Sectoral Engines of Growth in Developing Asia: Stylized Facts and Implications

8/22/2019 Sectoral Engines of Growth in Developing Asia: Stylized Facts and Implications

1/55

Economics and REsEaRch dEpaRtmEnt

serl Ege f Grw

develg a:

sylze F

il

Jesus Felipe, Miguel Len-Ledesma,Matteo Lanzafame, and Gemma Estrada

November 2007

RD WoRking PaPER SERiES no. 107


2/55


3/55

ERD Wrking Paper N. 107

Sectoral engineSof growthin DevelopingaSia: StylizeD factSanD implicationS

JeSuS felipe, miguel len-leDeSma,

matteo lanzafame, anD gemma eStraDa

november 2007

Jesus Felipe is Principal Economist in the Central and West Asia Department, Asian Development Bank; Miguel

Len-Ledesma is Reader, Department o Economics, in the University o Kent (Canterbury, U.K.); Matteo Lanzaameis Lecturer, Dipartimento di Economia, Statistica, Matematica e Sociologia V Pareto, Universita degli Studi di

Messina; and Gemma Estrada is Economics Ofcer in the Macroeconomics and Finance Research Division, Asian

Development Bank. This paper represents the views o the authors and does not represent those o the institutionsor countries they represent.


4/55

Asian Development Bank

6 ADB Avenue, Mandaluyong City1550 Metro Manila, Philippines

www.adb.org/economics

2007 by Asian Development BankNovember 2007

ISSN 1655-5252

The views expressed in this paperare those o the author(s) and do not

necessarily reect the views or policieso the Asian Development Bank.


5/55

FoREWoRD

The ERD Working Paper Series is a orum or ongoing and recently completedresearch and policy studies undertaken in the Asian Development Bank or onits behal. The Series is a quick-disseminating, inormal publication meant tostimulate discussion and elicit eedback. Papers published under this Seriescould subsequently be revised or publication as articles in proessional journalsor chapters in books.


6/55


7/55

CoNtENts

Abstract vii

I. INTRODUCTION 1

II. STRUCTURAL TRANSFORMATION IN DEVELOPING ASIA 2

III. STRUCTURAL CHANGE, INDUSTRIALIZATION, AND KALDORS LAWS 7

A. Kaldors Laws 8

IV. AN EXAMINATION OF KALDORS LAWS 11

A. Kaldors First Law 11B. Kaldors Second Law 13C. Kaldors Third Law 15

V. PRODUCTION STRUCTURE SIMILARITIES, TECHNOLOGY DIFFUSION,AND CATCH UP 19

VI. CONCLUSIONS 22

REFERENCES 31


8/55


9/55

AbstRACt

This paper provides an analysis o developing Asias growth experience romthe point o view o its structural transormation during the last three decades.The most salient eature o this transormation has been the signifcant decreasein the share o agriculture and the parallel increase in the share o services.The analysis uses Kaldors ramework to discuss whether industry plays the roleo engine o growth in developing Asia. The empirical results show frst, thatboth industry and services play such a role; and second, there is evidence o

endogenous, growth-induced technological progress. Likewise, the technology gapapproach supports the view that technological spillovers have ostered growthin developing Asia.


10/55


11/55

I. INtRoDUCtIoN

Except or those countries well endowed with natural resources such as oil, growth is alwayslinked to the structural transormation o the economy. Indeed, the growth experience o thedeveloped economies since the 19th century reveals that growth was associated to changes in thestructure o the economy. More recently, the experience o the successul Asian economies (Republico Korea [henceorth Korea]; Malaysia; Taipei,China etc.) also shows that high growth has beenassociated with deep changes in the structure o these economies. Moreover, many economists seethe development o a modern industrial sector as the key or propelling structural transormation.

Structural transormation is reected in changes in output and employment compositions.

An economy that grows as a result o transormation generates new activities characterized byhigher productivity and increasing returns to scale. The transition across dierent patterns oproduction and specialization also involves upgrading to higher value-added activities within eachsector through the introduction o new products and processes. These changes entail ar-reachingtransormations in terms o, among other things, economic geography and skill content o output.It is the countries that can sustain multiple transitions across dierent stages o their structuraltransormation that grow successully.

As Rodrik (2006) reminds us, development economists o the old school understood thekey role that structural transormation plays in the course o development. Among these, it wasprobably Nicholas Kaldor (1966 and 1967) who provided the most thorough explanation o whyindustry plays the role o engine o growth. Indeed, the so-called Kaldors Laws provide a solid

starting point or sector analyses o growth and structural change.The purpose o this paper is to analyze developing Asias growth experience in the context o

structural transormation. Growth and structural transormation are interrelated, since countries donot grow by simply reproducing themselves on a larger scale. Generally (unless all sectors o theeconomy grow at identical rates), countries become dierent as they grow, not only in terms owhatthey produce, but also in terms ohowthey do it (i.e., by using dierent inputs, includingmethods o production).

Specifcally, we attempt to answer the ollowing questions: (i) What has been the extent ostructural change in developing Asia during the last three decades? (ii) What is the contribution othe dierent sectors to the growth perormance o the Asian economies? (iii) What is the contributiono structural change to productivity growth and catching up?

The rest o the paper is structured as ollows. Section II documents the extent o structuraltransormation in developing Asia. Section III provides a brie summary o the literature on Kaldorslaws. Section IV discusses the empirical evidence provided by the laws. Section V complements theanalysis o growth and structural transormation in Asia through Kaldors laws with an analysis othe importance o structure and technology diusion. Section VI summarizes the main fndings.


12/55

November 2007

Sectoral eNgiNeSofgrowthiNDevelopiNgaSia: StylizeD factSaND implicatioNSJeSuSfelipe, miguel leN-leDeSma, matteo laNzafame, aND gemma eStraDa

II. stRUCtURAL tRANsFoRMAtIoN IN DEVELoPING AsIA

During the last three decades, most countries in developing Asia have undergone massivestructural change, in particular in terms o changes in both output and employment sectoral shares.Figures 1, 2, and 3 show scatterplots o the output and employment shares o agriculture, industry,and services vis--vis income per capita, pooling data since 1970 or the whole world. Figure 1shows that the shares o agricultural output and employment decline as countries become richer.Figure 2 shows that as countries income per capita increases, so do the shares o output andemployment in industry, although there seems to be a point beyond which these two shares startdeclining. Figure 2 also shows a wide dispersion in these shares or a given income per capita.Finally, Figure 3 shows that the shares o output and employment in services clearly increase withincrease in income per capita. The relationships described between sectors shares and income percapita is reerred to as logistic pattern. It is based on Engles law (demand explanation) and onthe dierential productivity growth rates across sectors (supply explanation).

figure 1agricultural outputanD employmentShareSverSuS per capita gDp, all economieS

(logarithmicScale), 19702004

Agricultural Output Share Agricultural Employment Share

100

50

20

5

100

50

20

5

Agr

iculture

(perce

ntofGDP)

Agr

iculture

(percento

femployment)

GDP per Capita, Constant US$2000 (in log scale)

Developing Asia Rest of the World



100 600 2000 10000 60000 100 600 2000 10000 60000

Sources: Asian Development Bank, Statistical Database System, downloaded 14 September 2006; National Bureau o Statistics (variousyears), China Statistical Yearbook; Sundrum (1997) and Chadha and Sahu (2002), cited in Anant et al. (2006); World Bank, WorldDevelopment Indicators online database, downloaded 4 August 2006. Data or Taipei,China were downloaded rom http://eng.stat.gov.tw/public/Data/782617221171.xls and http://eng.dgbas.gov.tw/public/data/dgbas03/bs2/yearbook_eng/y025I.pd on2 October 2006.

