasticity of effort wrt relative wages 0.70 0.43
Governmentτ Tax rate on total wages 0.198 0.123υI Share of spending on infrastructure 0.050 0.069ϕ Effi ciency parameter, public investment 0.5 0.4
Labor marketκS Specialised labor, unemp. benefit indexation 0.4 0.4κU Untrained labor, unemp. benefit indexation 0.4 0.4ξU Relative weight, untrained workers 0.06 0.08ξSY Relative weight, specialised workers 0.06 0.08wU0 Minimum wage indexation, untrained workers 0.522 0.546wSY0 Minimum wage indexation, specialised workers 0.740 0.699κU Elasticity wrt unemployment, untrained wage 0.12 0.12κS Elasticity wrt unemployment, specialised wage 0.12 0.12
185
Table 2.2: Initial Steady-State Values of Key Variables
Variable Description High MiddleIncome Income
θU Share of untrained workers in population 0.732 0.795θS Share of effective specialised workers in population 0.232 0.184θSR Share of effective specialised workers in innovation 0.019 0.004θSY Share of effective specialised workers in final good 0.145 0.109θUY Share of untrained workers in final good sector 0.606 0.708θUL Untrained unemployment rate 0.126 0.087θSL Specialised unemployment rate 0.068 0.071
(θR−θSR)/θSY Index of misallocation of talent 0.189 0.392ζSL Probability of specialised workers getting unemployed 0.293 0.385ζSY Prob. of specialised workers employed in final good 0.623 0.593ζSR Prob. of specialised workers employed in innovation 0.084 0.022ζUL Prob. of untrained workers getting unemployed 0.172 0.110ζUY Prob. of untrained workers getting employed 0.828 0.890ς Firms’payroll contribution rate 0.126 0.052
wU/wSweighted Relative wage ratio 0.550 0.750
kG Public-private capital ratio 0.189 0.147
Y/KP Final output-private capital ratio 0.286 0.429m Stock of innovation-private capital ratio 0.100 0.100
186
Table 2.3: High-Income Economy: Summary of Benchmark Policy Results
Steady state
value Impact Steady State Impact Steady State Impact Steady StateShare of untrained workers 0.732 0.0006 0.0027 0.0006 0.0017 0.0010 0.0025Effective share of specialised workers 0.232 0.0004 0.0020 0.0004 0.0013 0.0007 0.0018Share of specialised workers in innovation 0.019 0.0001 0.0008 0.0002 0.0005 0.0004 0.0003Expected wage premium 0.818 0.0073 0.0209 0.0154 0.0134 0.0173 0.0084Index of misallocation of talent 0.189 0.0016 0.0086 0.0025 0.0055 0.0031 0.0024Untrained unemployment rate 0.126 0.0114 0.0276 0.0014 0.0058 0.0004 0.0009Specialised unemployment rate 0.068 0.0003 0.0013 0.0005 0.0008 0.0013 0.0024Total unemployment rate 0.106 0.0081 0.0198 0.0011 0.0043 0.0005 0.0012Payroll contribution rate 0.126 0.0052 0.0211 0.0064 0.0137 0.0037 0.0073Growth rate of final output 0.008 0.0073 0.0011 0.0049 0.0007 0.0062 0.0004Social welfare 1.000 0.0053 0.0868 0.1704
Steady state
value Impact Steady State Impact Steady State Impact Steady StateShare of untrained workers 0.732 0.0004 0.0007 0.0003 0.0014 0.0005 0.0012Effective share of specialised workers 0.232 0.0003 0.0005 0.0002 0.0010 0.0003 0.0009Share of specialised workers in innovation 0.019 0.0006 0.0008 0.0001 0.0004 0.0001 0.0001Expected wage premium 0.818 0.0019 0.0050 0.0038 0.0110 0.0083 0.0039Index of misallocation of talent 0.189 0.0056 0.0078 0.0009 0.0046 0.0048 0.0049Untrained unemployment rate 0.126 0.0018 0.0066 0.0060 0.0146 0.0001 0.0004Specialised unemployment rate 0.068 0.0018 0.0032 0.0002 0.0007 0.0036 0.0041Total unemployment rate 0.106 0.0017 0.0054 0.0042 0.0105 0.0010 0.0013Payroll contribution rate 0.126 0.0102 0.0206 0.0027 0.0113 0.0016 0.0034Growth rate of final output 0.008 0.0110 0.0011 0.0038 0.0006 0.0094 0.0002Social welfare 1.000 0.2414 0.0027 0.0187
Steady state
value Impact Steady State Impact Steady StateShare of untrained workers 0.732 0.0110 0.0295 0.0003 0.0010Effective share of specialised workers 0.232 0.0080 0.0211 0.0003 0.0007Share of specialised workers in innovation 0.019 0.0014 0.0038 0.0019 0.0020Expected wage premium 0.818 0.0549 0.1241 0.0147 0.0131Index of misallocation of talent 0.189 0.0110 0.0309 0.0127 0.0141Untrained unemployment rate 0.126 0.0037 0.0203 0.0004 0.0018Specialised unemployment rate 0.068 0.0054 0.0133 0.0014 0.0014Total unemployment rate 0.106 0.0015 0.0108 0.0006 0.0017Payroll contribution rate 0.126 0.0004 0.0065 0.0014 0.0035Growth rate of final output 0.008 0.0005 0.0042 0.0346 0.0036Social welfare 1.000 0.0075 0.0024
*/ The respective individual policy shocks are: Reduction in wU? by 5 percent; κU reduced by 10 percent;
κS reduced by 10 percent; both κU and κS cut by 10 percent; ξU reduced by 37.5 percent; ξSY reduced by 37.5 percent; a decrease in advanced education cost by 5 percent; and an increase in share of public infrastructure investment by 20 percent.
Source: Authors' calculations.
HighIncome Economy, Policy Results: Absolute deviations from baseline*
Advanced EducationCost Cut
Increase in PublicInfrastructure Investment
Reduction in BaseMinimum Wage
Reduction in UntrainedWorkers' UB Indexation
Reduction in SpecialisedWorkers' UB Indexation
Reduction in both UBIndexation Parameters
Reduction in UntrainedWorkers' Union Markup
Reduction in SpecialisedWorkers' Union Markup
187
Table 2.4: Middle-Income Economy: Summary of Benchmark Policy Results
Steady state
value Impact Steady State Impact Steady State Impact Steady StateShare of untrained workers 0.795 0.0005 0.0014 0.0003 0.0007 0.0009 0.0023Effective share of specialised workers 0.184 0.0004 0.0011 0.0002 0.0006 0.0008 0.0018Share of specialised workers in innovation 0.004 0.0000 0.0001 0.0000 0.0001 0.0001 0.0000Expected wage premium 0.333 0.0024 0.0078 0.0057 0.0039 0.0218 0.0156Index of misallocation of talent 0.392 0.0010 0.0039 0.0011 0.0020 0.0009 0.0003Untrained unemployment rate 0.087 0.0085 0.0201 0.0004 0.0016 0.0001 0.0001Specialised unemployment rate 0.071 0.0001 0.0002 0.0000 0.0001 0.0009 0.0019Total unemployment rate 0.079 0.0065 0.0154 0.0003 0.0012 0.0002 0.0003Payroll contribution rate 0.052 0.0019 0.0073 0.0020 0.0037 0.0018 0.0032Growth rate of final output 0.039 0.0033 0.0002 0.0020 0.0001 0.0063 0.0001Social welfare 1.000 0.0040 0.0761 0.2230
Steady state
value Impact Steady State Impact Steady State Impact Steady StateShare of untrained workers 0.795 0.0007 0.0016 0.0003 0.0010 0.0005 0.0011Effective share of specialised workers 0.184 0.0005 0.0013 0.0003 0.0008 0.0004 0.0009Share of specialised workers in innovation 0.004 0.0001 0.0001 0.0000 0.0001 0.0000 0.0000Expected wage premium 0.333 0.0162 0.0117 0.0018 0.0058 0.0108 0.0077Index of misallocation of talent 0.392 0.0019 0.0023 0.0007 0.0029 0.0120 0.0121Untrained unemployment rate 0.087 0.0005 0.0015 0.0063 0.0149 0.0000 0.0000Specialised unemployment rate 0.071 0.0009 0.0018 0.0001 0.0001 0.0038 0.0043Total unemployment rate 0.079 0.0005 0.0015 0.0048 0.0114 0.0007 0.0007Payroll contribution rate 0.052 0.0038 0.0069 0.0014 0.0054 0.0008 0.0016Growth rate of final output 0.039 0.0082 0.0002 0.0024 0.0001 0.0149 0.0000Social welfare 1.000 0.2801 0.0030 0.0256
Steady state
value Impact Steady State Impact Steady StateShare of untrained workers 0.795 0.0205 0.0495 0.0001 0.0003Effective share of specialised workers 0.184 0.0161 0.0381 0.0001 0.0002Share of specialised workers in innovation 0.004 0.0008 0.0017 0.0006 0.0006Expected wage premium 0.333 0.0882 0.1820 0.0028 0.0022Index of misallocation of talent 0.392 0.0142 0.0313 0.0053 0.0055Untrained unemployment rate 0.087 0.0045 0.0225 0.0001 0.0003Specialised unemployment rate 0.071 0.0132 0.0318 0.0004 0.0004Total unemployment rate 0.079 0.0007 0.0085 0.0001 0.0003Payroll contribution rate 0.052 0.0018 0.0031 0.0002 0.0004Growth rate of final output 0.039 0.0042 0.0016 0.0496 0.0011Social welfare 1.000 0.0579 0.0022
*/ The respective individual policy shocks are: Reduction in wU? by 5 percent; κU reduced by 10 percent;
κS reduced by 10 percent; both κU and κS cut by 10 percent; ξU reduced by 37.5 percent; ξSY reduced by 37.5 percent; a decrease in advanced education cost by 5 percent; and an increase in share of public infrastructure investment by 20 percent.
Source: Authors' calculations.
Increase in PublicInfrastructure Investment
Advanced EducationCost Cut
Reduction in BaseMinimum Wage
Reduction in UntrainedWorkers' UB Indexation
Reduction in SpecialisedWorkers' UB Indexation
Reduction in SpecialisedWorkers' Union Markup
Reduction in UntrainedWorkers' Union Markup
Reduction in both UBIndexation Parameters
MiddleIncome Economy, Policy Results: Absolute deviations from baseline*
188
Table 2.5: Summary of Benchmark Composite Reform Programmes
HighIncome Economy Steady state
value Impact Steady State Impact Steady State Impact Steady StateShare of untrained workers 0.732 0.0011 0.0051 0.0112 0.0314 0.0116 0.0325Effective share of specialised workers 0.232 0.0008 0.0037 0.0081 0.0225 0.0084 0.0233Share of specialised workers in innovation 0.019 0.0007 0.0022 0.0014 0.0052 0.0034 0.0078Expected wage premium 0.818 0.0158 0.0473 0.0094 0.0524 0.0060 0.0387Index of misallocation of talent 0.189 0.0073 0.0221 0.0177 0.0488 0.0307 0.0649Untrained unemployment rate 0.126 0.0287 0.0648 0.0315 0.0762 0.0318 0.0772Specialised unemployment rate 0.068 0.0018 0.0046 0.0001 0.0040 0.0016 0.0024Total unemployment rate 0.106 0.0207 0.0466 0.0222 0.0512 0.0229 0.0522Payroll contribution rate 0.126 0.0185 0.0544 0.0190 0.0566 0.0203 0.0588Growth rate of final output 0.008 0.0106 0.0029 0.0105 0.0054 0.0457 0.0095Social welfare 1.000 0.1480 0.1820 0.1801
MiddleIncome Economy Steady state
value Impact Steady State Impact Steady State Impact Steady StateShare of untrained workers 0.795 0.0008 0.0020 0.0205 0.0489 0.0207 0.0493Effective share of specialised workers 0.184 0.0006 0.0016 0.0161 0.0377 0.0162 0.0380Share of specialised workers in innovation 0.004 0.0001 0.0003 0.0007 0.0019 0.0014 0.0028Expected wage premium 0.333 0.0045 0.0102 0.0717 0.1579 0.0688 0.1559Index of misallocation of talent 0.392 0.0037 0.0099 0.0262 0.0476 0.0321 0.0552Untrained unemployment rate 0.087 0.0225 0.0487 0.0259 0.0603 0.0259 0.0605Specialised unemployment rate 0.071 0.0002 0.0007 0.0096 0.0263 0.0090 0.0257Total unemployment rate 0.079 0.0173 0.0374 0.0174 0.0371 0.0175 0.0373Payroll contribution rate 0.052 0.0071 0.0200 0.0058 0.0152 0.0060 0.0156Growth rate of final output 0.039 0.0034 0.0005 0.0071 0.0018 0.0575 0.0031Social welfare 1.000 0.1682 0.1195 0.1207
*/ Programme A includes a decrease in κS by 6.25 percent; a decrease in κU by 6.25 percent; a decrease in wU0 by 10 percent; and a reduction in untrained union markup by 37.5 percent; Programme B includes a decrease in μ by 5 percent; an increase in advanced education period by 22 percent; a decrease in κS by 6.25 percent; a decrease in κU by 6.25 percent; a decrease in wU? by 10 percent; and a reduction in untrained union markup by 37.5 percent; Programme C adds an increase in public infrastructure investment by 20 percent to Programme B.
