Essays on Human Capital, Innovation, and Growth with

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


in Untrained Workers’UB Indexation . . . . . . . . . . . . . . . . . . 203


in Specialised Workers’UB Indexation . . . . . . . . . . . . . . . . . 204


in Both UB Indexation . . . . . . . . . . . . . . . . . . . . . . . . . . 205

6


in Untrained Union’s Mark-Up over Target Wage . . . . . . . . . . . 206


in Specialised Union’s Mark-Up over Target Wage . . . . . . . . . . . 207


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


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


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


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(


)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)(


)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


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


Ψ 0.3672 -0.0931 -0.1294 -0.0764ν 0.5700 -0.0532 -0.0055 -0.0859


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


Time


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



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


Time


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



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


Time


Time


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



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


Time


Time


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



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


Time


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



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


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)


θSt)]

KPt+1

KPt

= (σ

1 + ς t)YtKPt

(1− τ)βU + κUθULt (1 + ς t) (2.49)

+(1− τ)

1− µ(1− τ)


θSt)

[θSYt + θSRt

κR


βS

θSYt

+

ΩSθ

St (θSLt )κ

S − µ(1− τ)


κR


βSθSLtθSYt θSt

,

kGt =KGt

KPt

, (2.50)

YtKPt

=(kG)ω/(1−γ)Λ2


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


, (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


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


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


if j = E

if j = L,

cSR,,jt|t = [ηC(1 + ρ)

1 + ηC(1 + ρ)]

(1− τ)(1− ε)wSRt − 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)



if j = E

if j = L,

cSR,jt|t = (1− σ)



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



if j = E

if j = L, (2.70)

sSR,jt|t = σ



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)



if j = E

if j = L, (2.73)

cSR,jt|t+1 = σ(1 + rt+1)



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


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


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 =




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


× [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


× [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


+(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


+(1− τ)βS

θSYt

κRθSRtθSt

[θSYt + ΩS(θSLt )1+κS ]

− µ(1−τ)

[0.5(1+aCt )]χθSRtθSt

(θSYt + θSRt κR


+

ΩSθ

St (θSLt )κ

S − µ(1− τ)


κR


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)


θSt)

[θSYt + θSRt

κR


βS

θSYt

+

ΩSθ

St (θSLt )κ

S − µ(1− τ)


κR


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)

[1− µ(1− τ)


θSt)

][θSYt + θSRt

κR


βS

θSYt

+

ΩSθ

St (θSLt )κ

S − µ(1− τ)


κR


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,


βU + βS


, (2.102)

From (2.19) and (2.20),

Qt = (1− η)γ(YtMt

),

or equivalently, noting that Yt/Mt = (Yt/KPt )(mt)

−1,


)(mt)−1, (2.103)

179

where Yt/KPt is derived from equation (2.46):

YtKPt

=(kGt )ω/(1−γ)Λ2


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)

1− µ(1− τ)


θSt)

[θSYt + θSRt

κR


βS

θSYt

+

ΩSθ

St (θSLt )κ

S − µ(1− τ)


κR


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)


θSt)]

KPt+1

KPt

= (σ

1 + ς t)YtKPt

(1− τ)βU + κUθULt (1 + ς t) (2.109)

+(1− τ)

1− µ(1− τ)


θSt)

[θSYt + θSRt

κR


βS

θSYt

+

ΩSθ

St (θSLt )κ

S − µ(1− τ)


κR


βSθSLtθSYt θSt

,

together with the static equations

kGt =KGt

KPt

, (2.110)

YtKPt

=(kG)ω/(1−γ)Λ2


U]−1/(1−γ)

mt

(1−η)/ηγ/(1−γ)

, (2.111)


)(mt)−1, (2.112)

θUt = µ1/χ

1− ζUYt (1− τ) + [ζULt κU − (1− ε)ζSLt κS]ΩU(θULt )κ

U


βU

βS(θSYtθUYt

)

−1/χ

,

(2.113)

θSt =1− (θUt )2

2, (2.114)

θSRt =




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)


βU + βS


, (2.120)

ζUYt =θUYtθUt


, (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


+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



+κRκSζSRt ζSLt

YtN

+ΛV ,

(2.132)

V S,Lt = (ηC +

1

1 + ρ) ln

(1− ε)

κS − µ(1− τ)

(θUt )χ



+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



+ 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



+ ln(1− τ)(1− ε)

(

κRβS ζSY

(1+ς)(1−ε)θSY + κRκS ζSL)

− µ

(θU )χ



+ 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


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


Steady state


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.


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


MiddleIncome Economy Steady state


*/ 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.


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




*/ 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.


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





*/ 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.


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



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



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



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


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


Composite Reform Programme B(Absolute deviations from baseline)


Time


Time

Time

Time

Time

Time

Time

Time


Untrained unemployment rate Specialised unemployment rate


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


Composite Reform Programme C(Absolute deviations from baseline)


Time


Time

Time

Time

Time

Time

Time

Time


Untrained unemployment rateSpecialised unemployment rate


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

Documents

Essays on Human Capital, Innovation, and Growth with