Appendix Tables 1, 2, and 3 show output and employment shares o the three sectors othe economy as well as o the manuacturing subsector by decade or developing Asia. The share


13/55

SectioN iiStructural traNSformatioNiNDevelopiNgaSia

erD workiNgpaper SerieSNo. 107

o agricultural output in total output has declined signifcantly in all regions during the last 30years. Especially signifcant are the declines that occurred in the Peoples Republic o China (PRC)and India. Parallel to this decline, there has been an increase in the share o services also in allregions. The share o industry has increased signifcantly in some parts o developing Asia (e.g.,

ASEAN-4,1

Other Southeast Asia, Other South Asia); remained the same in the PRC; and increasedby a small margin in India.

The share o employment in agriculture has also declined across the region, except in Centraland West Asia (although in 20002004, agriculture was still the largest employer in developing Asiain 12 out o 23 countries or which data was available). This is the result o the convulsion thatthe region underwent ater the collapse o the Soviet Union. In general, the decline in agriculturalemployment has occurred at a much slower pace than that in output.

As in the case o the output share, there has been a generalized increase in the share oemployment in services in all regions. Employment in industry has increased signifcantly in theASEAN-4 countries (except the Philippines) and by a small margin in India; it has not changed inthe PRC; and has suered a decline in the newly industrialized economies or NIEs (especially Hong

Kong, China) and across most o Central and West Asia.

figure 2inDuStry outputanD employment ShareSverSuS per capita gDp, all economieS


Industrial Output Share Industrial Employment Share

Industry

(percentofGDP)





100

50

20

100

50

20

Industry

(percentofemploymen

t)

100 600 2000 10000 60000 100 600 2000 10000 60000

Sources: Asian Development Bank, Statistical Database System, downloaded 14 September 2006; National Bureau o Statistics (variousyears), China Statistical Yearbook; Sundrum (1997) and Chadha and Sahu (2002), cited in Anant et al. (2006); World Bank, WorldDevelopment Indicators online database, downloaded 4 August 2006. Data or Taipei,China were downloaded rom http://eng.stat.gov.tw/public/Data/782617221171.xls and http://eng.dgbas.gov.tw/public/data/dgbas03/bs2/yearbook_eng/y025I.pd on

1 The ASEAN-4 economies are Indonesia, Malaysia, Philippines, and Thailand.


14/55

November 2007


2 October 2006.

The NIEs have undergone severe deindustrialization as manuacturing has lost signifcantweight in total output between the 1970s and 2000-2004 (see Rowthorn and Ramaswamy [1997and 1999] and Pieper [2000]). This is not a negative phenomenon, but the natural consequence othe industrial dynamism o these economies. It is a eature o economic development that reectstheir success. In terms o manuacturing employment, all our NIEs have clearly deindustrialized,especially Hong Kong, China where the share decreased by about 25 percentage points in two decades.The declines in the other three economies are signifcant but smaller. These developments shouldnot be interpreted as ailure o these economies, but as the result o the natural and dynamicprocess o development, i.e., the transition to service-led economies. Rowthorn and Ramaswamy(1997 and 1997) have noted that this group o countries is going through a process similar to thato the countries o the Organisation or Economic Co-operation and Development (OECD), althoughit must be noted that it is a process aecting Taipei,China and, especially, Hong Kong, China andto a much lesser extent Korea and Singapore. This is the result o transerring production acilitiesto the PRC. In Korea and Singapore, the share o manuacturing has remained at about 27% sincethe 1980s.

figure 3ServiceS outputanD employmentShareSverSuS per capita gDp, all economieS


Services Output Share Services Employment Share

GDP per Capita (Constant US$2000) GDP per Capita (Constant US$2000)

100 600 2000 10000 60000 100 600 2000 10000 60000

100

50

20

100

20

5

Services

(percentofGD

P)

Services

(percentofemploy

ment)

Developing Asia Rest of the World Developing Asia Rest of the World

Sources: Asian Development Bank, Statistical Database System, downloaded 14 September 2006; National Bureau o Statistics (variousyears), China Statistical Yearbook; Sundrum (1997) and Chadha and Sahu (2002), cited in Anant et al. (2006); World Bank, WorldDevelopment Indicators online database, downloaded 4 August 2006. Data or Taipei,China were downloaded rom http://eng.stat.gov.tw/public/Data/782617221171.xls and http://eng.dgbas.gov.tw/public/data/dgbas03/bs2/yearbook_eng/y025I.pd on2 October 2006.


15/55

SectioN iiStructural traNSformatioNiNDevelopiNgaSia


The ASEAN-4 economies and other Southeast Asia have increased their manuacturing sharessignifcantly, both in terms o output and employment. The exception is the Philippines, which hadthe highest manuacturing output share among the ASEAN-4 in the 1970s, but by 20002004 theshare had decreased by about 3 percentage points and was the lowest in the group.

Although Indonesia, Malaysia, and Thailand are cases o what can be labeled as successulindustrialization, this must be qualifed with the ollowing two observations. First, other than Korea;Kyrgyz Republic; Malaysia; and Taipei,China, none o the other economies in Appendix Table 3 had in20002004 a share o employment in manuacturing as high as that o the OECD countries. Second,in terms o labor productivity (Figure 4), there is a large dierential between most developingAsian countries and the OECD average. Indeed, it appears that many countries across developingAsia have industrialized at low levels o productivity. This could be due to two reasons: (i) thatthe product mix o new employment has been toward relatively low-productivity industries; and/or(ii) that the increase in employment has taken place in low-productivity techniques.

Other than the NIEs, todays level o productivity in the rest o the developing Asia is stillbelow the OECD average during the early 1970s. The level o productivity in the secondary sector is

signifcantly higher than that in agriculture. And, the level o labor productivity in the service sectoris above that in industry and manuacturing. Labor productivity in industry in the Philippines in20002004 was below the 1978 level, and in Indonesia it has barely increased. In most countries,labor productivity in agriculture is still very low.

Only the NIEs have achieved labor productivity levels that approach those o the OECD countries,and within this group, Singapore and Hong Kong, China are city-states with very small rural sectors.Korea and Taipei,China are signifcantly behind. Moreover, although in all Asian countries productivityhas improved signifcantly (with the noted exception o the Philippines), the absolute gap withrespect to the OECD productivity level has widened. In the case o Malaysia, the country with thehighest productivity levels outside the OECD and the NIEs, the absolute productivity dierentialwith respect to the OECD in industry has almost doubled, rom $21,786 in 19801985 to $38,946

in 20002004, despite the act that Malaysias productivity in industry increased by a actor o 1.7.In other countries and sectors the gap has widened by even larger amounts.


16/55

November 2007


figure 4total labor proDuctivity, logarithmic Scale (conStant2000 uS DollarS)

100000

10000

1000

100

100000

10000

1000

100

100000

10000

1000

100

OECD

PRC

India

197075 7679 8085 8689 9095 9699 200004

Singapore

OECD

Taipei,ChinaHongkong, China

197075 7679 8085 8689 9095 9699 200004

100000

10000

1000

Rep. of Korea

OECD versus PRC and India

OECD versus ASEAN-4 OECD versus Other Asian Developing Countries

OECD versus NIEs

1 97 0 75 7 6 79 8 0 85 8 6 89 9 0 95 9 6 99 2 00 0 04 1 97 0 75 7 6 79 8 0 85 8 6 89 9 0 95 9 6 99 20 00 0 4

Indonesia

OECD

Philippines

Malaysia

Thailand Azerbaijan

OECD

Pakistan

Kyrgyz Rep.

Viet Nam

Note: The 19801985, 19861989, 19901995, and 20002004 data or India reer only to 1983, 1988, 1994, and 2000 fgures, respectively.Similarly, the 20002004 data or PRC, Indonesia, Kyrgyz Republic, and Pakistan reer only to 20002002; the 19861989 dataor Indonesia only to 1989; the 19761979 data or the Philippines only to 1978; and the 19701975 fgure or Pakistan only to19731975.