Source: Authors' calculations.
Programme C
Policy Results: Absolute deviations from baseline*Programme A Programme CProgramme B
Programme A Programme B
189
Table2.6:SensitivityAnalysis:PolicyExperimentsfor(i)ReductioninBaseMinimum
Wage,and(ii)ReductioninUntrained
Workers’UBIndexation
Ben
chm
ark
= 0.
24
β
U=
0.4,
βS =
0.2
Ben
chm
ark
= 0.
24
β
U=
0.35
, βS =
0.2
Redu
ctio
n in
Bas
e M
inim
um W
age:
aIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
te
Shar
e of
unt
rain
ed w
orke
rs0.
732
0.0
006
0.0
027
0.0
004
0.0
015
0.0
016
0.0
048
0.79
50
.000
50
.001
40
.000
30
.000
60
.001
30
.002
4
Effe
ctiv
e sh
are
of sp
ecia
lised
wor
kers
0.23
20.
0004
0.00
200.
0003
0.00
110.
0012
0.00
350.
184
0.00
040.
0011
0.00
020.
0005
0.00
100.
0019
Shar
e of
spec
ialis
ed w
orke
rs in
inno
vatio
n0.
019
0.00
010.
0008
0.00
010.
0005
0.00
030.
0011
0.00
40.
0000
0.00
010.
0000
0.00
010.
0000
0.00
01
Expe
cted
wag
e pr
emiu
m0.
818
0.00
730.
0209
0.00
000.
0080
0.00
160.
0098
0.33
30.
0024
0.00
780
.000
50.
0031
0.0
015
0.00
32
Inde
x of
mis
allo
catio
n of
tale
nt0.
189
0.0
016
0.0
086
0.0
010
0.0
054
0.0
027
0.0
111
0.39
20
.001
00
.003
90
.000
60
.002
30
.001
50
.004
6
Untr
aine
d un
empl
oym
ent r
ate
0.12
60
.011
40
.027
60
.006
70
.017
90
.011
70
.028
90.
087
0.0
085
0.0
201
0.0
048
0.0
128
0.0
087
0.0
205
Spec
ialis
ed u
nem
ploy
men
t rat
e0.
068
0.0
003
0.0
013
0.0
001
0.0
010
0.00
020
.000
40.
071
0.00
010.
0002
0.00
010.
0000
0.00
060.
0008
Tota
l une
mpl
oym
ent r
ate
0.10
60
.008
10
.019
80
.004
70
.012
90
.008
30
.020
50.
079
0.0
065
0.0
154
0.0
037
0.0
098
0.0
065
0.0
156
Payr
oll c
ontr
ibut
ion
rate
0.12
60
.005
20
.021
10
.003
00
.013
90
.005
20
.021
40.
052
0.0
019
0.0
073
0.0
011
0.0
047
0.0
018
0.0
072
Grow
th ra
te o
f fin
al o
utpu
t0.
008
0.00
730.
0011
0.00
430.
0006
0.01
010.
0012
0.03
90.
0033
0.00
020.
0019
0.00
010.
0047
0.00
02
Soci
al w
elfa
re1.
000
0.00
530.
0037
0.0
060
1.00
00.
0040
0.00
250
.007
3
Redu
ctio
n in
Unt
rain
ed w
orke
rs'
Ben
chm
ark
= 0.
24
β
U=
0.4,
βS =
0.2
Ben
chm
ark
= 0.
24
β
U=
0.35
, βS =
0.2
Une
mpl
oym
ent B
enef
it In
dexa
tion:
bIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
te
Shar
e of
unt
rain
ed w
orke
rs0.
732
0.0
006
0.0
017
0.0
006
0.0
013
0.0
017
0.0
031
0.79
50
.000
30
.000
70
.000
20
.000
50
.000
80
.001
2
Effe
ctiv
e sh
are
of sp
ecia
lised
wor
kers
0.23
20.
0004
0.00
130.
0004
0.00
100.
0013
0.00
220.
184
0.00
020.
0006
0.00
020.
0004
0.00
060.
0009
Shar
e of
spec
ialis
ed w
orke
rs in
inno
vatio
n0.
019
0.00
020.
0005
0.00
020.
0004
0.00
040.
0007
0.00
40.
0000
0.00
010.
0000
0.00
000.
0001
0.00
01
Expe
cted
wag
e pr
emiu
m0.
818
0.01
540.
0134
0.01
010.
0071
0.00
920.
0063
0.33
30.
0057
0.00
390.
0038
0.00
230.
0032
0.00
16
Inde
x of
mis
allo
catio
n of
tale
nt0.
189
0.0
025
0.0
055
0.0
025
0.0
048
0.0
037
0.0
072
0.39
20
.001
10
.002
00
.001
00
.001
70
.001
40
.002
3
Untr
aine
d un
empl
oym
ent r
ate
0.12
60
.001
40
.005
80
.000
80
.003
70
.001
80
.006
70.
087
0.0
004
0.0
016
0.0
002
0.0
010
0.0
005
0.0
018
Spec
ialis
ed u
nem
ploy
men
t rat
e0.
068
0.0
005
0.0
008
0.0
005
0.0
008
0.00
000
.000
20.
071
0.00
000.
0001
0.00
000.
0000
0.00
030.
0004
Tota
l une
mpl
oym
ent r
ate
0.10
60
.001
10
.004
30
.000
70
.002
90
.001
30
.004
80.
079
0.0
003
0.0
012
0.0
002
0.0
007
0.0
003
0.0
013
Payr
oll c
ontr
ibut
ion
rate
0.12
60
.006
40
.013
70
.006
20
.012
40
.006
40
.013
90.
052
0.0
020
0.0
037
0.0
020
0.0
035
0.0
020
0.0
037
Grow
th ra
te o
f fin
al o
utpu
t0.
008
0.0
049
0.00
070
.005
40.
0005
0.0
053
0.00
080.
039
0.0
020
0.00
010
.002
10.
0001
0.0
025
0.00
01
Soci
al w
elfa
re1.
000
0.0
868
0.0
865
0.0
936
1.00
00
.076
10
.076
20
.081
5
a/ R
educ
tion
in w
U ₀ b
y 5
perc
ent (
for E
U5,
from
0.5
22 to
0.4
96; f
or LA
5, f
rom
0.5
46 to
0.5
19).
b/ R
educ
tion
in κ
U b
y 10
per
cent
(for
bot
h EU
5 a
nd LA
5, κ
U redu
ces f
rom
0.4
0 to
0.3
6).
Sour
ce: A
utho
rs' c
alcu
latio
ns.
Base
line
Base
line
Abs
olut
e De
viat
ions
from
Bas
elin
e
Abs
olut
e De
viat
ions
from
Bas
elin
e
Abs
olut
e De
viat
ions
from
Bas
elin
e
Sens
itivi
ty A
naly
sis:
Pol
icy
Expe
rimen
t for
(i) R
educ
tion
in B
ase
Min
imum
Wag
e, a
nd (i
i) Re
duct
ion
in U
ntra
ined
wor
kers
' UB
Inde
xatio
n
High
Inco
me
Econ
omy
Base
line
Mid
dle
Inco
me
Econ
omy
Abs
olut
e De
viat
ions
from
Bas
elin
e
Base
line
nUnU nU
nU
190
Table2.7:SensitivityAnalysis:PolicyExperimentsfor(i)ReductioninSpecialisedWorkers’UBIndexation,and(ii)Reductionin
BothUBIndexation
Redu
ctio
n in
Spe
cial
ised
wor
kers
'
B
ench
mar
k=
0.24
βU
= 0.
2, β
S = 0
.4
B
ench
mar
k=
0.24
βU
= 0.
1, β
S = 0
.45
Une
mpl
oym
ent B
enef
it In
dexa
tion:
cIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
te
Shar
e of
unt
rain
ed w
orke
rs0.
732
0.00
100.
0025
0.00
120.
0025
0.00
060.
0016
0.79
50.
0009
0.00
230.
0010
0.00
230.
0004
0.00
12
Effe
ctiv
e sh
are
of sp
ecia
lised
wor
kers
0.23
20
.000
70
.001
80
.000
90
.001
90
.000
50
.001
20.
184
0.0
008
0.0
018
0.0
008
0.0
018
0.0
003
0.0
010
Shar
e of
spec
ialis
ed w
orke
rs in
inno
vatio
n0.
019
0.00
040.
0003
0.00
040.
0002
0.00
040.
0003
0.00
40.
0001
0.00
000.
0001
0.00
000.
0001
0.00
01Ex
pect
ed w
age
prem
ium
0.81
80
.017
30
.008
40
.008
80.
0005
0.0
209
0.0
144
0.33
30
.021
80
.015
60
.014
30
.007
60
.024
40
.020
8In
dex
of m
isal
loca
tion
of ta
lent
0.18
90
.003
10
.002
40
.002
30
.001
40
.002
80
.002
70.
392
0.0
009
0.0
003
0.0
002
0.00
070
.001
20
.001
2Un
trai
ned
unem
ploy
men
t rat
e0.
126
0.0
004
0.0
009
0.0
003
0.0
007
0.0
005
0.0
015
0.08
70
.000
10.
0001
0.00
000.
0001
0.0
002
0.0
005
Spec
ialis
ed u
nem
ploy
men
t rat
e0.
068
0.0
013
0.0
024
0.0
011
0.0
020
0.0
010
0.0
019
0.07
10
.000
90
.001
90
.000
80
.001
60
.000
50
.001
2To
tal u
nem
ploy
men
t rat
e0.
106
0.0
005
0.0
012
0.0
004
0.0
009
0.0
006
0.0
015
0.07
90
.000
20
.000
30
.000
20
.000
20
.000
20
.000
6Pa
yrol
l con
trib
utio
n ra
te0.
126
0.0
037
0.0
073
0.0
036
0.0
069
0.0
037
0.0
073
0.05
20
.001
80
.003
20
.001
80
.003
10
.001
70
.003
2Gr
owth
rate
of f
inal
out
put
0.00
80
.006
20.
0004
0.0
048
0.00
030
.008
90.
0004
0.03
90
.006
30.
0001
0.0
052
0.00
010
.007
90.
0001
Soci
al w
elfa
re1.
000
0.1
704
0.1
661
0.2
098
1.00
00
.223
00
.222
10
.405
3
Redu
ctio
n in
Bot
h U
nem
ploy
men
t
B
ench
mar
k
β
U=
0.4,
βS =
0.2
βU
= 0.
2, β
S = 0
.4
B
ench
mar
k
β
U=
0.35
, βS =
0.2
βU
= 0.
1, β
S = 0
.45
Bene
fits'
Inde
xatio
n:d
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Shar
e of
unt
rain
ed w
orke
rs0.
732
0.00
040.
0007
0.00
080.
0010
0.00
020.
0005
0.79
50.
0007
0.00
160.
0018
0.00
270.
0003
0.00
08
Effe
ctiv
e sh
are
of sp
ecia
lised
wor
kers
0.23
20
.000
30
.000
50
.000
60
.000
70
.000
20
.000
40.
184
0.0
005
0.0
013
0.0
014
0.0
022
0.0
002
0.0
007
Shar
e of
spec
ialis
ed w
orke
rs in
inno
vatio
n0.
019
0.00
060.
0008
0.00
080.
0011
0.00
060.
0007
0.00
40.
0001
0.00
010.
0001
0.00
010.
0001
0.00
01Ex
pect
ed w
age
prem
ium
0.81
80
.001
90.
0050
0.00
100.
0064
0.0
048
0.00
100.
333
0.0
162
0.0
117
0.0
106
0.0
065
0.0
181
0.0
155
Inde
x of
mis
allo
catio
n of
tale
nt0.
189
0.0
056
0.0
078
0.0
067
0.0
096
0.0
050
0.0
072
0.39
20
.001
90
.002
30
.001
40
.001
90
.002
10
.002
8Un
trai
ned
unem
ploy
men
t rat
e0.
126
0.0
018
0.0
066
0.0
016
0.0
065
0.0
018
0.0
067
0.08
70
.000
50
.001
50
.000
20
.001
00
.000
50
.001
9Sp
ecia
lised
une
mpl
oym
ent r
ate
0.06
80
.001
80
.003
20
.002
20
.003
60
.001
60
.003
00.
071
0.0
009
0.0
018
0.0
016
0.0
025
0.0
006
0.0
013
Tota
l une
mpl
oym
ent r
ate
0.10
60
.001
70
.005
40
.001
70
.005
50
.001
70
.005
40.
079
0.0
005
0.0
015
0.0
004
0.0
012
0.0
005
0.0
017
Payr
oll c
ontr
ibut
ion
rate
0.12
60
.010
20
.020
60
.010
40
.020
90
.010
10
.020
40.
052
0.0
038
0.0
069
0.0
039
0.0
069
0.0
038
0.0
068
Grow
th ra
te o
f fin
al o
utpu
t0.
008
0.0
110
0.00
110
.008
00.
0012
0.0
140
0.00
090.
039
0.0
082
0.00
020
.005
60.
0001
0.0
096
0.00
02So
cial
wel
fare
1.00
00
.241
40
.220
60
.271
21.
000
0.2
801
0.2
018
0.4
399
c/ R
educ
tion
in κ
S by
10 p
erce
nt (f
or b
oth
EU5
and
LA5
, κS re
duce
s fro
m 0
.40
to 0
.36)
.
d/ R
educ
tion
in b
oth κS a
nd κ
U b
y 10
per
cent
(for
bot
h EU
5 a
nd LA
5, b
oth κS a
nd κ
S dec
line
from
0.4
0 to
0.3
6).
Sour
ce: A
utho
rs' c
alcu
latio
ns.
Sens
itivi
ty A
naly
sis:
Pol
icy
Expe
rimen
t for
(i) R
educ
tion
in S
peci
alise
d w
orke
rs' U
B In
dexa
tion,
and
(ii)
Redu
ctio
n in
Bot
h U
B In
dexa
tion
High
Inco
me
Econ
omy
Base
line
Abs
olut
e De
viat
ions
from
Bas
elin
e
Base
line
Abs
olut
e De
viat
ions
from
Bas
elin
e
Mid
dle
Inco
me
Econ
omy
Base
line
Abs
olut
e De
viat
ions
from
Bas
elin
e
Base
line
Abs
olut
e De
viat
ions
from
Bas
elin
e
nSnS
191
Table2.8:SensitivityAnalysis:PolicyExperimentsfor(i)ReductioninUntrainedUnionMark-up,and(ii)ReductioninSpecialised
UnionMark-upoverTargetWage
Redu
ctio
n in
Unt
rain
ed w
orke
rs' u
nion
mar
kup
Ben
chm
ark
= 0.
24
β
U=
0.4,
βS =
0.2
Ben
chm
ark
= 0.
24
β
U=
0.35
, βS =
0.2
over
targ
et w
age:
eIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
te
Shar
e of
unt
rain
ed w
orke
rs0.
732
0.0
003
0.0
014
0.0
002
0.0
008
0.0
008
0.0
025
0.79
50
.000
30
.001
00
.000
20
.000
50
.000
90
.001
8
Effe
ctiv
e sh
are
of sp
ecia
lised
wor
kers
0.23
20.
0002
0.00
100.
0001
0.00
060.
0006
0.00
190.
184
0.00
030.
0008
0.00
020.
0004
0.00
070.
0014
Shar
e of
spec
ialis
ed w
orke
rs in
inno
vatio
n0.
019
0.00
010.
0004
0.00
000.
0003
0.00
010.
0006
0.00
40.
0000
0.00
010.
0000
0.00
000.
0000
0.00
01Ex
pect
ed w
age
prem
ium
0.81
80.
0038
0.01
100.
0000
0.00
420.
0009
0.00
520.
333
0.00
180.
0058
0.0
004
0.00
230
.001
00.
0024
Inde
x of
mis
allo
catio
n of
tale
nt0.
189
0.0
009
0.0
046
0.0
005
0.0
028
0.0
014
0.0
059
0.39
20
.000
70
.002
90
.000
40
.001
70
.001
10
.003
4Un
trai
ned
unem
ploy
men
t rat
e0.
126
0.0
060
0.0
146
0.0
035
0.0
094
0.0
061
0.0
153
0.08
70
.006
30
.014
90
.003
50
.009
50
.006
40
.015
3Sp
ecia
lised
une
mpl
oym
ent r
ate
0.06
80
.000
20
.000
70
.000
10
.000
50.
0001
0.0
002
0.07
10.
0001
0.00
010.
0001
0.00
000.
0005
0.00
06To
tal u
nem
ploy
men
t rat
e0.
106
0.0
042
0.0
105
0.0
025
0.0
068
0.0
043
0.0
109
0.07
90
.004
80
.011
40
.002
70
.007
30
.004
80
.011
6Pa
yrol
l con
trib
utio
n ra
te0.
126
0.0
027
0.0
113
0.0
016
0.0
074
0.0
027
0.0
115
0.05
20
.001
40
.005
40
.000
80
.003
50
.001
30
.005
4Gr
owth
rate
of f
inal
out
put
0.00
80.
0038
0.00
060.
0022
0.00
030.
0053
0.00
060.
039
0.00
240.
0001
0.00
140.
0001
0.00
350.
0001
Soci
al w
elfa
re1.
000
0.00
270.
0019
0.0
032
1.00
00.
0030
0.00
180
.005
4
Redu
ctio
n in
Spe
cial
ised
wor
kers
' uni
on m
ark
up
B
ench
mar
k=
0.24
βU
= 0.
2, β
S = 0
.4
B
ench
mar
k=
0.24
βU
= 0.
1, β
S = 0
.45
over
targ
et w
age:
fIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
te
Shar
e of
unt
rain
ed w
orke
rs0.
732
0.00
050.
0012
0.00
050.
0010
0.00
030.
0008
0.79
50.
0005
0.00
110.
0004
0.00
100.
0002
0.00
06
Effe
ctiv
e sh
are
of sp
ecia
lised
wor
kers
0.23
20
.000
30
.000
90
.000
30
.000
70
.000
20
.000
60.
184
0.0
004
0.0
009
0.0
003
0.0
008
0.0
002
0.0
005
Shar
e of
spec
ialis
ed w
orke
rs in
inno
vatio
n0.
019
0.00
010.
0001
0.00
010.
0001
0.00
010.
0001
0.00
40.
0000
0.00
000.
0000
0.00
000.
0000
0.00
00Ex
pect
ed w
age
prem
ium
0.81
80
.008
30
.003
90
.003
50.
0002
0.0
099
0.0
066
0.33
30
.010
80
.007
70
.006
10
.003
20
.012
10
.010
2In
dex
of m
isal
loca
tion
of ta
lent
0.18
90
.004
80
.004
90
.003
80
.003
70
.004
60
.005
00.
392
0.0
120
0.0
121
0.0
102
0.0
100
0.0
121
0.0
124
Untr
aine
d un
empl
oym
ent r
ate
0.12
60
.000
10
.000
40
.000
10
.000
30
.000
20
.000
70.
087
0.00
000.
0000
0.00
000.
0000
0.0
001
0.0
002
Spec
ialis
ed u
nem
ploy
men
t rat
e0.
068
0.0
036
0.0
041
0.0
029
0.0
032
0.0
034
0.0
039
0.07
10
.003
80
.004
30
.003
30
.003
60
.003
60
.004
0To
tal u
nem
ploy
men
t rat
e0.
106
0.0
010
0.0
013
0.0
008
0.0
010
0.0
010
0.0
014
0.07
90
.000
70
.000
70
.000
60
.000
60
.000
70
.000
9Pa
yrol
l con
trib
utio
n ra
te0.
126
0.0
016
0.0
034
0.0
013
0.0
026
0.0
016
0.0
034
0.05
20
.000
80
.001
60
.000
70
.001
30
.000
80
.001
6Gr
owth
rate
of f
inal
out
put
0.00
80.
0094
0.00
020.
0074
0.00
010.
0108
0.00
020.
039
0.01
490.
0000
0.01
240.
0000
0.01
900.
0000
Soci
al w
elfa
re1.
000
0.0
187
0.0
151
0.0
167
1.00
00
.025
60
.022
50
.024
6
e/ R
educ
tion
in m
ark
up b
y 37
.5 p
erce
nt (f
or E
U5,
ξU from
0.0
6 to
0.0
375;
for L
A5,
ξU from
0.0
8 to
0.0
5).
f/ R
educ
tion
in m
ark
up b
y 37
.5 p
erce
nt (f
or E
U5,
ξSY fr
om 0
.06
to 0
.037
5; fo
r LA
5, ξSY
from
0.0
8 to
0.0
5).
Sour
ce: A
utho
rs' c
alcu
latio
ns.
Sen
sitiv
ity A
naly
sis:
Pol
icy E
xper
imen
t for
(i) R
educ
tion
in U
ntra
ined
wor
kers
' uni
on m
ark
up, a
nd (i
i) Re
duct
ion
in S
peci
alis
ed w
orke
rs' u
nion
mar
kup
Base
line
Base
line
Abs
olut
e De
viat
ions
from
Bas
elin
e
A
bsol
ute
Devi
atio
ns fr
om B
asel
ine
High
Inco
me
Econ
omy
Mid
dle
Inco
me
Econ
omy
Base
line
Abs
olut
e De
viat
ions
from
Bas
elin
e
Base
line
Abs
olut
e De
viat
ions
from
Bas
elin
e
nU
nU
nSnS
192
Table2.9:SensitivityAnalysis:PolicyExperimentsfor(i)AdvancedEducationCostCut,and(ii)CompositeReformProgramme
A
Ben
chm
ark
=
0.95
= 0.
24
B
ench
mar
k
= 0.
95=
0.24
Adva
nced
Edu
catio
n Co
st C
ut:g
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Shar
e of
unt
rain
ed w
orke
rs0.
732
0.0
110
0.0
295
0.0
094
0.0
217
0.0
096
0.0
215
0.79
50
.020
50
.049
50
.011
20
.029
00
.018
90
.041
4Ef
fect
ive
shar
e of
spec
ialis
ed w
orke
rs0.
232
0.00
800.
0211
0.00
680.
0157
0.00
700.
0155
0.18
40.
0161
0.03
810.
0089
0.02
270.
0149
0.03
21Sh
are
of sp
ecia
lised
wor
kers
in in
nova
tion
0.01
90.
0014
0.00
380.
0012
0.00
280.
0012
0.00
300.
004
0.00
080.
0017
0.00
040.
0010
0.00
070.
0016
Expe
cted
wag
e pr
emiu
m0.
818
0.0
549
0.1
241
0.0
483
0.0
962
0.0
556
0.1
065
0.33
30
.088
20
.182
00
.050
50
.115
00
.096
10
.183
6In
dex
of m
isal
loca
tion
of ta
lent
0.18
90
.011
00
.030
90
.009
40
.022
80
.011
00
.025
80.
392
0.0
142
0.0
313
0.0
080
0.0
193
0.0
188
0.0
387
Untr
aine
d un
empl
oym
ent r
ate
0.12
60
.003
70
.020
30
.003
10
.015
10
.003
30
.015
50.
087
0.0
045
0.0
225
0.0
025
0.0
136
0.0
042
0.0
195
Spec
ialis
ed u
nem
ploy
men
t rat
e0.
068
0.00
540.
0133
0.00
460.
0099
0.00
380.
0079
0.07
10.
0132
0.03
180.
0073
0.01
870.
0106
0.02
30To
tal u
nem
ploy
men
t rat
e0.
106
0.0
015
0.0
108
0.0
013
0.0
082
0.0
016
0.0
091
0.07
90
.000
70
.008
50
.000
50
.006
10
.001
00
.008
9Pa
yrol
l con
trib
utio
n ra
te0.
126
0.00
040
.006
50.
0004
0.0
047
0.00
000
.006
40.
052
0.00
180.
0031
0.00
100.
0017
0.00
130.
0010
Grow
th ra
te o
f fin
al o
utpu
t0.
008
0.00
050.
0042
0.00
110.
0026
0.00
290.
0028
0.03
90.
0042
0.00
160.
0039
0.00
080.
0102
0.00
13So
cial
wel
fare
1.00
00.
0075
0.00
880
.000
21.
000
0.05
790.
0966
0.04
41
Ben
chm
ark
= 0.
24=
0.7
Ben
chm
ark
= 0.
24=
0.7
Com
posi
te R
efor
m P
roga
mm
e A:
hIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
te
Shar
e of
unt
rain
ed w
orke
rs0.
732
0.0
011
0.0
051
0.0
007
0.0
028
0.0
016
0.0
062
0.79
50
.000
80
.002
00
.000
40
.000
80
.000
90
.002
1Ef
fect
ive
shar
e of
spec
ialis
ed w
orke
rs0.
232
0.00
080.
0037
0.00
050.
0021
0.00
110.
0045
0.18
40.
0006
0.00
160.
0004
0.00
060.
0007
0.00
17Sh
are
of sp
ecia
lised
wor
kers
in in
nova
tion
0.01
90.
0007
0.00
220.
0006
0.00
160.
0010
0.00
330.
004
0.00
010.
0003
0.00
010.
0002
0.00
020.
0004
Expe
cted
wag
e pr
emiu
m0.
818
0.01
580.
0473
0.0
054
0.01
420.
0141
0.04
410.
333
0.0
045
0.01
020
.013
40
.002
80
.004
90.
0101
Inde
x of
mis
allo
catio
n of
tale
nt0.
189
0.0
073
0.0
221
0.0
058
0.0
160
0.0
096
0.0
294
0.39
20
.003
70
.009
90
.002
70
.006
90
.004
10
.010
9Un
trai
ned
unem
ploy
men
t rat
e0.
126
0.0
287
0.0
648
0.0
171
0.0
438
0.0
288
0.0
655
0.08
70
.022
50
.048
70
.013
10
.032
70
.022
50
.048
7Sp
ecia
lised
une
mpl
oym
ent r
ate
0.06
80
.001
80
.004
60
.001
40
.003
90
.001
80
.004
90.
071
0.0
002
0.0
007
0.0
002
0.0
010
0.0
002
0.0
007
Tota
l une
mpl
oym
ent r
ate
0.10
60
.020
70
.046
60
.012
40
.031
80
.020
80
.047
20.
079
0.0
173
0.0
374
0.0
101
0.0
253
0.0
173
0.0
374
Payr
oll c
ontr
ibut
ion
rate
0.12
60
.018
50
.054
40
.013
30
.040
60
.018
70
.055
30.
052
0.0
071
0.0
200
0.0
051
0.0
149
0.0
071
0.0
200
Grow
th ra
te o
f fin
al o
utpu
t0.
008
0.01
060.