17/55

SectioN iiiStructural chaNge, iNDuStrializatioN, aND kalDorSlawS

erD workiNgpaper SerieSNo. 107 7

III. stRUCtURAL CHANGE, INDUstRIALIZAtIoN, AND KALDoRs LAWs

The evidence presented so ar clearly points toward a rapid process o structural transormationin developing Asia. In order to understand the potential role o this transormation, it is importantto view these changes in light o the development theory literature. It is in this context that theKaldorian sectoral growth acts or laws (Kaldor 1966 and 1967) become very relevant as an approachto the issue o how structural change has aected growth in developing Asia, and what is the rolethat the dierent sectors have played. The Kaldorian acts bring together the notion o engine ogrowth sectors, economies o scale, and sectoral shits in a simple yet inormative way. Thisramework recognizes that some sectors may play a more important role in pulling the rest o theeconomy and generating productivity gains through economies o scale.

Kaldors laws allow us to address empirically the ollowing questions: (i) Is manuacturingstill an engine o growth in Asia? (ii) Can services play a role as engine o growth? (iii) What arethe most dynamic sectors in Asian countries? (iv) Can we expect continued growth in Asia, giventhe recent sectoral developments? It should be noted that we view Kaldors laws more as a serieso stylized acts and historical regularities rather than a theory o economic development. These

acts are compatible with a diverse range o theories o growth. What is important is that thesecorrelations are presented at the sectoral level and, hence, are helpul in analyzing and comparingpatterns o economic growth and the role o structure. In this sense, our objective is estimating,rather than testing, these laws, ollowing the distinction put orward by Leamer and Levinsohn(1995).

The role attributed to manuacturing in the process o take-o and subsequent catch-up isusually a key element o sector studies o growth. It is no surprise, thereore, that economistsand policymakers worry about swings in manuacturing. Though economies like Australia, Canada,New Zealand, the Scandinavian countries, and others relied heavily on the primary sector or theirdevelopment, they all experienced periods o strong industrial growth and diversifcation as essentialcomponents o their sustained economic growth. Rodrik (2006) has argued that sustained growth

requires a dynamic industrial base. One can, thereore, speak o the logic o industrialization (Nixson1990, 313) and understand why many developing countries have adopted strategies toward rapidindustrialization, oten starting with industries that use relatively simple technologies and that havethe potential to be labor-intensive and thus absorb labor, such as textiles, clothing, and shoes. Theexperience o the industrial economies shows that establishing a broad and robust domestic industrialbase holds the key to successul development, and the reason that industrialization matters lies inthe potential or strong productivity and income growth o the sector. This potential is associatedalso with a strong investment drive in the sector, rapidly rising productivity, and a growing shareo the sector in total output and employment. The presence o scale economies associated withthe secondary sector, gains rom specialization and learning, as well as avorable global marketconditions imply that the creation o leading industrial subsectors, along with related technological

and social capabilities, remains a key policy challenge. Today, there is wide variety across countriesin terms o resource endowments, pace o capital accumulation, and policy choices. This impliesthat there is ample room or diversity in industrial development.


18/55

November 2007


Figure 5 shows the scatter plot o the annual growth rate o output vis--vis the absolutechange in the share o manuacturing in total output or the 1970s up to 20002004. The fguredocuments the positive correlation between both variables. Economies in the frst quadrant withthe highest increases in the manuacturing share and in the output growth rate are Cambodia,

Indonesia, Korea, Lao Peoples Democratic Republic (Lao PDR), Malaysia, and Thailand.figure 5

output growthverSuS changein manufacturing output Share, 1970S20002004

40

20

-20

-40

0

Absolute

Change,

Sharein

Manufacturing

Output

Growth Rate, Output (percent)

Developing Asia Rest of the Word

-5 0 5 10 15

Mftgshare=6.099 + 1.090 Yt-stat: (6.48) (4.32)R2: 0.10, No. of obs: 178

Note: Positive change in the share indicates that the share at the end o the period was higher.Source: Authors estimates.

Notwithstanding this observation, given the high growth rate that the service sector o anumber o Asian countries has achieved recently (see Appendix Table 4), and consequently theincreasing share o services in total output, one may wonder i industrialization is a step that maybe bypassed today.

A. Kaldr Law

Kaldors frst lawstates that the aster the rate o growth o manuacturing output, the asterthe rate o growth o gross domestic product (GDP), giving to manuacturing the role o engineo growth. The characteristics o manuacturing, and industry in general, as a sector with stronginputoutput linkages coners this sector this potential. This role is based not only on this aspect,but also on the act that capital accumulation and technical progress are strongest in the industrial


19/55

SectioN iiiStructural chaNge, iNDuStrializatioN, aND kalDorSlawS


sector, having important spillover eects on the rest o the economy. This means that the strongerthe rate o growth o manuacturing, the stronger the rate o growth o the rest o the economy.Kaldor viewed the high growth rates characteristics o middle-income countries as an attribute othe process o industrialization.

In his seminal work and or empirical purposes, Kaldor specifed the laws as relationshipsbetween growth rates because he estimated a cross section o countries with data at two pointsin time. Kaldors frst law, i.e., that manuacturing acts as the engine o growth, can be examinedthrough a regression o nonmanuacturing output growth ( Ynm ) on manuacturing output growth

( Ym ), was thereore specifed as

Y a a Y nm m= +1 2 (1)

where a2 indicates the strength and size o the impact (elasticity) o the manuacturing sectorsgrowth on the rest o the economy. This coefcient, thereore, can be viewed as the main indicatoro the engine o growth role o this sector. Similar regressions are estimated or agriculture,industry, and services to assess their capacity as engines o growth.

Kaldors second law states that there is a strong positive relationship between the growth omanuacturing production and the growth o manuacturing productivity. This law is also knownas Verdoorns Law and has been interpreted as evidence in support o the existence o increasingreturns in the manuacturing sector (see, or example, McCombie et al. 2002). The expansion ooutput leads to a process o macro-dynamic increasing returns that derive in productivity gains.This can also be interpreted rom the point o view o employment creation: sectors subject to scaleeconomies have lower employment elasticities with respect to output, as productivity grows as aby-product o output expansion.2 As productivity growth equals output growth minus employmentgrowth, regressing productivity on total growth could induce spurious correlation. For this reason,Verdoorns law, i.e., the induced productivity growth eect linked to increasing returns, is specifedas a regression o manuacturing employment growth ( em ) on manuacturing output growth ( Ym ).

Algebraically,

e b b Y m m= +1 2 (2)

Kaldors hypothesis is that output expansion induces a less than proportional employmentexpansion that leads to productivity gains. The coefcient b2 , the elasticity o employment withrespect to output, is an indicator o the degree o increasing returns. The closer to 1, the smallerthe induced productivity growth and returns to scale. Traditionally, estimates o the coefcientor manuacturing are close to 0.5. With a ew assumptions about the capitaloutput ratio, a 0.5coefcient implies increasing returns in a standard production unction (see Ros 2000, 1303).The interpretation o this coefcient is that each additional percentage point in the growth ooutput is associated with a 0.5% increase in employment and a 0.5% increase in the growth o

productivity. As in the case o the frst law, similar regressions are estimated or agriculture, industry,and services.

As mentioned earlier, rather than interpreting the Kaldorian model o growth as a theoreticalexplanation o the ultimate causes o growth, these hypotheses are ormulated empirically through

2 It is also the result o a second mechanism, wherein employment growth in industry tends to increase the rate o growtho productivity in other sectors. This is the consequence o diminishing returns to labor in other sectors, absorption

o surplus labor rom these sectors, as well as o aster increase in the ow o goods into consumption.