0029
0.00
320.
0018
0.01
040.
0024
0.03
90.
0034
0.00
050
.000
30.
0003
0.00
340.
0003
Soci
al w
elfa
re1.
000
0.1
480
0.1
478
0.1
514
1.00
00
.168
20
.170
50
.168
6
g/ D
ecre
ase
in a
dvan
ced
educ
atio
n co
st b
y 5
perc
ent (
for E
U5,
μ d
ecre
ases
from
0.0
80 to
0.0
76; f
or LA
5, μ
dec
reas
es fr
om 0
.120
to 0
.114
).
h/ F
or E
U5,
this
incl
udes
a d
ecre
ase
in κ
U from
0.4
00 to
0.3
75; a
dec
reas
e in
κS fr
om 0
.400
to 0
.375
; dec
reas
e in
wU₀
b
y 10
per
cent
from
0.5
22 to
0.4
70; a
nd a
redu
ctio
n in
unt
rain
ed u
nion
mar
kup
by
37.5
per
cent
(ξU fr
om 0
.06
to 0
.037
5).
F
or LA
5, t
his i
nclu
des a
dec
reas
e in
κU fr
om 0
.400
to 0
.375
; a d
ecre
ase
in κ
S from
0.4
00 to
0.3
75; a
dec
reas
e in
wU0
b
y 10
per
cent
from
0.5
46 to
0.4
91; a
nd a
redu
ctio
n in
unt
rain
ed u
nion
mar
kup
by
37.5
per
cent
(ξU fr
om 0
.08
to 0
.05)
.
Sour
ce: A
utho
rs' c
alcu
latio
ns.
Sens
itivi
ty A
naly
sis:
Pol
icy
Expe
rimen
t for
(i) A
dvan
ced
Educ
atio
n Co
st C
ut ,
and
(ii) C
ompo
site
Ref
orm
Pro
gram
me
A
High
Inco
me
Econ
omy
Mid
dle
Inco
me
Econ
omy
Base
line
Abs
olut
e De
viat
ions
from
Bas
elin
e
Base
line
Abs
olut
e De
viat
ions
from
Bas
elin
e
Base
line
Base
line
Abs
olut
e De
viat
ions
from
Bas
elin
e
A
bsol
ute
Devi
atio
ns fr
om B
asel
ine
nSn
S
nUnU
ee
VnU
V
193
Table2.10:SensitivityAnalysis:PolicyExperimentsfor(i)CompositeReformProgrammeB,and(ii)CompositeReformProgramme
C
Ben
chm
ark
= 0.
24=
0.7
Ben
chm
ark
= 0.
24=
0.7
Com
posi
te R
efor
m P
roga
mm
e B:
iIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teSh
are
of u
ntra
ined
wor
kers
0.73
20
.011
20
.031
40
.009
20
.021
30
.010
10
.025
70.
795
0.0
205
0.0
489
0.0
185
0.0
391
0.0
193
0.0
425
Effe
ctiv
e sh
are
of sp
ecia
lised
wor
kers
0.23
20.
0081
0.02
250.
0067
0.01
540.
0074
0.01
850.
184
0.01
610.
0377
0.01
460.
0303
0.01
510.
0329
Shar
e of
spec
ialis
ed w
orke
rs in
inno
vatio
n0.
019
0.00
140.
0052
0.00
110.
0034
0.00
150.
0061
0.00
40.
0007
0.00
190.
0006
0.00
150.
0009
0.00
23Ex
pect
ed w
age
prem
ium
0.81
80
.009
40
.052
40
.024
10
.053
40
.003
40
.021
30.
333
0.0
717
0.1
579
0.0
747
0.1
431
0.0
660
0.1
339
Inde
x of
mis
allo
catio
n of
tale
nt0.
189
0.0
177
0.0
488
0.0
147
0.0
348
0.0
183
0.0
535
0.39
20
.026
20
.047
60
.024
20
.039
90
.027
10
.049
5Un
trai
ned
unem
ploy
men
t rat
e0.
126
0.0
315
0.0
762
0.0
186
0.0
503
0.0
313
0.0
743
0.08
70
.025
90
.060
30
.015
00
.040
60
.025
70
.058
9Sp
ecia
lised
une
mpl
oym
ent r
ate
0.06
80
.000
10.
0040
0.0
004
0.00
150
.000
80.
0001
0.07
10.
0096
0.02
630.
0085
0.02
070.
0086
0.02
16To
tal u
nem
ploy
men
t rat
e0.
106
0.0
222
0.0
512
0.0
133
0.0
346
0.0
223
0.0
513
0.07
90
.017
40
.037
10
.009
50
.024
90
.017
50
.037
8Pa
yrol
l con
trib
utio
n ra
te0.
126
0.0
190
0.0
566
0.0
135
0.0
415
0.0
193
0.0
580
0.05
20
.005
80
.015
20
.003
60
.010
50
.005
90
.016
4Gr
owth
rate
of f
inal
out
put
0.00
80.
0105
0.00
540.
0030
0.00
290.
0109
0.00
370.
039
0.00
710.
0018
0.00
310.
0012
0.00
720.
0010
Soci
al w
elfa
re1.
000
0.1
820
0.1
722
0.1
795
1.00
00
.119
50
.114
10
.121
3
Ben
chm
ark
= 0.
24=
0.7
Ben
chm
ark
= 0.
24=
0.7
Com
posi
te R
efor
m P
roga
mm
e C:
jIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teIm
pact
Stea
dys
tate
Impa
ctSt
eady
sta
teSh
are
of u
ntra
ined
wor
kers
0.73
20
.011
60
.032
50
.009
80
.022
70
.010
80
.027
80.
795
0.0
207
0.0
493
0.0
187
0.0
396
0.0
195
0.0
431
Effe
ctiv
e sh
are
of sp
ecia
lised
wor
kers
0.23
20.
0084
0.02
330.
0071
0.01
640.
0079
0.01
990.
184
0.01
620.
0380
0.01
470.
0307
0.01
530.
0334
Shar
e of
spec
ialis
ed w
orke
rs in
inno
vatio
n0.
019
0.00
340.
0078
0.00
310.
0058
0.00
440.
0100
0.00
40.
0014
0.00
280.
0013
0.00
230.
0019
0.00
37Ex
pect
ed w
age
prem
ium
0.81
80.
0060
0.0
387
0.0
144
0.0
468
0.01
150
.008
70.
333
0.0
688
0.1
559
0.0
727
0.1
422
0.0
628
0.1
319
Inde
x of
mis
allo
catio
n of
tale
nt0.
189
0.0
307
0.0
649
0.0
278
0.0
503
0.0
368
0.0
784
0.39
20
.032
10
.055
20
.030
10
.047
00
.035
60
.061
0Un
trai
ned
unem
ploy
men
t rat
e0.
126
0.0
318
0.0
772
0.0
188
0.0
511
0.0
317
0.0
759
0.08
70
.025
90
.060
50
.015
00
.040
80
.025
70
.059
2Sp
ecia
lised
une
mpl
oym
ent r
ate
0.06
80
.001
60.
0024
0.0
017
0.00
020
.002
80
.002
30.
071
0.00
900.
0257
0.00
800.
0202
0.00
790.
0208
Tota
l une
mpl
oym
ent r
ate
0.10
60
.022
90
.052
20
.013
90
.035
50
.023
10
.052
90.
079
0.0
175
0.0
373
0.0
097
0.0
251
0.0
177
0.0
381
Payr
oll c
ontr
ibut
ion
rate
0.12
60
.020
30
.058
80
.014
70
.043
60
.021
00
.061
10.
052
0.0
060
0.0
156
0.0
039
0.0
109
0.0
062
0.0
169
Grow
th ra
te o
f fin
al o
utpu
t0.
008
0.04
570.
0095
0.03
780.
0063
0.04
580.
0070
0.03
90.
0575
0.00
310.
0533
0.00
230.
0575
0.00
18So
cial
wel
fare
1.00
00
.180
10
.170
20
.182
21.
000
0.1
207
0.1
152
0.1
258
i/ F
or E
U5,
this
incl
udes
a d
ecre
ase
in μ
by
5 pe
rcen
t (fr
om 0
.080
to 0
.076
); an
incr
ease
in a
dvan
ced
educ
atio
n pe
riod
by 2
2 pe
rcen
t (ε
to 0
.140
0); a
dec
reas
e in
κS fr
om 0
.400
to 0
.375
;
a d
ecre
ase
in κ
U from
0.4
00 to
0.3
75; a
dec
reas
e in
wU₀
by
10 p
erce
nt fr
om 0
.522
to 0
.470
; and
a re
duct
ion
in u
ntra
ined
uni
on m
ark
up (ξ
U from
0.0
6 to
0.0
375)
.
For
LA5
, thi
s inc
lude
s a d
ecre
ase
in μ
by
5 pe
rcen
t (fr
om 0
.120
to 0
.114
); an
incr
ease
in a
dvan
ced
educ
atio
n pe
riod
by 2
2 pe
rcen
t (ε
to 0
.150
0); a
dec
reas
e in
κS fr
om 0
.400
to 0
.375
;
a d
ecre
ase
in κ
U from
0.4
00 to
0.3
75; a
dec
reas
e in
wU₀
by
10 p
erce
nt fr
om 0
.546
to 0
.491
; and
a re
duct
ion
in u
ntra
ined
uni
on m
ark
up (ξ
U from
0.0
8 to
0.0
5).
j/ F
or E
U5,
this
incl
udes
a d
ecre
ase
in μ
by
5 pe
rcen
t (fr
om 0
.080
to 0
.076
); an
incr
ease
in a
dvan
ced
educ
atio
n pe
riod
by 2
2 pe
rcen
t (ε
to 0
.140
0); a
dec
reas
e in
κS fr
om 0
.400
to 0
.375
;
a d
ecre
ase
in κ
U from
0.4
00 to
0.3
75; a
dec
reas
e in
wU₀
by
10 p
erce
nt f
rom
0.5
22 to
0.4
70; a
redu
ctio
n in
unt
rain
ed u
nion
mar
kup
(ξU fr
om 0
.06
to 0
.037
5);
a
nd a
n in
crea
se in
infr
astr
uctu
re in
vest
men
t by
20 p
erce
nt (ν I
from
0.0
21 to
0.0
252)
.
For
LA5
, thi
s inc
lude
s a d
ecre
ase
in μ
by
5 pe
rcen
t (fr
om 0
.120
to 0
.114
); an
incr
ease
in a
dvan
ced
educ
atio
n pe
riod
by 2
2 pe
rcen
t (ε
to 0
.150
0); a
dec
reas
e in
κS fr
om 0
.400
to 0
.375
;
a d
ecre
ase
in κ
U from
0.4
00 to
0.3
75; a
dec
reas
e in
wU₀
by
10 p
erce
nt fr
om 0
.546
to 0
.491
; and
a re
duct
ion
in u
ntra
ined
uni
on m
ark
up (ξ
U from
0.0
8 to
0.0
5);
a
nd a
n in
crea
se in
infr
astr
uctu
re in
vest
men
t by
20 p
erce
nt (ν I
from
0.0
69 to
0.0
828)
.
Sour
ce: A
utho
rs' c
alcu
latio
ns.
Sen
sitiv
ity A
naly
sis: P
olic
y Ex
perim
ent f
or (i
) Com
posit
e Re
form
Pro
gram
me
B, a
nd (i
i) Co
mpo
site
Refo
rm P
rogr
amm
e C
High
Inco
me
Econ
omy
Mid
dle
Inco
me
Econ
omy
Base
line
Abs
olut
e De
viat
ions
from
Bas
elin
e
Base
line
Abs
olut
e De
viat
ions
from
Bas
elin
e
Base
line
Abs
olut
e De
viat
ions
from
Bas
elin
eBa
selin
e
A
bsol
ute
Devi
atio
ns fr
om B
asel
ine
nUnU nU
nUnUV
nUV
nUV
nUV
194
Table 2.11: High-Income Economy: Sensitivity Analysis, comparison betweenBenchmark Model and Model without UB consideration
Baseline
steady state Benchmark
value Impact Steadystate Impact Steadystate Impact Steadystate Impact SteadystateShare of untrained workers 0.732 0.0006 0.0027 0.0000 0.0003 0.0006 0.0017 0.0001 0.0002Effective share of specialised workers 0.232 0.0004 0.0020 0.0000 0.0002 0.0004 0.0013 0.0000 0.0001Share of specialised workers in innovation 0.019 0.0001 0.0008 0.0001 0.0005 0.0002 0.0005 0.0002 0.0003Expected wage premium** 0.818 0.0073 0.0209 0.0002 0.0019 0.0154 0.0134 0.0011 0.0012Index of misallocation of talent 0.189 0.0016 0.0086 0.0011 0.0061 0.0025 0.0055 0.0019 0.0039Untrained unemployment rate 0.126 0.0114 0.0276 0.0112 0.0262 0.0014 0.0058 0.0012 0.0047Specialised unemployment rate 0.068 0.0003 0.0013 0.0006 0.0024 0.0005 0.0008 0.0008 0.0015Total unemployment rate 0.106 0.0081 0.0198 0.0081 0.0190 0.0011 0.0043 0.0010 0.0037Payroll contribution rate 0.126 0.0052 0.0211 0.0053 0.0208 0.0064 0.0137 0.0065 0.0135Growth rate of final output 0.008 0.0073 0.0011 0.0074 0.0007 0.0049 0.0007 0.0048 0.0005Social welfare 1.000 0.0053 0.0091 0.0868 0.0845
Baseline Impact Steadystate Impact Steadystate Impact Steadystate Impact SteadystateShare of untrained workers 0.732 0.0010 0.0025 0.0001 0.0003 0.0004 0.0007 0.0002 0.0005Effective share of specialised workers 0.232 0.0007 0.0018 0.0001 0.0002 0.0003 0.0005 0.0001 0.0004Share of specialised workers in innovation 0.019 0.0004 0.0003 0.0006 0.0007 0.0006 0.0008 0.0007 0.0010Expected wage premium** 0.818 0.0173 0.0084 0.0022 0.0017 0.0019 0.0050 0.0033 0.0029Index of misallocation of talent 0.189 0.0031 0.0024 0.0042 0.0055 0.0056 0.0078 0.0061 0.0092Untrained unemployment rate 0.126 0.0004 0.0009 0.0007 0.0029 0.0018 0.0066 0.0019 0.0074Specialised unemployment rate 0.068 0.0013 0.0024 0.0008 0.0012 0.0018 0.0032 0.0015 0.0026Total unemployment rate 0.106 0.0005 0.0012 0.0007 0.0024 0.0017 0.0054 0.0017 0.0059Payroll contribution rate 0.126 0.0037 0.0073 0.0037 0.0081 0.0102 0.0206 0.0102 0.0208Growth rate of final output 0.008 0.0062 0.0004 0.0063 0.0008 0.0110 0.0011 0.0111 0.0013Social welfare 1.000 0.1704 0.1732 0.2414 0.2428
Baseline Impact Steadystate Impact Steadystate Impact Steadystate Impact SteadystateShare of untrained workers 0.732 0.0003 0.0014 0.0000 0.0002 0.0005 0.0012 0.0000 0.0001Effective share of specialised workers 0.232 0.0002 0.0010 0.0000 0.0001 0.0003 0.0009 0.0000 0.0001Share of specialised workers in innovation 0.019 0.0001 0.0004 0.0000 0.0003 0.0001 0.0001 0.0002 0.0003Expected wage premium** 0.818 0.0038 0.0110 0.0001 0.0010 0.0083 0.0039 0.0004 0.0008Index of misallocation of talent 0.189 0.0009 0.0046 0.0006 0.0032 0.0048 0.0049 0.0053 0.0063Untrained unemployment rate 0.126 0.0060 0.0146 0.0059 0.0138 0.0001 0.0004 0.0003 0.0013Specialised unemployment rate 0.068 0.0002 0.0007 0.0003 0.0012 0.0036 0.0041 0.0033 0.0036Total unemployment rate 0.106 0.0042 0.0105 0.0042 0.0100 0.0010 0.0013 0.0011 0.0018Payroll contribution rate 0.126 0.0027 0.0113 0.0027 0.0110 0.0016 0.0034 0.0016 0.0037Growth rate of final output 0.008 0.0038 0.0006 0.0038 0.0004 0.0094 0.0002 0.0093 0.0004Social welfare 1.000 0.0027 0.0047 0.0187 0.0206
Baseline Impact Steadystate Impact Steadystate Impact Steadystate Impact SteadystateShare of untrained workers 0.732 0.0110 0.0295 0.0092 0.0190 0.0011 0.0051 0.0001 0.0009Effective share of specialised workers 0.232 0.0080 0.0211 0.0067 0.0137 0.0008 0.0037 0.0001 0.0007Share of specialised workers in innovation 0.019 0.0014 0.0038 0.0011 0.0025 0.0007 0.0022 0.0006 0.0016Expected wage premium** 0.818 0.0549 0.1241 0.0629 0.1252 0.0158 0.0473 0.0024 0.0058Index of misallocation of talent 0.189 0.0110 0.0309 0.0092 0.0201 0.0073 0.0221 0.0062 0.0177Untrained unemployment rate 0.126 0.0037 0.0203 0.0031 0.0133 0.0287 0.0648 0.0284 0.0630Specialised unemployment rate 0.068 0.0054 0.0133 0.0045 0.0086 0.0018 0.0046 0.0023 0.0064Total unemployment rate 0.106 0.0015 0.0108 0.0013 0.0073 0.0207 0.0466 0.0206 0.0459Payroll contribution rate 0.126 0.0004 0.0065 0.0003 0.0044 0.0185 0.0544 0.0186 0.0543Growth rate of final output 0.008 0.0005 0.0042 0.0007 0.0027 0.0106 0.0029 0.0108 0.0023Social welfare 1.000 0.0075 0.0145 0.1480 0.1419
Baseline Impact Steadystate Impact Steadystate Impact Steadystate Impact SteadystateShare of untrained workers 0.732 0.0112 0.0314 0.0089 0.0192 0.0116 0.0325 0.0099 0.0214Effective share of specialised workers 0.232 0.0081 0.0225 0.0065 0.0138 0.0084 0.0233 0.0072 0.0154Share of specialised workers in innovation 0.019 0.0014 0.0052 0.0011 0.0035 0.0034 0.0078 0.0032 0.0061Expected wage premium** 0.818 0.0094 0.0524 0.0166 0.0743 0.0060 0.0387 0.0018 0.0622Index of misallocation of talent 0.189 0.0177 0.0488 0.0155 0.0367 0.0307 0.0649 0.0289 0.0530Untrained unemployment rate 0.126 0.0315 0.0762 0.0309 0.0713 0.0318 0.0772 0.0314 0.0728Specialised unemployment rate 0.068 0.0001 0.0040 0.0012 0.0015 0.0016 0.0024 0.0024 0.0025Total unemployment rate 0.106 0.0222 0.0512 0.0221 0.0499 0.0229 0.0522 0.0227 0.0511Payroll contribution rate 0.126 0.0190 0.0566 0.0192 0.0566 0.0203 0.0588 0.0204 0.0588Growth rate of final output 0.008 0.0105 0.0054 0.0108 0.0037 0.0457 0.0095 0.0459 0.0078Social welfare 1.000 0.1820 0.1648 0.1801 0.1648
*/ Calibrated based on Europe5. The exact details of each individual policy experiment listed are documented in their respective tables, in Table 2.62.10.**/ All scenarios account for unemployment probabilities and unemployment benefits, except the scenario without unemployment benefit consideration.
Source: Authors' calculations.
Reduction in Untrained workers' UB Indexation:Without unemployment
benefit consideration
Reduction in Specialised workers' UB Indexation: Reduction in Both UB Indexation:
Absolute deviations from baselineHighIncome Economy: Sensitivity Analysis, comparison between Benchmark Model and Model without unemployment benefit (UB) consideration*
Reduction in Base Minimum Wage:Without unemployment
benefit considerationBenchmark
Advanced Education Cost Cut: Composite Reform Programme A:
Composite Reform Programme B: Composite Reform Programme C:
Reduction in Untrained workers' union markup: Reduction in Specialised workers' union markup:
195
Table 2.12: Middle-Income Economy: Sensitivity Analysis, comparison betweenBenchmark Model and Model without UB consideration
Baseline
steady state Benchmark
value Impact Steadystate Impact Steadystate Impact Steadystate Impact SteadystateShare of untrained workers 0.795 0.0005 0.0014 0.0000 0.0000 0.0003 0.0007 0.0000 0.0000Effective share of specialised workers 0.184 0.0004 0.0011 0.0000 0.0000 0.0002 0.0006 0.0000 0.0000Share of specialised workers in innovation 0.004 0.0000 0.0001 0.0000 0.0001 0.0000 0.0001 0.0000 0.0000Expected wage premium** 0.333 0.0024 0.0078 0.0000 0.0002 0.0057 0.0039 0.0002 0.0001Index of misallocation of talent 0.392 0.0010 0.0039 0.0006 0.0029 0.0011 0.0020 0.0008 0.0015Untrained unemployment rate 0.087 0.0085 0.0201 0.0084 0.0196 0.0004 0.0016 0.0003 0.0013Specialised unemployment rate 0.071 0.0001 0.0002 0.0002 0.0007 0.0000 0.0001 0.0002 0.0003Total unemployment rate 0.079 0.0065 0.0154 0.0065 0.0151 0.0003 0.0012 0.0003 0.0010Payroll contribution rate 0.052 0.0019 0.0073 0.0019 0.0074 0.0020 0.0037 0.0021 0.0038Growth rate of final output 0.039 0.0033 0.0002 0.0032 0.0001 0.0020 0.0001 0.0020 0.0001Social welfare 1.000 0.0040 0.0032 0.0761 0.0765
Baseline Impact Steadystate Impact Steadystate Impact Steadystate Impact SteadystateShare of untrained workers 0.795 0.0009 0.0023 0.0000 0.0000 0.0007 0.0016 0.0000 0.0001Effective share of specialised workers 0.184 0.0008 0.0018 0.0000 0.0000 0.0005 0.0013 0.0000 0.0000Share of specialised workers in innovation 0.004 0.0001 0.0000 0.0001 0.0001 0.0001 0.0001 0.0001 0.0002Expected wage premium** 0.333 0.0218 0.0156 0.0003 0.0002 0.0162 0.0117 0.0004 0.0003Index of misallocation of talent 0.392 0.0009 0.0003 0.0016 0.0021 0.0019 0.0023 0.0024 0.0035Untrained unemployment rate 0.087 0.0001 0.0001 0.0003 0.0011 0.0005 0.0015 0.0006 0.0023Specialised unemployment rate 0.071 0.0009 0.0019 0.0002 0.0004 0.0009 0.0018 0.0004 0.0007Total unemployment rate 0.079 0.0002 0.0003 0.0003 0.0009 0.0005 0.0015 0.0005 0.0019Payroll contribution rate 0.052 0.0018 0.0032 0.0017 0.0031 0.0038 0.0069 0.0038 0.0068Growth rate of final output 0.039 0.0063 0.0001 0.0062 0.0002 0.0082 0.0002 0.0081 0.0002Social welfare 1.000 0.2230 0.2183 0.2801 0.2772
Baseline Impact Steadystate Impact Steadystate Impact Steadystate Impact SteadystateShare of untrained workers 0.795 0.0003 0.0010 0.0000 0.0000 0.0005 0.0011 0.0000 0.0000Effective share of specialised workers 0.184 0.0003 0.0008 0.0000 0.0000 0.0004 0.0009 0.0000 0.0000Share of specialised workers in innovation 0.004 0.0000 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000 0.0001Expected wage premium** 0.333 0.0018 0.0058 0.0000 0.0002 0.0108 0.0077 0.0002 0.0001Index of misallocation of talent 0.392 0.0007 0.0029 0.0005 0.0022 0.0120 0.0121 0.0123 0.0129Untrained unemployment rate 0.087 0.0063 0.0149 0.0062 0.0145 0.0000 0.0000 0.0001 0.0005Specialised unemployment rate 0.071 0.0001 0.0001 0.0001 0.0005 0.0038 0.0043 0.0035 0.0036Total unemployment rate 0.079 0.0048 0.0114 0.0048 0.0112 0.0007 0.0007 0.0007 0.0010Payroll contribution rate 0.052 0.0014 0.0054 0.0014 0.0055 0.0008 0.0016 0.0008 0.0015Growth rate of final output 0.039 0.0024 0.0001 0.0024 0.0001 0.0149 0.0000 0.0149 0.0001Social welfare 1.000 0.0030 0.0023 0.0256 0.0247
Baseline Impact Steadystate Impact Steadystate Impact Steadystate Impact SteadystateShare of untrained workers 0.795 0.0205 0.0495 0.0176 0.0351 0.0008 0.0020 0.0000 0.0001Effective share of specialised workers 0.184 0.0161 0.0381 0.0138 0.0273 0.0006 0.0016 0.0000 0.0001Share of specialised workers in innovation 0.004 0.0008 0.0017 0.0007 0.0012 0.0001 0.0003 0.0001 0.0002Expected wage premium** 0.333 0.0882 0.1820 0.1158 0.2158 0.0045 0.0102 0.0003 0.0008Index of misallocation of talent 0.392 0.0142 0.0313 0.0123 0.0230 0.0037 0.0099 0.0031 0.0085Untrained unemployment rate 0.087 0.0045 0.0225 0.0038 0.0163 0.0225 0.0487 0.0224 0.0482Specialised unemployment rate 0.071 0.0132 0.0318 0.0114 0.0226 0.0002 0.0007 0.0007 0.0019Total unemployment rate 0.079 0.0007 0.0085 0.0006 0.0070 0.0173 0.0374 0.0173 0.0373Payroll contribution rate 0.052 0.0018 0.0031 0.0016 0.0021 0.0071 0.0200 0.0072 0.0202Growth rate of final output 0.039 0.0042 0.0016 0.0040 0.0012 0.0034 0.0005 0.0033 0.0004Social welfare 1.000 0.0579 0.0667 0.1682 0.1701
Baseline Impact Steadystate Impact Steadystate Impact Steadystate Impact SteadystateShare of untrained workers 0.795 0.0205 0.0489 0.0175 0.0351 0.0207 0.0493 0.0178 0.0358Effective share of specialised workers 0.184 0.0161 0.0377 0.0138 0.0273 0.0162 0.0380 0.0140 0.0278Share of specialised workers in innovation 0.004 0.0007 0.0019 0.0006 0.0014 0.0014 0.0028 0.0013 0.0022Expected wage premium** 0.333 0.0717 0.1579 0.0784 0.1804 0.0688 0.1559 0.0737 0.1774Index of misallocation of talent 0.392 0.0262 0.0476 0.0242 0.0399 0.0321 0.0552 0.0301 0.0471Untrained unemployment rate 0.087 0.0259 0.0603 0.0254 0.0572 0.0259 0.0605 0.0254 0.0574Specialised unemployment rate 0.071 0.0096 0.0263 0.0076 0.0175 0.0090 0.0257 0.0072 0.0172Total unemployment rate 0.079 0.0174 0.0371 0.0175 0.0381 0.0175 0.0373 0.0176 0.0382Payroll contribution rate 0.052 0.0058 0.0152 0.0061 0.0170 0.0060 0.0156 0.0063 0.0173Growth rate of final output 0.039 0.0071 0.0018 0.0068 0.0013 0.0575 0.0031 0.0573 0.0026Social welfare 1.000 0.1195 0.1130 0.1207 0.1145
*/ Calibrated based on LatinAmerica5. The exact details of each individual policy experiment listed are documented in their respective tables, in Table 2.62.10.**/ All scenarios account for unemployment probabilities and unemployment benefits, except the scenario without unemployment benefit consideration.