20/55

10 November 2007


regressions (1) and (2) and interpreted as stylized acts that can shed light on the questions posedabove. Thus, these two hypotheses provide a set o growth acts at the sectoral level that can beused in conjunction with several theoretical interpretations to ormulate a well-inormed analysiso the prospective growth perormance o the Asian countries. Kaldors laws, when viewed as a set

o empirical regularities, appear to be consistent with many growth models that do not rely ondiminishing returns to capital. The division o labor and ideas-driven growth models o Romer (1986and 1990), Lucas (1988), and Aghion and Howitt (1992) are all consistent with Kaldors secondlaw, although they are set up in economies without an explicit sectoral structure.

Kaldors third lawstates that when manuacturing grows, the rest o the sectors (not subjectto increasing returns) will transer labor to manuacturing, raising the overall productivity o theeconomy. Dynamic sectors absorb workers rom the stagnant ones in which the level and growtho labor productivity is very low. This raises the overall productivity level o the economy and itsrate o growth. The key mechanisms that explain how structural change aects productivity growththrough compositional eects were developed by Baumol et al. (1985 and 1989). According to theirview, backward economies with a large pool o employment in low-productivity activities (normally

agriculture) experience a bonus rom structural change. This structural bonus arises as a resulto the transer o labor rom low- to high-productivity activities. This will automatically increasethe productivity level o the economy. This happens even i this transer o resources constitutesmainly a shit rom agriculture to services. However, as the logistic pattern o structural changedrives resources toward services, and given that productivity growth in this sector is usually slowerthan in industry, countries experience a structural burden. This burden means that the processo structural change has a negative impact on productivity growth. In the limit, as most o thelabor orce has moved into the services activities, economies experience asymptotic stagnancyas productivity growth is mostly determined by the services sector.3

The relationship between Kaldors third law and Baumols asymptotic stagnancy theory is evident.The importance o a sector depends not only on its role in generating scale economies, but also onhow it absorbs resources rom other sectors, leading to structural bonus and structural burdeneects. Although a sector with low productivity growth can absorb resources rom agriculture leadingto increased productivity levels, this source o economic growth is asymptotically exhausted. Inthe transition process, Kaldors third law will be an important source o growth but, in the limit,induced productivity growth is the key to generating growth (see, or example, Fagerberg 2000,Timmer and Szirmai 2000).

3 This description o how resource transers in the process o structural change aect growth is very useul to analyzeThis description o how resource transers in the process o structural change aect growth is very useul to analyzecompositional eects. However, three aspects have to be noted. First, the concept o asymptotic stagnancy is a

relative one. That is, growth is driven by activities whose productivity grows at a relatively slower rate than industry,

but productivity growth may still be high in absolute terms. Second, it is assumed that services are necessarily aslow productivity growth sector. However, the distinction between stagnant and dynamic sectors has become blurred

in recent decades by technological advances that have provoked very important changes in the organization and

productivity o many services activities. Finally, although it is almost tautological that employment shits toward themore labor-intensive activities, the model does not consider that growth in the dierent sectors is interdependent.

That is, the expansion o some activities, especially those with increasing returns, can have an important impact on

productivity in other activities. New growth theory has emphasized how the expansion o markets leads to increaseddivision o labor and intermediate products leading to more sophisticated production processes that can be enjoyed

by all the sectors in the economy. Similarly, some activities with more traditional inputoutput linkages can act as

engines o growth through backward and orward eects on other sectors. Innovation and knowledge accumulationare but another source o sectoral spillovers that link together the developments o dierent sectors independently o

their relative size in the economy.


21/55

SectioN ivaNexamiNatioNofkalDorSlawS


IV. AN EXAMINAtIoN oF KALDoRs LAWs

Regressions o Kaldors frst two laws were conducted using a panel o 17 developing Asiancountries or 19802004. The panel is unbalanced as data or some countries or some years aremissing at the beginning o the sample. The lack o consistent time-series data on employment orBangladesh, India, and Lao PDR prevented us rom including these countries in the regression othe second law. The exploitation o both cross-sectional and time series data allows us to includeall relevant inormation that would be thrown away in pure cross-sectional average estimates. Themodels are estimated in log levels using cointegration techniques, as the traditional growth ratesspecifcation may be simply capturing business cycle correlations that are not the ocus o theinvestigation.

This panel o time-series allows us to address other potentially relevant problems. The frst oneis the bias that might be associated rom the endogeneity o the regressors. Although our interestis in the stylized act stemming rom the reduced orm, and not in a structural interpretation,endogeneity may induce biases in the estimated coefcients. For this reason we use a ully modifedordinary least squares (FMOLS) panel cointegration estimator, as advocated by Pedroni (2000). This

is an estimator or heterogeneous panels that allows us to obtain a panel estimate o the coefcientas well as country-specifc coefcients. A homogeneous cointegration vector is estimated, but fxedeects and short-run dynamics are allowed to be unit-specifc.

The second problem that may arise is that there may be a high degree o correlation betweenthe dierent variables across countries. World shocks aecting variables such as the terms otrade may induce cross-sectional correlation and also correlation between the regressed variables,unrelated to the Kaldorian hypotheses that are the ocus o the analysis. For this reason, we alsoprovide estimates o the panel cointegration coefcients including heterogeneous (country-specifc)unobserved components estimated by obtaining principal components.4 This estimate allows or ahigh degree o heterogeneity as well, but assumes common slope coefcients. We reer to theseestimates as UC (unobserved component) elasticities.

A. Kaldr Fir Law

Estimation results o the frst law are summarized in Figure 6, which shows the estimatedlong-run elasticities o output o the rest o the economy with respect to output in each one othe sectors.5 The sector with the largest engine o growth elasticity, ater controlling or commonshocks, is industry. This is ollowed by services and manuacturing. Agriculture appears to have avery large impact using the FMOLS estimate. The introduction o the unobserved component (UCelasticity) reduces the size o the elasticity signifcantly. This is because agricultural output is likelyto be highly correlated across countries due to common shocks stemming rom, or example, climateconditions and terms o trade shocks. The larger elasticity o industry relative to manuacturing

reects the act that industrial activities such as electricity and other utilities have importantorward and backward linkages with the rest o the economy.

These results also indicate that both industry and services have acted as the engines o growthin developing Asia during the period analyzed. It is important to note that services have a largerimpact than manuacturing. This is not due to mere compositional eects, as we have avoided

4 See Forni et al. (2001). We included only the frst principal component in the model.See Forni et al. (2001). We included only the frst principal component in the model.ee Forni et al. (2001). We included only the frst principal component in the model.5 Cointegration tests showed that all variables in this specifcation were cointegrated.


22/55

1 November 2007


this source o spuriousness by using the output o the rest o the sectors, and not total output,as the dependent variable.

figure 6engineof growth effectS (elaSticitieS)

Services

Manufacturing

Industry

Agriculture

UC Elasticity

FMOLS Elasticity

0 0.5 1.0 1.5 2.0

Individual economy results are shown in Table 1. The strongest elasticities o the service sectorcan be ound in Thailand, Korea, Cambodia, Bangladesh, and Lao PDR, respectively; whereas ormanuacturing, Taipei,China and the Philippines present the largest eects. Cambodia and Bangladeshhave signifcantly larger elasticities or overall industry than or manuacturing. This may reect

the importance o utilities or construction in these countries. The results or Hong Kong, China arerather surprising, with a low elasticity or services and a negative (though insignifcant) elasticityor manuacturing. The reason or these results is that the nonservices sector in Hong Kong, Chinais only a very small raction o output (15% in 2003), making the results not reliable. The FMOLSresults or agriculture are biased as in this sector common shocks account or a large raction othe variation o output.