Source: Authors' calculations.
Composite Reform Programme C:
Reduction in Specialised workers' UB Indexation: Reduction in Both UB Indexation:
Reduction in Untrained workers' union markup: Reduction in Specialised workers' union markup:
Advanced Education Cost Cut: Composite Reform Programme A:
Composite Reform Programme B:
Without unemploymentbenefit consideration
Without unemploymentbenefit consideration
MiddleIncome Economy: Sensitivity Analysis, comparison between Benchmark Model and Model without unemployment benefit (UB) consideration*Absolute deviations from baseline
Reduction in Base Minimum Wage: Reduction in Untrained workers' UB Indexation:
Benchmark
196
Figure 2.1: Overview of Production Structure and Labour Market
Final Good Sector Innovation Sector
Stype labour supplyUtype labour supply
Wage rate, Stypelabour
Wage rates, Utype labourStype labour
Trade unionMinimum wage
Production Structure and the Labour Market
Training decision(Beginning of adulthood)
Unemployment
Intermediate GoodSector
Training cost
Unemployment benefits
Mandated compensation
Highest abilities
Blueprints
Efficiency wage
197
Figure 2.2: Individual and Composite Experiments: Steady-state effects
Individual and Composite Experiments: Steadystate effects(Absolute deviations from baseline)
Source: Authors' calculation.
Highincome economy Middleincome economy
Minimum wage cut
Untrained UB cut
Specialised UB cut
Untrained markup cut
Specialised markup cut
Training cost cut
Programme A
Programme B
Programme C
0.06 0.04 0.02 0.00
Total Unemployment Rate
Minimum wage cut
Untrained UB cut
Specialised UB cut
Untrained markup cut
Specialised markup cut
Training cost cut
Programme A
Programme B
Programme C
0.00 0.00 0.01
Final Output Growth Rate
Minimum wage cut
Untrained UB cut
Specialised UB cut
Untrained markup cut
Specialised markup cut
Training cost cut
Programme A
Programme B
Programme C
0.16 0.08 0.00 0.08 0.16
Social Welfare
Minimum wage cut
Untrained UB cut
Specialised UB cut
Untrained markup cut
Specialised markup cut
Training cost cut
Programme A
Programme B
Programme C
0.12 0.08 0.04 0.00
Index of Misallocation of Talent
198
Figure 2.3: Transitional Dynamics of Composite Reform Programme A
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0
0.005
0.004
0.003
0.002
0.001
0
0.06
0.05
0.04
0.03
0.02
0.01
0
0
0.0005
0.001
0.0015
0.002
0.0025
0.01
0
0.01
0.02
0.03
0.04
0.05
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.025
0.02
0.015
0.01
0.005
0
0.006
0.005
0.004
0.003
0.002
0.001
0
Share of untrained workers
Composite Reform Programme A(Absolute deviations from baseline)
Specialiseduntrained wage premium
Time
Payroll contribution rateGrowth rate of final output
Time
Time
Time
Time
Time
Time
Time
Index of misallocation of talent Share of specialised workers in innovation
Untrained unemployment rate Specialised unemployment rate
Highincome economy Middleincome economy
10 20 30 40 50 60
10 30 40 50 6020
10 30 40 50 6020
10 30 40 50 6020
10 20 30 40 50 60
10 20 30 40 50 60
10 20 30 40 50 60
10 30 40 50 6020
199
Figure 2.4: Transitional Dynamics of Composite Reform Programme B
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0
0.005
0
0.005
0.01
0.015
0.02
0.025
0.03
0.06
0.05
0.04
0.03
0.02
0.01
0
0
0.001
0.002
0.003
0.004
0.005
0.006
0.2
0.15
0.1
0.05
0
0
0.01
0.02
0.03
0.04
0.055
0.05
0.045
0.04
0.035
0.03
0.025
0.02
0.015
0.06
0.05
0.04
0.03
0.02
0.01
0
Share of untrained workers
Composite Reform Programme B(Absolute deviations from baseline)
Specialiseduntrained wage premium
Time
Payroll contribution rateGrowth rate of final output
Time
Time
Time
Time
Time
Time
Time
Index of misallocation of talent Share of specialised workers in innovation
Untrained unemployment rate Specialised unemployment rate
Highincome economy Middleincome economy
10 20 30 40 50 60
10 20 30 40 50 60
10 20 30 40 50 60
10 20 30 40 50 60
10 30 40 50 6020
10 30 40 50 6020
10 30 40 50 6020
10 30 40 50 6020
200
Figure 2.5: Transitional Dynamics of Composite Reform Programme C
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0
0.005
0
0.005
0.01
0.015
0.02
0.025
0.03
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0
0
0.001
0.002
0.003
0.004
0.005
0.006
0.007
0.008
0.009
0.2
0.15
0.1
0.05
0
0.05
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.06
0.05
0.04
0.03
0.02
0.01
0
Share of untrained workers
Composite Reform Programme C(Absolute deviations from baseline)
Specialiseduntrained wage premium
Time
Payroll contribution rateGrowth rate of final output
Time
Time
Time
Time
Time
Time
Time
Index of misallocation of talent Share of specialised workers in innovation
Untrained unemployment rateSpecialised unemployment rate
Highincome economy Middleincome economy
10 20 30 40 50 60
10 20 30 40 50 60
10 20 30 40 50 60
10 20 30 40 50 60
10 30 40 50 6020
10 30 40 50 6020
10 30 40 50 6020
10 30 40 50 6020
201
Figure2.6:SensitivityAnalysis:TransitionalDynamicsofAPermanentDecreaseinBaseMinimum
Wage
202
Figure2.7:SensitivityAnalysis:TransitionalDynamicsofAPermanentDecreaseinUntrainedWorkers’UBIndexation
203
Figure2.8:SensitivityAnalysis:TransitionalDynamicsofAPermanentDecreaseinSpecialisedWorkers’UBIndexation
204
Figure2.9:SensitivityAnalysis:TransitionalDynamicsofAPermanentDecreaseinBothUBIndexation
205
Figure2.10:SensitivityAnalysis:TransitionalDynamicsofAPermanentDecreaseinUntrainedUnion’sMark-UpoverTargetWage
206
Figure2.11:SensitivityAnalysis:TransitionalDynamicsofAPermanentDecreaseinSpecialisedUnion’sMark-UpoverTarget
Wage
207
Figure2.12:SensitivityAnalysis:TransitionalDynamicsofAPermanentDecreaseinAdvancedEducationCost
208
Summary and ConclusionThe thesis statements and contributions of this dissertation have been well-
discussed throughout the entire document. As such, I will only briefly summarise
the content of each chapters below.
In Chapter 1, a continuous time growth model with heterogeneous labour and
foreign MNCs is developed to examine industrial transformation in a developing
host economy. With FDI modelled at the disaggregated level of foreign experts,
a stylised foreign MNC composition-determination framework is formalised to ex-
plain Dunning’s ‘internalisation advantage’(1977) as being driven by the presence
of asymmetric views of foreign experts on the productivity of domestic workers. As
productivity is a transformation of ability, the skills acquisition decision and foreign
subsidiaries’operational mode choice are determined along the same ability distrib-
ution in the model. These, coupled with the modelling of an additional asymmetry
between Vertical MNCs and other MNCs, enable the model to be parameterised
and analysed using numerical policy experiments. Both the long-run properties and
transitional dynamics are examined, with the results obtained largely consistent
with some well-documented stylised facts in the FDI literature. In short, the key
policy implications derived based on the analysis include: (i) the implementation
of foreign investment liberalisation measures in a typical developing host economy
would not be a matter of straightforward provision of investment incentives. In-
deed, in the presence of asymmetries, an investment liberalisation measure that
is balanced and targeting all types of foreign firms is more innovation- and skills
acquisition-promoting than disproportionate ones biased towards selected types of
foreign firms; (ii) it is important to combine human capital and FDI-promoting
policies in promoting industrial transformation, especially if the government of a
host economy intends to minimise disruption of industrial transformation; and (iii)
the policy complementarities are stronger the higher the technological diffusion rate
within a developing host economy.
209
In Chapter 2, the unemployment, growth, and welfare effects of labour market
reforms are examined in an innovation-driven, OLG model of endogenous growth
with a heterogeneous labour force, labour market rigidities, and structural unem-
ployment. The chapter also introduces an interesting concept of misallocation of
talent, which allows for an additional effi ciency assessment of policy outcomes. The
model is parameterised for stylised high- and middle-income economies and used
to perform a range of policy experiments. These include both individual labour
market reforms and composite reform programmes. Both the steady-state proper-
ties and transitional dynamics of these policies are again examined in this chapter.
Two-way causality between growth and unemployment is documented for all the
policies examined, with a long-run growth-welfare tradeoff consistently observed for
some individual labour market policies, such as the cuts in unemployment benefit
indexation (which is especially bad for welfare). Moreover, it is found that the pop-
ular policy recommendation of an ambitious expansion in tertiary education (via a
drastic cut in effective cost of education), while growth- and welfare-enhancing, can
create an absorption or oversupply problem for the specialised workers (à la Spain),
if it were not accompanied by corresponding measures promoting labour demand.
Governments must therefore refrain from adopting policies that contribute to a vast
increase in the numbers of university graduates, and focuses instead on quality- or
productivity-enhancing measures. One such policy examined in this chapter is the
public investment in infrastructure. Public investment in infrastructure, through its
productivity-enhancing effects across all sectors, may help to boost employment and
mitigate the oversupply problem. Lastly, in terms of policy choice in a composite
reform, the analysis suggests that, if unemployment or social welfare matters more
than growth to policymakers, comprehensive reform programmes may generate neg-
ative externalities. As such, overly ambitious labour market reform programmes,
notably those with unemployment benefit cuts, may be costly and ineffective.
210
Bibliography[1] Acemoglu, Daron, and Jaume Ventura, “The World Income Distribution,”
Quarterly Journal of Economics, 110 (May 2002), 659-94.
[2] Adascalitei, Dragos, and Clemente Pignatti Morano, “Labour Market Re-
forms since the Crisis: Drivers and Consequences,”Research Department Working
Paper No. 5, International Labour Offi ce (October 2015).
[3] Afonso, António, and Miguel St. Aubyn, “Macroeconomic Rates of Return of
Public and Private Investment: Crowding-in and Crowding-out Effects,”Manchester
School, 77 (September 2009), 21-39.
[4] Afonso, António, Ludger Schuknecht, and Vito Tanzi, “Public Sector Effi -
ciency: An International Comparison,”Working Paper No. 242, European Central
Bank (July 2003).
[5] Agénor, Pierre-Richard, “Schooling and Public Capital in a Model of Endoge-
nous Growth,”Economica, 78 (January 2011), 108-32.
[6] – – , Public Capital, Growth and Welfare, Princeton University Press (Prince-
ton, New Jersey: 2012).
[7] – – , “Caught in the Middle? The Economics of Middle-Income Traps,”
unpublished, University of Manchester (March 2016). Forthcoming, Journal of Eco-
nomic Surveys.
[8] Agénor, Pierre-Richard, and Joshua Aizenman, “Macroeconomic Adjustment
with Segmented Labor Markets,” Journal of Development Economics, 58 (April
1999), 277-96.
[9] Agénor, Pierre-Richard, and Baris Alpaslan, “Infrastructure and Industrial
Development with Endogenous Skill Acquisition,”Centre for Growth and Business
Cycle Research Discussion Paper No. 195, University of Manchester (October 2014).
[10] Agénor, Pierre-Richard, and Otaviano Canuto, “Gender Equality and Eco-
nomic Growth in Brazil: A Long-Run Analysis,” Journal of Macroeconomics, 43
(March 2015a), 155-72.
211
[11] – – , “Middle-income Growth Traps,”Research in Economics, 69 (December
2015b), 641-60.
[12] Agénor, Pierre-Richard, and Hinh T. Dinh, “Public Policy and Industrial
Transformation in the Process of Development,” Policy Research Working Paper
No. 6405, World Bank (April 2013).
[13] Agénor, Pierre-Richard, and Khazanah team, “Khazanah Modeling Project:
MV20 Model Technical Report,”unpublished, Khazanah Nasional (Kuala Lumpur:
2012).