23/55



table 1engineof growth elaSticitieSby economy

ServiceS manufacturing inDuStry agriculture

Bangladesh 1.03 0.65 1.06 1.80Cambodia 1.05 0.28 2.55 2.83

PRC 0.90 0.67 0.57 2.58

Hong Kong, China 0.20 -0.75 0.3 0.71

India 0.60 0.84 0.79 2.28

Indonesia 1.00 0.51 0.87 2.02

Lao PDR 1.03 0.44 0.4 1.78

Malaysia 0.91 0.57 0.84 3.79

Myanmar 0.75 0.86 0.46 0.42

Pakistan 0.89 0.85 0.87 1.29

Philippines 0.55 1.18 0.77 1.88

Singapore 0.89 0.96 0.83 1.41

Korea 1.03 0.66 0.83 5.51

Sri Lanka 0.74 0.58 0.92 2.88

Thailand 1.16 0.61 0.84 2.65

Viet Nam 0.93 0.75 0.98 2.09

Taipei,China 0.46 1.67 0.44 0.28

Note: Authors FMOLS estimates.

b. Kaldr secnd Law

Figure 7 provides the panel employment elasticities or the dierent sectors and or the totaleconomy. These provide the basis or the analysis o the second law. 6 The results show that thesector with the lowest employment elasticities (highest degree o induced productivity growth) ismanuacturing with an elasticity o 0.5, in line with the traditional estimates o this eect. Thisis ollowed by industry and services. Although in the case o services the UC estimation producedunreliable results,7 the estimated elasticity o 0.68 indicates that there is a substantial degree oincreasing returns in this sector, as it is statistically dierent rom one. Agriculture shows the highestemployment elasticity, and the UC estimate is not signifcantly dierent rom one. The act thatthe total employment elasticity or all the sectors is the lowest reects the act that intersectoraltransers o labor between sectors also play an important role in inducing productivity growth.

6 The Verdoorn eect equations are all cointegrated except in the case o agriculture, where we could not reject the null

o no panel-cointegration.7 The frst principal component accounted or less than 25% o the total variance, indicating only small cross-sectional

correlation. For this reason we preer to rely on the FMOLS estimation or this sector.


24/55

1 November 2007


figure 7employmentelaSticitieSwith reSpectto output: the verDoorn effect

Services

Manufacturing

Industry

Agriculture

Total

UC Elasticity

FMOLS Elasticity

0 0.2 0.4 0.6 0.8

Table 2 shows employment elasticities. The PRCs employment elasticity o industry is verylow. This is ollowed by the elasticities or Singapore; Taipei,China; and Korea. The latter are themost mature economies in the sample. In the case o services, Taipei,China; Singapore; Myanmar;Indonesia; and Hong Kong, China have the lowest employment elasticities. With the exceptions o

Myanmar and Indonesia, these economies are also more advanced. It appears that economies that areeither growing very rapidly or are already more mature are the most capable o generating inducedproductivity growth in both sectors, and also enjoy increasing returns. This sectoral dynamism isimportant to avoid the middle income trap.


25/55



table 2employmentelaSticitieSby economy

total inDuStry agriculture ServiceS manufacturing

Cambodia 0.86 1.39 1.15 1.42 1.20

PRC 0.22 0.10 n.a. 0.75 n.a.

Hong Kong, China 0.29 1.30 1.26 0.6 1.17

Indonesia 0.41 0.98 0.31 0.48 0.47

Malaysia 0.48 0.64 n.a. 0.58 0.55

Myanmar 0.53 0.57 0.34 0.45 0.40

Pakistan 0.41 0.28 0.68 0.70 0.20

Philippines 0.84 1.01 0.50 1.05 0.81

Singapore 0.40 0.19 0.96 0.48 0.12

Korea 0.32 0.26 n.a. 0.68 0.14

Sri Lanka 0.37 0.37 0.76 0.68 0.28

Thailand 0.35 0.59 n.a. 0.67 0.53

Viet Nam 0.34 0.35 0.60 0.68 n.a.

Taipei,China 0.25 0.21 0.82 0.41 0.14

n.a. means not available.Source: Authors FMOLS estimates.

C. Kaldr third Law

Discussions o Kaldors third law have been carried out in regression context with little success(see McCombie 1980).8 Traditional estimates suer rom spurious correlation and identifcation

problems. A more useul approach consists in decomposing the growth rate o labor productivity( q ) into three components:

( ) ( )( )

q

q q q

t

i ti i

I

i

ti i ti i

II

i=

+ =

0 01

3

0 0

== = +

1

3

0 0

1

3

0

( )q q

q

ti i i

III

i

(3)

where q denotes the level o labor productivity and denotes the sectors employment share in totalemployment (both variables in periods O and t; ireers to the three sectors, agriculture, industry,and services). What is the interpretation o the three terms in the decomposition?

(i) The frst term (I) is the static structural reallocation eect (SSRE). This is the contributionto productivity levels o the transer o resources rom low- to high-productivity sectors.It is related to Baumols structural bonus hypothesis. We expect that the transer will

8 Recently, Wells and Thirlwall (2003) have estimated Kaldors laws with data or Arican countries. The third law is a

regression o overall productivity growth on the growth o industry and the growth o employment outside industry. Thecoefcient o industrys growth is positive while that o employment growth outside industry is negative, indicating

that the slower employment growth outside industry, the aster overall productivity grows.


26/55

1 November 2007


increase the average level o productivity o the economy as employment shares shitrom agriculture to services. This eect is calculated by shiting employment shares,keeping initial productivity levels o each sector constant.

(ii) The second term second term (II) is the dynamic structural reallocation eect (DSRE), and

represents the contribution o the resource transer to productivity growth. It is relatedto Baumols structural burden hypothesis as employment transers toward servicesasector with (in general, though not always) lower productivity growthreduce the overallproductivity growth o the economy. The eect is calculated as the interaction betweenemployment shits and productivity growth.

(iii) The fnal term (III) is the within-sector productivity growth (WS). It is the contributiono productivity growth within each sector to overall productivity growth. This is thegrowth o productivity that is not explained by sectoral shits. It is calculated by keepingemployment shares constant and allowing productivity levels to change.

The importance o making the distinction between the static and dynamic structural reallocationeects is that it helps distinguish between the structural bonus and burden eects o employmentreallocation. Countries with large agricultural sectors have a lot to gain rom the bonus o surpluslabor in low-productivity activities. However, i the growth o employment is predominantly insectors with lower scope or productivity growth, there is a burden eect. I productivity growthin services is lower than in manuacturing, this imposes a relative burden (though not absolute).Note, however, that in Kaldors interpretation, this reallocation is induced by growth o the leadingsector. This hypothesis cannot be examined directly by shit-share analysis, but it is clear that itis the growing sectors that will draw resources rom those contracting.

The contribution o sectoral shits to productivity growth embedded in the third law is presentedin Figures 8, 9, and 10. We have decomposed productivity growth into its three components orthree dierent subperiods, 19801989, 19901999, and 20002004. The average contributiono each eect throughout 19802004 is approximately 33% or the static structural reallocationeect (SSRE), 14% or the dynamic structural reallocation eect (DSRE), and 81% or within-sector productivity growth eect (WS). For 19801989, these percentages are 15.6%, 22%, and106.5%, respectively. For 19901999, they are 40%, 6% and 66%, respectively; whereas or thefnal period the fgures are 44%, 13%, and 69%, respectively. This shows that the SSRE gainedimportance during the last 15 years. The fgures, however, mask large dierences across countries.Nevertheless, they point toward the WS eect as the main driver o overall productivity growth.Within-sector productivity growth is related to productivity gains stemming rom scale economiesand, importantly, technology absorption rom rontier economies such as Europe, Japan, and UnitedStates (US).9 The impact o the SSRE is non-negligible, accounting or one third o productivitygrowth. This eect is mainly the result o the transer o labor rom agriculture into services. Asexpected, the DSRE is negative and related to the structural burden hypothesis.