[14] Agénor, Pierre-Richard, and Peter Montiel, Development Macroeconomics,
Princeton University Press (New Jersey: 2008).
[15] Agénor, Pierre-Richard, and Kyriakos C. Neanidis, “Innovation, Public Cap-
ital, and Growth,”Journal of Macroeconomics, 44 (June 2015), 252-75.
[16] Allanson, Paul and Catia Montagna, “Multiproduct firms and market struc-
ture: An explorative application to the product life cycle,”International Journal of
Industrial Organization, 23 (September 2005), 587-97.
[17] Almeida, Rita, Jere Behrman, and David Robalino, The Right Skills for the
Job: Rethinking Policies for Workers, World Bank Publications (Washington DC:
2012).
[18] Amsden, Alice, The Rise of “The Rest”: Challenges to the West from Late-
Industrializing Economies, Oxford University Press (New York: 2001).
[19] Ang, James B., and Jakob B. Madsen, “What Drives Ideas Production across
the World,”Macroeconomic Dynamics, 19 (January 2015), 79-115.
[20] Athukorala, Prema-chandra, “Product Fragmentation and Trade Patterns
in East Asia,”Asian Economic Papers, 4 (October 2005), 1-27.
[21] Athukorala, Prema-chandra, and Hal Hill, “Asian Trade: Long-term Pat-
terns and Key Policy Issues,” Asian-Pacific Economic Literature, 24 (November
2010), 52-82.
[22] Bell, Martin, and Paulo N. Figueiredo, “Building Innovative Capabilities
212
in Latecomer Emerging Market Firms: Some Key Issues,” in Innovative Firms in
Emerging Market Countries, ed. by Edmund Amann and John Cantwell, Oxford
University Press (Oxford: 2012), 24-109.
[23] Benhabib, Jess, Jesse Perla, and Christopher Tonetti, “Catch-up and Fall-
back through Innovation and Imitation,”Journal of Economic Growth, 19 (March
2014), 1-35.
[24] Bernal-Verdugo, Lorenzo E., Davide Furceri, Dominique M. Guillaume, “La-
bor Market Flexibility and Unemployment: New Empirical Evidence of Static and
Dynamic Effects,”Working Paper No. 12/64, International Monetary Fund (March
2012).
[25] Bhattacharyya, Chandril, and Manash R. Gupta, “Unionised Labour Mar-
ket, Unemployment Allowances, Productive Public Expenditure and Endogenous
Economic Growth,”Metroeconomica, 66 (July 2015), 397-425.
[26] Blanchflower, David, and Alex Bryson, “Changes Over Time in Union Rel-
ative Wage Effects in the UK and the US Revisited,”Working Paper No. 9395,
National Bureau of Economic Research (August 2002).
[27] Blomström, Magnus, “The Economics of International Investment Incen-
tives,” in International Investment Perspectives ed. by Hans Christiansen, OECD
Publishing (Paris: 2002), 165—83.
[28] Blomström, Magnus, and Ari Kokko, , “Human Capital and Inward FDI,”
Working Paper No. 167, European Institute of Japanese Studies (January 2003).
[29] Blomström, Magnus, and Fredrik Sjöholm, “Technology Transfer and Spillovers:
Does Local Participation with Multinationals matter?,”European Economic Review,
43 (April 1999), 915-23.
[30] Böhm, Sebastian, Volker Grossmann, and Thomas M. Steger, “Does Ex-
pansion of Higher Education Lead to Trickle-down Growth?,” Journal of Public
Economics, 132 (December 2015), 79-94.
[31] Bom, Pedro R., and Jenny E. Ligthart, “What Have we Learned from Three
213
Decades of Research on the Productivity of Public Capital?,”Journal of Economic
Surveys, 28 (December 2014), 889-916.
[32] Borensztein, E., Jose De Gregorio, and Jong-Wha Lee, “How does Foreign
Direct Investment affect Economic Growth?,”Journal of International Economics,
45 (June 1998), 115-35.
[33] Bouis, Romain, and Romain Duval, “Raising Potential Growth After the
Crisis,”Economics Department Working Papers, No. 835, OECD Publishing (Paris:
2011).
[34] Braconier, Henrik, Pehr-Johan Norbäck, and Dieter Urban, “Multinational
Enterprises and Wage Costs: Vertical FDI revisited,”Journal of International Eco-
nomics, 67 (December 2005), 446-70.
[35] Brainard, S Lael, “An Empirical Assessment of the Proximity-Concentration
Trade-off between Multinational Sales and Trade,”American Economic Review, 87
(September 1997), 520-44.
[36] Brambilla, Irene, Galina Hale, and Cheryl Long, “Foreign Direct Investment
and the Incentives to Innovate and Imitate,”Scandinavian Journal of Economics,
111 (December 2009), 835-61.
[37] Breda, Thomas, “Firms’Rents, Workers’Bargaining Power and the Union
Wage Premium,”Economic Journal, 125 (December 2015), 1616-52.
[38] Briglauer, Wolfgang, Christian Holzleitner, and Ingo Vogelsang, “The Need
for more Effi cient Public Funding of New Communications Infrastructure in EU
Member States,”Information Economics and Policy, forthcoming (September 2016).
[39] Brunner, Martin, and Holger Strulik, “Solution of Perfect Foresight Saddle-
point Problems: A Simple Method and Applications,”Journal of Economic Dynam-
ics and Control, 26 (May 2002), 737-53.
[40] Bucci, Alberto, Fabio Fiorillo, and Stefano Staffolani, “Can Market Power
Influence Employment, Wage Inequality and Growth?,”Metroeconomica, 54 (May
2003), 129-60.
214
[41] Cacciatore, Matteo, and Giuseppe Fiori, “The Macroeconomic Effects of
Goods and Labor Market Deregulation,”Review of Economic Dynamics, 20 (March
2016), 1-24.
[42] Cacciatore, Matteo., Romain Duval, and Giuseppe Fiori, “Short-Term Gain
or Pain? A DSGE Model-Based Analysis of the Short-Term Effects of Structural
Reforms in Labour and Product Markets,”Economics Department Working Papers,
No. 948, OECD Publishing (Paris: 2012).
[43] Cahuc, Pierre, and Philippe Michel, “Minimum Wage, Unemployment and
Growth,”European Economic Review, 40 (August 1996), 1463-82.
[44] Calderón, César, and Luis Servén, “Infrastructure in Latin America,”Policy
Research Working Paper No. 5317, World Bank (May 2010).
[45] Calderón, César, Enrique Moral-Benito, and Luis Servén, “Is Infrastruc-
ture Capital Productive? A Dynamic Heterogeneous Approach,”Journal of Applied
Econometrics, 30 (June 2015), 177-98.
[46] Carranza, Luis, Christian Daude, and Angel Melguizo, “Public Infrastruc-
ture Investment and Fiscal Sustainability in Latin America: Incompatible Goals?”
Journal of Economic Studies, 41 (January 2014), 29-50.
[47] Chang, Juin-jen, and Hsiao-wen Hung, “Trade Unions, Unemployment,
Economic Growth and Income Inequality,”Macroeconomic dynamics, 20 (January
2016), 404-28.
[48] Chen, Xi, and Michael Funke, “The Dynamics of Catch-up and Skill and
Technology Upgrading in China,”Journal of Macroeconomics, 38 (December 2013),
465-80.
[49] Ciconne, Antonio, and Elias Papaioannou, “Human Capital, the Structure of
Production, and Growth,”Review of Economics and Statistics, 91 (February 2009),
66-82.
[50] Cortazar, Rene, “Unemployment Insurance Systems for Latin America,”
in Labor Market Policies in Canada and Latin America: Challenges of the New
215
Millennium, ed. by Albert Berry, Springer (New York: 2001), 97-107.
[51] Dabla-Norris, Era, Jim Brumby, Annette Kyobe, Zac Mills, and Chris Pa-
pageorgiou, “Investing in Public Investment: An Index of Public Investment Effi -
ciency,”Journal of Economic Growth, 17 (September 2012), 235-66.
[52] Daveri, Francesco, and Guido Tabellini, “Unemployment, Growth and Tax-
ation in Industrial Countries,”Economic Policy, 30 (April 2000), 49-104.
[53] D’Costa, Anthony P., “Software Outsourcing and Development Policy Impli-
cations: An Indian Perspective,”International Journal of Technology Management,
24 (July-August 2002), 705-23.
[54] De la Croix, David, and Philippe Michel, A Theory of Economic Growth:
Dynamics and Policy in Overlapping Generations, Cambridge University Press (Oc-
tober 2002).
[55] Dinopoulos, Elias, and Paul S. Segerstrom, “A Schumpeterian Model of
Protection and Relative Wages,”American Economic Review, 89 (September 1999),
450-72.
[56] Djankov, Simeon, and Bernard Hoekman, “Foreign Investment and Produc-
tivity Growth in Czech Enterprises,”World Bank Economic Review, 14 (January
2000), 49-64.
[57] Dunning, John H., “Trade, Location of Economic Activity and the Multina-
tional Enterprise: A Search for an Eclectic Approach,” in The International Allo-
cation of Economic Activity, ed. by B. Ohlin, P.O. Hesselborn and P.M. Wijkman,
Holmes and Meier (London: 1977), 395-418.
[58] Dunning, John H., and Sarianna M. Lundan, Multinational Enterprises and
the Global Economy, Edward Elgar (Cheltenham: 2008).
[59] Eicher, Theo, and Cecilia García-Peñalosa, “Inequality and Growth: The
Dual Role of Human Capital in Development,”Journal of Development Economics,
66 (October 2001), 173-97.
[60] Faeth, Isabel, “Determinants of Foreign Direct Investment - A Tale of Nine
216
Theoretical Models,”Journal of Economic Surveys, 23 (February 2009), 165-96.
[61] Felipe, Jesus, Arnelyn Abdon, and Utsav Kumar, “Tracking the Middle-
Income Trap: What Is It, Who Is in It, and Why?,”Working Paper No. 715, Levy
Economics Institute (April 2012).
[62] Ferreira, Pedro C., Samuel D. Pessoa, and Fernando A. Veloso, “On the
Evolution of Total Factor Productivity in Latin America,”Economic Inquiry, 51
(January 2013), 16-30.
[63] Ferrer, Ana, and W. Craig Riddell, “Unemployment Insurance Savings Ac-
counts in Latin America: Overview and Assessment,”Social Protection Discussion
Paper No. 910, World Bank (June 2009).
[64] Flaaen, Aaron, Ejaz Ghani, and Saurabh Mishra, “How to Avoid Middle
Income Traps? Evidence from Malaysia,”Policy Research Working Paper No. 6427,
World Bank (April 2013).
[65] Flynn, James R., What is Intelligence?, Cambridge University Press (New
York: 2007).
[66] Funke, Michael, and Holger Strulik, “On Endogenous Growth with Physical
Capital, Human Capital, and Product Variety,” European Economic Review, 44
(March 2000), 491-515.
[67] Gancia, Gino, and Fabrizio Zilibotti, “Horizontal Innovation in the Theory
of Growth and Development,” in Handbook of Economic Growth, Vol. 1A, ed. by
Philippe Aghion and Steven Durlauf, North Holland (Amsterdam: 2005).
[68] Gandelman, Néstor, “A Comparison of Saving Rates: Micro Evidence from
Seventeen Latin American and Caribbean Countries,” Working Paper No. 602,
Inter-American Development Bank (July 2015).
[69] Gander, James, Steve Reynolds, and Richard Fowles, “FDI Flow Volatil-
ity and ASEAN Members: An Exploratory Approach,”Department of Economics
Working Paper No. 6, University of Utah (Utah: 2009).
[70] Gersbach, Hans, and Armin Schmutzler, “Foreign Direct Investment and
217
R&D-Offshoring,”Oxford Economic Papers, 63 (January 2011), 134-57.
[71] Glass, Amy J., “Imitation as a Stepping Stone to Innovation,”unpublished,
Texas A&M University (June 2010).
[72] Haaland, Jan I., and Ian Wooton, “Multinational Firms: Easy Come, Easy
Go?”CEPR Discussion Papers, 2660 (January 2001).
[73] Haddad, Mona, and Anne E. Harrison, “Are there Positive Spillovers from
Direct Foreign Investment?: Evidence from Panel Data for Morocco,” Journal of
Development Economics, 42 (October 1993), 51-74.
[74] Hanson, Gordon, Raymond J. Mataloni, Jr., and Matthew J. Slaughter, “Ex-
pansion Strategies of U.S. Multinational Firms,”Working Paper No. 8433, National
Bureau of Economic Research (August 2001).
[75] Heer, Burkhard, and Albrecht Morgenstern, “The Labor Market Effects of
Indexing Unemployment Benefits to Previous Earnings,”Public Finance Review, 33
(May 2005), 385-402.
[76] Helpman, Elhanan, “A Simple Theory of International Trade with Multina-
tional Corporations,”Journal of Political Economy, 92 (June 1984), 451-71.
[77] Hill, Hal, Tham S. Yean, and H. Ragayah, “Malaysia: A Success Story Stuck
in the Middle?”World Economy, 35 (December 2012), 1687-711.
[78] Hobday, Michael, Howard Rush, and John Bessant, “Approaching the Inno-
vation Frontier in Korea: The Transition Phase to Leadership,”Research Policy, 33
(December 2004), 1433-57.