9 Technology adoption can be thought o as a unction o explicit research and development investment, human capital,

oreign direct investment, and also structural composition o output in the sector.


27/55



figure 8proDuctivity growth DecompoSition: percent contributionof each effect, 19801989

Thailand

Taipei,China

Korea, Rep. of

Singapore

Philippines

Pakistan

Malaysia

Indonesia

Hong Kong, China

-200 -150 -100 -150 0 50 100 150 200Within

Dynamic Structural Reallocation Effect

Static Structural Reallocation Effect


-150 0 50 100 150 200

Viet Nam

Thailand

Taipei,China

Sri Lanka

Korea, Rep. of

Singapore

Philippines

Pakistan

Myanmar

Malaysia

Indonesia

Hong Kong, China

PRC

Within




28/55

1 November 2007



-50-100 0 50100 150 200

Viet Nam

Thailand

Taipei,China

Sri Lanka

Korea, Rep. of

Singapore

Philippines

Pakistan

Myanmar

Malaysia

Hong Kong, China

PRC

250

Within



During 19801989, WS productivity growth drove productivity growth. The exception wasIndonesia, where the negative impact o DSRE is very large. SSRE is small or most economies exceptIndonesia and Thailand. During 19901999, SSRE acquires a more relevant role, driving productivity

growth in countries such as Indonesia, Philippines, and Thailand. In 20002004, SSRE was the maindriver o productivity growth in Sri Lanka, Viet Nam, Pakistan, and Thailand. This eect is usuallylarger or economies with large agricultural surplus labor and high growth. It is worth noting that,in the case o the PRC, WS productivity growth has been the main driver o total productivity growthduring the last 15 years despite having a large share o employment in agriculture. For economieslike Hong Kong, China; Singapore; and Korea, WS productivity growth has been also the main drivero productivity growth, as one would expect given their levels o economic development and smallershares o agricultural employment. In general, DSRE is small, reecting the act that services sectorproductivity growth has not lagged too ar behind that o industry.

The main conclusions o this section allow us to answer the main questions posed at the starto Section III:

(i) Is manuacturing still an engine o growth in developing Asia? Manuacturing, andespecially the wider industrial sector, appears to act as an engine o growth as it drivesgrowth in other sectors and is subject to strong economies o scale. That is, there isstill a traditional role or industry as an engine o growth. However, the industrial sectordoes not appear to drive employment transers rom agriculture, which is an importantsource o productivity gains.


29/55

SectioNvproDuctioNStructureSimilaritieS, techNologyDiffuSioN, aND catch up


(ii) Can services play a role as engine o growth? The evidence shows that services havea strong and large impact on the growth o the other sectors. Indeed, this impact islarger than industrys. Although to a lesser extent than in industry, services appear tohave signifcant productivity growth-inducing eects through the exploitation o scale

economies. Services also appear to be driving productivity gains through actor reallocationeects.

(iii) What are the most dynamic sectors in developing Asian countries? Although there areimportant dierences across countries, both industry and services can be thought oas the dynamic sectors o Asian economies. The evidence points toward a key role orindustry but, very importantly, services appear to have been able to play this dynamic roleas well. The old distinction between industry and services as the dynamic and stagnantsectors o an economy, respectively, does not appear to hold true in the context o theAsian countries.

(iv) Can we expect continued growth in Asia, given the recent developments at the sectorlevel? The scope or growth is still very large. This is because productivity growth is

likely to continue in many Asian economies through two sources. First, actor reallocationtoward services, especially or the middle-low income countries in the sample, is notlikely to be exhausted as a source o growth in the short run. Second, WS productivitygrowth through catching-up and exploitation o scale economies is likely to continuebeing the main driver o productivity gains in the uture.

Overall, this implies that there is signiicant evidence o endogenous, growth-inducedtechnological progress in developing Asia.

V. PRoDUCtIoN stRUCtURE sIMILARItIEs, tECHNoLoGY DIFFUsIoN, AND CAtCH UP

In this section we address the third question posed in the introduction, namely, what is the

contribution o structural change to productivity growth and catching up? While regression analysiso the frst two Kaldorian hypotheses or the Asian countries has provided signifcant evidence oendogenous, growth-induced technological progress, or countries lagging behind the technologicalrontier, endogenous technological progress will be partly dependent on the acquisition and masteringo more advanced production techniques rom the leader countries, which in turn will be determinedby such actors as national research and development, human capital, and trade openness.

Furthermore, i technology is (at least to a certain extent) sector-specifc, its diusion romthe most advanced to the less advanced countries will be more intense and aster the higher thedegree o structural (or sectoral) similarity between them. As a result, ceteris paribus, technologicalprogress will be aster or a less advanced country, the more its production structure resembles thato the technological leader. This reasoning is in line with Abramovitz (1986 and 1993), who has

argued that the extent to which developing economies can beneft rom the superior technologydeveloped in advanced countries depends on their absorption capability. The latter is itsel acomposite variable, determined by social as well as economic and structural actors, such as thedegree o technological congruence with countries on the technological rontier.

Here we propose a simple approach to measuring the signifcance o the extent to which theproductivity growth perormance o the Asian countries benefted rom technological spillovers romthe most advanced countries owing via a structural channel.


30/55

0 November 2007


First, in the spirit o the technology gap approach to growth and convergence (Gerschenkron1962, Nelson and Wright 1992), we defne a measure o the potential or technology transer romthe most advanced to the less advanced countries as given by the labor productivity ratio between

the two, i.e.,

GAP t q t

q t

iL

i

( )( )

( )

=

, where tdenotes time, q tL( ) is the level o labor productivity in thetechnologically most advanced country, and q ti( ) its counterpart in the less advanced country i.

Second, we devise a measure o structural similarity making use o Krugmans specializationindex (or K-index) developed by Midelart-Knarvirk et al. (2000).10 At each point in time, the indexis constructed as the sum over the ksectors o the absolute dierences between the sectors shareso value-added in country iand in the technological leader. Its value ranges between zero and twoand increases with the degree o specialization, i.e., it is higher the more a countrys productionstructure diers rom that o the technological leader. For instance, a K-index value o 0.5 indicatesthat 25% o the country is production structure is out o line with that o the technologicallymost advanced country, in the sense that one quarter o its total output does not correspond tothe average sectoral composition in the latter.11

This way, one can build a structurally weighted gap variable by frst designing a measure o

structural weights as W tK t

iL

iL( ) = ( )

12

, where 0 1 ( ) W tiL

, which increases with the degree o structural

similarity, i.e., as K tiL

( ) alls. The structurally weighted gap variable is WGAP t W GAP t i iL i ( ) ( ).= Thisvariable can then be introduced in a growth regression (see Temple 1999) to capture the idea thatthe impact o technology spillovers on the less advanced countries growth perormance will bedependent not only on the size o the technology gap but also on the degree o structural similaritybetween technological leaders and ollowers. We examine this hypothesis by making use o a simplereduced-orm growth equation.

Given the nature o the hypothesis under examination, fnely sectorally disaggregated data areessential or estimation proposes. Taking this into account, we restrict our attention to manuacturing

and construct the structural weights W tiL ( ) using data or 28 sectors rom the United Nations IndustrialDevelopment Organization.12 The remaining data are taken rom the World Bank World Development

Indicators (WDI) and the International Labour Organisation (ILO).13

10 When applied to country-level bilateral comparisons it is constructed as: K t abs v t v t iL ik

L

k

k( ) ( ) - ( )= , where

v tx t

x ti

k ik

ik

k

( )( )

( )

and x tik( ) denotes country is value-added in sector kat time tand reers to the technological leader.Instead o value-added, Midelart-Knarvirk et al. (2000) employed the gross value o output as a measure o activity

level, on the grounds that this makes the results o the analysis less likely to be biased by the eects o structuralshits in outsourcing to other sectors. This option was precluded by data unavailability in our case.11 The upper bound o the index equals two because, by construction, it takes into account both positive and negativeThe upper bound o the index equals two because, by construction, it takes into account both positive and negative

deviations across sectors. Thus, when calculating the implied-percentage deviation the value in question must behalved: in the example, (0.5/2)%=25%.

12 The data are rom the Industrial Statistics Database 2006 at the 3-digit level o ISIC Code (Revision 2) (UNIDO

2006).13 The source o the manuacturing value added series or the US is the Department o Commerce, Bureau o Economic

Analysis.


31/55

SectioNvproDuctioNStructureSimilaritieS, techNologyDiffuSioN, aND catch up


To smooth out cyclical eects, structural weights W tiL

( ) were computed as 3-year movingaverages o annual values. The regression was estimated by means o panel data techniques usingan unbalanced panel o annual data over 19822002 or nine Asian economies, namely Bangladesh;PRC; Hong Kong, China; Indonesia; Malaysia; Singapore; Korea; Sri Lanka; and Taipei,China. The

regression estimated is:

q GAP SWGAP it i i it j

j it

j

jj

= + +==

1

2

1

2

(4)

where qit is the rate o labor productivity growth in country iand the gap variables are constructedtaking both the US and Japan as the two technological leaders with respect to the less advancedAsian economies in our sample. We used the fxed-eects least squares dummy variables estimatorand, since we are dealing with annual data and a airly short time-series, we allow or just onelag. Results are reported in Table 3.14

table 3technology gap regreSSion

variableS coefficient StanDarD error t-valueGAP_JP 0.008 0.005 1.85

GAP_JP(-1) 0.050* 0.020 2.54

SWGAP_JP 0.050** 0.017 2.95

SWGAP_JP(-1) 0.050 0.037 1.31

GAP_US 0.016 0.008 1.95

GAP_US(-1) 0.064** 0.020 3.16

SWGAP_US 0.070** 0.025 2.77

SWGAP_US(-1) 0.074** 0.025 2.92

* and ** indicate signifcance at the 5% and 1% level, respectively.Note: JP and _US denote that the leader country is Japan or the United States, respectively; (-1) indicates the frst lags.

The R-squared o the regression is 0.49, suggesting that the gap and structurally weightedgap variables explain roughly hal o the variation in labor productivity growth. GAP_JP is notstatistically signifcant; however, it becomes highly signifcant and takes on the expected positive

sign when it is interacted with W tiL

( ) to take account o structural similarities, i.e., in SWGAP_JP.The coefcient o GAP_JP(-1) is also signifcant and positive so that, overall, one can read theresults in Table 1 as supporting the view that Japan, as the technological leader country in theregion, plays a signifcant role as a source o technological spillovers to the other Asian economiesin our sample.

When the technological gap is measured with respect to the US, results are dierent. Boththe structurally weighted gap variable SWGAP_US and its frst lag SWGAP_US(-1) turn out to besignifcant at the 1% level, but the ormer takes on a negative sign so that the overall impact othe two on labor productivity growth is negligible. Furthermore, the frst lag o the GAP_US variableis also negative and signifcant.

14 The introduction o more lags o the independent variables did not change the results qualitatively. Moreover, estimationo a dynamic version o the growth regression using the Generalized Method o Moments resulted in an insignifcant

coefcient on the lagged dependent variable, leading us to avor the static-version results reported in Table 3.


32/55

November 2007


VI. CoNCLUsIoNs

The most salient eature o developing Asias transormation during the last three decades hasbeen the signifcant decrease in the share o agriculture and the parallel increase in the share oservices. Some parts o developing Asia have clearly industrialized in the sense that the shares oindustry and manuacturing in total output have increased (e.g., Indonesia, Malaysia, Thailand).But many other countries in the region have not seen an increase in these shares. The richesteconomies in the region, the NIEs, are undergoing a deindustrialization process. This simply reectstheir shit to high value-added services. Various other countries in the region have had difcultiesin industrializing. India and the Philippines are among the most signifcant examples, althoughor India, recent data seem to indicate that its manuacturing share has increased. Others (e.g.,the Pacifc economies) ace industrialization as a very difcult process, since they have limitedopportunities to start with. It is important to note that the patterns o structural transormationo output and employment are dierent, as the decline in agricultural employment is taking placeat a much slower pace than that o output. This has led, in many countries across the region, torather asymmetric output and employment structures. Indeed, one could say that much o the

region looks like a service economy in terms o output, but like an agricultural economy in termso employment. An additional important eature o structural change in developing Asia is that,despite its rapid growth, the level o labor productivity in most o the region still lags ar behindthat o the industrial countries. Given that investments were made in highly productive industryand services segments o the economy, this implies that there are still many other large segmentso the economy with very low productivity. Structural transormation in developing Asia is takingplace through a combination o modern and sophisticated industry and services with high and risingproductivity levels, with many other backward ones (probably where a large part o the labor romagriculture is being transerred) that operate at very low productivity levels. Finally, the analysiso structural transormation rom the point o view o technology and scale indicates that only aew countries in the region have undergone a signifcant upgrade.

Regressions o Kaldors laws indicate that both industry and services appear to have acted asengines o growth in the Asian economies. The manuacturing sector is subject to strong increasingreturns, although in services the degree o increasing returns is indeed non-negligible too. Althoughthe share o industry and manuacturing is shrinking in total employment, this does not necessarilyimply that their role as the most dynamic sectors has decreased. Induced productivity growth inmanuacturing can indeed be seen as the reason or its decline as a share o output or countriesthat had previously industrialized. Notable exceptions are the PRC and India. In the ormer,industrial activity remains relatively very important and in the latter large-scale industrializationhas not occurred. Services appear to have contributed largely to growth as they drag employmentrom the less productive agricultural sector. Although induced productivity growth in services issmaller than in industry, services appear to be a remarkably dynamic sector. Both actors togetherhave contributed to the importance o services as an engine o growth. As the large reserves o

employment in agriculture are exhausted, the contribution o services to productivity growth islikely to decrease as its productivity growth is lower than that o industry.

This will largely depend on the composition o services between dynamic and stagnant activities.However, there is no reason to believe that, in the medium run, growth will decline due to theincrease in the share o services in total output. There are three main reasons or this:


33/55


(i) There is still a very large scope or structural change, especially in the less developedeconomies o Asia.

(ii) The role o the dynamic industrial sector remains very relevant or economies whereindustrialization occurred previously.

(iii) The scope o within-sector productivity growth is still very large. This is likely to beacilitated by structural change itsel, which increases the capacity o Asian economiesto absorb oreign technology. This catching-up process is likely to lead to importantproductivity gains in services.

Finally, the technology gap approach, as ormalized in the ramework used here, provides asimple way to analyze the impact o technology diusion on the growth perormance o the Asiancountries. The results support the view that technological spillovers oster growth when Japan istaken as the technological leader, but is not the case when the leader-country is the US. Moreover,structural similarity seems to be playing a signifcant part in the process o technology diusionboth rom Japan and the US, although the overall inuence rom the latter is airly small.