[79] Horstmann, Ignatius J., and James R. Markusen, “Strategic Investments and
the Development of Multinationals,”International Economic Review, 28 (February
1987), 109-21.
[80] – – —, “Endogenous Market Structures in International Trade,”Journal of
International Economics, 32 (February 1992), 109-29.
[81] – – —, “Exploring New Markets: Direct Investment, Contractual Relations
and the Multinational Enterprise,” International Economic Review, 37 (February
218
1996). 1-19.
[82] Iacopetta, Maurizio, “Formal Education and Public Knowledge,”Journal of
Economic Dynamics and Control, 35 (May 2011), 676-93.
[83] Inter-American Development Bank (IDB), Good Jobs Wanted: Labor Mar-
kets in Latin America, IDB (Washington DC: 2004).
[84] – – —, Rethinking Productive Development: Sound Policies and Institutions
for Economic Transformation, IDB publications (Washington DC: 2014).
[85] International Monetary Fund, “Time for a Supply-Side Boost? The Macro-
economic Effects of Labor and Product Market Reforms in Advanced Economies,”
Chapter 3 in World Economic Outlook, IMF Publications (Washington DC: 2016).
[86] Javorcik, Beata Smarzynska, “Does Foreign Direct Investment Increase the
Productivity of Domestic Firms? In Search of Spillovers Through Backward Link-
ages,”American Economic Review, 94 (June 2004), 605-27.
[87] Jones, Charles I., “Growth and Ideas,” in Handbook of Economic Growth,
ed. by Philippe Aghion and Steven Durlauf, Elsevier , 1B (Amsterdam: 2005),
1063-111.
[88] Judd, Kenneth L., Numerical Methods in Economics, MIT Press (Cam-
bridge, MA: 1998).
[89] Kam, Andrew, “International Production Networks and Host Country Pro-
ductivity: Evidence from Malaysia,”Asian-Pacific Economic Literature, 27 (May
2013), 127-46.
[90] Kamps, Christophe, “New Estimates of Government Net Capital Stocks for
22 OECD Countries, 1960—2001,”IMF Staff Papers, 53 (April 2006), 120-50.
[91] Kharas, Homi, Albert Gaspard Zeufack, and Hamdan Majeed, Cities, People
& the Economy: A Study on Positioning Penang, Khazanah Nasional and World
Bank (Kuala Lumpur: 2010).
[92] Kottaridi, Constantina, and Thanasis Stengos, “Foreign Direct Investment,
Human Capital and Non-linearities in Economic Growth,” Journal of Macroeco-
219
nomics, 32 (September 2010), 858-71.
[93] Leuven, Edwin, and Hessel Oosterbeek, “Overeducation and mismatch in the
labor market,”in Handbook of the Economics of Education, ed. by E. Hanushek, S.
Machin, and L. Woessmann, 4, North Holland (Amsterdam: 2011), 283—326.
[94] Lim, King Yoong, “Is Innovation Shortfall in Malaysia a Structural Issue?
Evidence from PICS on FDI, Ownership Mode, and Investment Climate,”unpub-
lished, University of Manchester (2015).
[95] Liu, Zhiqiang, “Foreign Direct Investment and Technology Spillovers: The-
ory and Evidence,”Journal of Development Economics, 85 (February 2008), 176-93.
[96] Loayza, Norman, Pablo Fajnzylber, and César Calderón, Economic Growth
in Latin America and the Caribbean: Stylized Facts, Explanations, and Forecasts,
World Bank (March 2005).
[97] Lora, Eduardo, and Deisy J. Fajardo, “Employment and Taxes in Latin
America: An Empirical Study of the Effects of Payroll, Corporate Income and
Value-added Taxes on Labor Outcomes,”Working Paper No. 334, Inter-American
Development Bank (September 2012).
[98] LSE Growth Commission, Investing for Prosperity: Skills, Infrastructure
and Innovation, London School of Economics (London: 2013).
[99] Lucas, Robert E., and Benjamin Moll, “Knowledge Growth and the Alloca-
tion of Time,”Journal of Political Economy, 122 (February 2014), 1-51.
[100] Lusinyan Lusine, and Dirk Muir, “Assessing the Macroeconomic Impact of
Structural Reforms: The Case of Italy,”Working Paper Series, 13/22, International
Monetary Fund (New York: 2013).
[101] Machin, Stephen, and Sandra McNally, “Tertiary Education Systems and
Labour Markets,”Education and Training Policy Division, OECD (Paris: 2007).
[102] Markusen, James R., “Multinationals, Multi-plant Economies, and the
Gains from Trade,”Journal of International Economics, 16 (May 1984), 205-26.
[103] – – —, “The Boundaries of Multinational Enterprises and the Theory of
220
International Trade,”Journal of Economic Perspectives, 9 (Spring 1995), 169-89.
[104] – – —, “Multinational Firms, Location and Trade,”World Economy, 21
(August 1998), 733-56.
[105] Markusen, James R., and Keith E. Maskus, “Discriminating among Al-
ternative Theories of the Multinational Enterprise,”Review of International Eco-
nomics, 10 (November 2002), 694-707.
[106] Markusen, James R., and Natalia Trofimenko, “Teaching Locals New Tricks:
Foreign Experts as a Channel of Knowledge Transfers,” Journal of Development
Economics, 88 (January 2009), 120-31.
[107] Markusen, James R., and Anthony J. Venables, “Foreign Direct Investment
as a Catalyst for Industrial Development,”European Economic Review, 43 (February
1999), 335-56.
[108] McDonnell, Anthony, Ryan Lamare, Patrick Gunnigle, and Jonathan Lavelle,
“Developing Tomorrow’s Leaders– Evidence of Global Talent Management in Multi-
national Enterprises,”Journal of World Business, 45 (April 2010), 150-60.
[109] Meckl, Jurgen, “Effi ciency-Wage Unemployment and Endogenous Growth,”
Labour, 15 (December 2001), 579-602.
[110] – – , “Accumulation of Technological Knowledge, Wage Differentials, and
Unemployment,”Journal of Macroeconomics, 26 (March 2004), 65-82.
[111] Melitz, Marc J., “The Impact of Trade on Intra-Industry Reallocations and
Aggregate Industry Productivity,”Econometrica, 71 (November 2003), 1695-725.
[112] Mercenier, Jean, and Philippe Michel, “Discrete-Time Finite Horizon Ap-
promixation of Infinite Horizon Optimization Problems with Steady-State Invari-
ance,”Econometrica, 62 (May 1994), 635-56.
[113] Millán, Ana, José María Millán, Concepción Román, and André van Stel,
“How does Employment Protection Legislation Influence Hiring and Firing Decisions
by the Smallest Firms?,”Economics Letters, 121 (December 2013), 444-8.
[114] Montuenga, Víctor, Inmaculada García, and Melchor Fernández, “Wage
221
Flexibility: Evidence from Five EU Countries based on the Wage Curve,”Economic
Letters, 78 (February 2003), 169-74.
[115] Mortensen, Dale T., and Christopher A. Pissarides, “Job Reallocation,
Employment Fluctuations and Unemployment,” in Handbook of Macroeconomics,
ed. by J.B. Taylor and M. Woodford, Vol 1, Elsevier (Amsterdam: 1999), 1171-228.
[116] Nelson, Richard R., and Howard Pack, “The Asian Miracle and Modern
Growth Theory,”Economic Journal, 109 (July 1999), 416-36.
[117] Neumark, David, and William Wascher, “Minimum Wages and Employ-
ment: A Review of Evidence from the NewMinimumWage Research,”NBERWork-
ing Papers 12663, National Bureau of Economic Research (Massachusetts: 2006).
[118] OECD, “Tax Incentives for Investment: A Global Perspective Experiences
in MENA and Non-MENACountries,”inMaking Reforms Succeed: Moving Forward
with the MENA Investment Policy Agenda, OECD Publishing (Paris: 2008), 225-61.
[119] – – , Education at a Glance 2015, OECD (Paris: 2016).
[120] Olney, William W., “A Race to the Bottom? Employment Protection
and Foreign Direct Investment,”Journal of International Economics, 91 (November
2013), 191-203.
[121] Oman, Charles, Policy Competition for Foreign Direct Investment, OECD
Development Centre Studies (Paris: 2000).
[122] Parello, Carmelo P., “Labor Market Rigidity and Productivity Growth in a
Model of Innovation-Driven Growth,”Economic Modelling, 28 (May 2011), 1058-67.
[123] Perez-Sebastian, Fidel, “Public Support to Innovation and Imitation in a
Non-Scale Growth Model,” Journal of Economic Dynamics and Control, 31 (De-
cember 2007), 3791-821.
[124] Pessoa, Argentino, “‘Ideas’Driven Growth: The OECD Evidence,”Por-
tuguese Economic Journal, 4 (March 2005), 46-67.
[125] PricewaterhouseCoopers, Talent Mobility 2020: The Next Generation of
International Assignments, PwC (New York: 2012).
222
[126] Rasiah, Rajah, “Is Malaysia’s Electronics Industry Moving Up the Value
Chain?”inMalaysian Development Challenges: Graduating from the Middle, ed. by
H. Hill, S.Y. Tham and M.Z. Ragayah, Routledge (London: 2012), 194-212.
[127] Romer, Paul M., “Endogenous Technological Change,”Journal of Political
Economy (October 1990), S71-102.
[128] Rustichini, Aldo, and James A. Schmilz Jr., “Research and Imitation in
Long-run Growth,”Journal of Monetary Economics, 27 (April 1991), 271-92.
[129] Saggi, Kamal, “Trade, Foreign Direct Investment, and International Tech-
nology Transfer: A Survey,”World Bank Research Observer, 17 (Fall 2002), 191-235.
[130] Sander, Frederico Gil, and Marek Hanusch, Malaysia Economic Monitor:
Modern Jobs, World Bank (Washington, DC: 2012).
[131] Scullion, Hugh, and Chris Brewster, “The Management of Expatriates:
Messages from Europe?”Journal of World Business, 36 (Winter 2001), 346-65.
[132] Scullion, Hugh, David G. Collings, and Patrick Gunnigle, “International
Human Resource Management in the 21st Century: Emerging Themes and Con-
temporary Debates,”Human Resource Management Journal, 17 (November 2007),
309-19.
[133] Sequeira, Tiago Neves, “R&D Spillovers In An Endogenous Growth Model
With Physical Capital, Human Capital, And Varieties,”Macroeconomic Dynamics,
15 (April 2011), 223-39.
[134] Serebrisky, Tomás, Ancor Suárez-Alemán, Diego Margot, and Maria Cecilia
Ramirez, Financing Infrastructure in Latin America and the Caribbean: How, how
much and by whom?, Inter-American Development Bank (Washington DC: 2015).
[135] Shapiro, Carl, and Joseph E. Stiglitz, ”Equilibrium Unemployment as a
Worker Discipline Device,”American Economic Review, 74 (June 1984), 433-44.
[136] Tafunell, Xavier, and Cristian Ducoing, “Non-Residential Capital Stock in
Latin America, 1875-2008: New Estimates and International Comparisons,”Aus-
tralian Economic History Review, 56 (March 2016), 46-69.
223
[137] Trimborn, Timo, Karl-Josef Koch, and Thomas M. Steger, “Multidimen-
sional Transitional Dynamics: A Simple Numerical Procedure,”Macroeconomic Dy-
namics, 12 (June 2008), 301-19.
[138] Van Schaik, A., and H. de Groot, “Unemployment, Growth and Effi ciency
Wages,” in Growth, Unemployment and Deindustrialization, ed. by H. de Groot,
Edward Elgar (Cheltenham: 2000).
[139] Varga, Janos, Werner Roeger, and Jan in’t Veld, “Growth Effects of Struc-
tural Reforms in Southern Europe: The Case of Greece, Italy, Spain and Portugal,”
Empirica, 41 (May 2014), 323-63.
[140] Vogel, David, and Robert A. Kagan, Dynamics of Regulatory Change: How
Globalization Affects National Regulatory Policies, University of California Press
(Berkeley: 2004).
[141] Wauthy, Xavier, and Yves Zenou, “Effi ciency Wages, Labor Heterogeneity
and the Financing of the Training Cost,”Discussion Paper No. 1997072, Université
catholique de Louvain (Louvain: 1997).
[142] World Bank, Jobs, World Bank publications (Washington DC: 2012a).
[143] – – , Golden Growth: Restoring the Lustre of the European Economic
Model, World Bank publications (Washington DC: 2012b).
[144] World Economic Forum, Global Competitiveness Reports 2015-2016 (Geneva:
2015).
[145] Yamada, Gustavo, “The Boom in University Graduates and the Risk of
Underemployment,”IZA World of Labor No. 165 (July 2015).
[146] Yusuf, Shahid, and Kaoru Nabeshima, Tiger Economies Under Threat : A
Comparative Analysis of Malaysia’s Industrial Prospects and Policy Options, World
Bank Publications (Washington, DC: 2009).
[147] Zagler, Martin, “Economic Growth, Structural Change, and Search Unem-
ployment,”Journal of Economics, 96 (January 2009), 63-78.
[148] – – , “Endogenous Growth, Effi ciency Wages, and Persistent Unemploy-
224
ment,”Review of Economics and Finance, 1 (Februrary 2011), 34-42.
[149] Zeufack, Albert, and King Yoong Lim, Can Malaysia Achieve Innovation-
led Growth?, Khazanah Nasional (Kuala Lumpur: 2013).
225