SectioNvicoNcluSioNS


34/55


35/55


economy

outputShareS

employmentShareS

170

10

10

00000

10

10

00000

A

I

s

A

I

s

A

I

s

A

I

s

A

I

s

A

I

s

A

I

s

Mongolia

17.3

6

27.3

0

55.3

4

33.7

3

24.9

0

41.3

7

23.1

2

24.3

6

52.5

2

47.0

9

21.8

4

31.0

7

45.9

2

14.4

3

39.6

4

Tajikistan

32.0

6

40.1

1

27.8

3

31.1

9

36.4

3

32.3

7

25.9

0

37.2

9

36.8

1

57.7

6

19.9

6

24.7

7

Turkmenistan

27.7

7

37.0

6

35.1

7

23.4

9

45.5

0

31.0

1

23.2

3

44.4

5

32.3

3

44.7

7

Uzbekistan

29.3

4

35.5

4

35.1

3

32.7

9

30.1

1

37.1

0

33.3

7

23.2

9

43.3

4

42.5

0

20.9

4

36.5

6

34.4

0

20.3

0

45.3

0

PacifcIland

FijiIslands

24.72

21.8

8

53.4

0

20.3

8

20.9

7

58.6

6

19.05

25.6

8

55.2

7

16.0

2

26.2

7

57.7

1

2.2

2

35.3

9

62.3

5

Kiribati

19.06

52.2

8

28.6

6

29.6

6

7.8

1

62.5

3

21.29

6.3

8

72.3

2

15.4

6

10.7

5

73.7

9

9.6

1

13.8

6

76.5

3

MarshallIslands

14.45

14.5

9

70.9

6

10.9

3

20.8

2

68.2

5

Micronesia

44.5

4

PapuaNew

Guinea

32.45

28.0

0

39.5

5

32.2

1

27.4

8

40.3

2

28.75

37.0

2

34.2

2

27.9

4

43.3

9

28.6

8

73.3

3

3.6

5

23.0

2

Palau

11.14

11.0

6

77.8

0

3.0

7

13.7

9

83.1

4

Samoa

19.63

26.9

4

53.4

3

14.5

2

26.6

1

58.8

6

SolomonIslands

26.6

8

13.6

5

59.1

5

Timor-Leste

43.15

16.7

7

40.0

8

27.2

0

16.2

1

56.6

0

Tonga

44.65

12.2

9

43.0

6

37.9

1

14.2

0

47.8

9

33.63

14.7

9

51.5

8

28.1

4

15.8

0

56.0

6

37.1

3

Vanuatu

21.95

6.0

5

72.0

0

22.7

0

9.3

0

68.0

0

16.84

10.4

0

72.7

6

14.1

8

8.8

9

76.9

2

meansdatanotavailable.

A

meansagriculture,

Imeansindustry,

Smeansservices.

Note:Themaindatasour

ceistheWorldDevelopmentIndicators(WorldBankonlinedatabase,

downloaded4August2006).

Agricultureincludesagriculture,

fshery,andorestry.

Industryincludesmanuacturing,mining,constructionandutilitie

s.

Servicesincludetransport,

trade,

fnance,p

ublicadministration,andothers.

Sectorshare

sarecomputedin

nominalterms.

Itis

importanttoaddthatwecheckedthequality

othedataothisdatabaseandoundsome

problems.

Wecalculatedthenumberocases

wherethechange

betweentwoconsec

utiveyearsintheshare(bothoutputandem

ployment,oreachsector)waslarger(i.e.,

increaseintheshare)orsmaller(i.e.,

decreaseintheshare)than

5percentagepoints

.Exceptinextremecircumstances,e.g.,warso

rnaturaldisasters,sectorsharescannotchangebythismuchbetween2years.

Wediscoveredthattherewere

plentyosuchcasesoroutput,andsubstantiallyeweroremployment.

Appendixtable1continued

appeNDix


36/55

November 2007


appenDix table 2manufacturing output ShareS, by DecaDe

economy 170 10 10 00000

PRC 37.27 36.26 32.90h 34.50l

India 15.32 16.43 16.58 15.71

NIE

Hong Kong, China 21.18 9.43 4.32

Korea 21.61 27.51 27.14 27.82

Singapore 24.84a 26.09 26.11 27.39

Taipei,China 32.43 34.95 27.11 22.80

AsEAN-

Indonesia 10.42 15.35 23.72 29.04

Malaysia 16.82 20.42 27.05 31.21

Philippines 25.72 25.03 23.29 22.94

Thailand 18.98 23.32 29.55 34.00

oher suhea

Asia

Cambodia 11.08 19.40

Lao PDR 9.27d 14.20 18.67

Myanmar 9.64 9.07 6.90 8.49m

Viet Nam 19.69e 15.23 19.94

oher suh Aia

Bangladesh 13.76 14.87 15.73

Bhutan 5.29 10.39 7.79m

Maldives

Nepal 4.11 5.24 8.77 8.85Pakistan 15.89 15.98 16.44 15.99

Sri Lanka 19.02 15.39 15.68 15.90

Cenral Aia and Mnglia

Armenia 27.56 22.68

Azerbaijan 14.08 7.87

Kazakhstan 13.30i 16.33

Kyrgyz 20.04 16.19

Mongolia 31.04 18.70 6.37

Tajikistan 27.70e 25.43 32.35

Turkmenistan 26.30j 15.47m

Uzbekistan 25.06 11.96j 9.40

Pacifc Iland

Fiji 11.79 10.59 14.44 15.02n

continued next page.


37/55


economy 170 10 10 00000

Kiribati 1.62b 1.16 0.98 0.89n

Marshall Islands 1.63 4.54o

Micronesia 0.40g

Papua New Guinea 7.26 10.06 8.89 8.50m

Palau 0.97 1.19n

Samoa 17.10j 15.37

Timor-Leste 2.78k 3.29

Tonga 6.63a 5.42 4.85 4.61

Vanuatu 3.90c 4.45 4.88 4.21o

means data not availablea Reers to 19751979 average.b Reers to 19781979 average.c Reers to 1979.d Reers to 1989.e Reers to 19851989 average. Reers to 19871989 average.g Reers to 1983.h Reers to 19901992 average.i Reers to 19921999 average.

j Reers to 19941999 average.k Reers to 1999.l Reers to 2000.m Reers to 20002003 average.n Reers to 20002002 average.o Reers to 20002001 average.Sources o basic data: World Development Indicators (World Bank onlinedatabase, downloaded 4 August 2006); Directorate General o

Budget, Accounting, and Statistics, Taipei,China (various years).

appenDix table 2. cninued.

appeNDix


38/55

November 2007


appenDix table 3manufacturing employmentShareS, by DecaDe

manufacturingaS percentof total employment

10 10 00000

PRC 15.11b 13.47 11.16g

India 11.05a 10.92a 11.22a

NIE

Hong Kong, China 35.89 19.02 10.20h

Singapore 27.91 24.53 18.31i

Korea 23.93 23.40 19.44

Taipei,China 33.41 28.66 27.40

AsEAN-

Indonesia 9.68c 11.73 13.15g

Malaysia 15.95 22.59 21.94

Philippines 9.93 10.06 9.65Thailand 8.87 12.33 14.58

oher Ecnmie

Azerbaijan 9.36e 5.44g

Kyrgyztan 20.11 19.19

Pakistan 13.66 10.99 12.66

Viet Nam 8.32 10.33

oECD 21.58d 19.20d 16.89d

a For India, the fgure or each decade reers only to a single year, as ollows: 1983, 1993/1994, 1999/2000.b Reers to the period 19871989.c Reers to the average or the years 1980, 1982, 1985, and 1989.

d For OECD, the number o countries covered each decade are: 18 or the 1980s; 20 or the 1990s; 21 or 2000/2004.e Reers to the period 19921999. Reers to the period 19961999.g Reers to the period 20002002.h Reers to the period 20002001.i Reers to the period 20012003.Sources: ILO (2006); Directorate-General o Budget, Accounting, and Statistics, Taipei,China (various years); Sundaram and Tendulkar

(2006).


39/55


appenDix

table4

averagetotalg

Dpgrowth

anD

Sectoralgrowth

by

DecaDe

economy

170

10

10

000

0

total

agri-

culture

inDuStry

ServiceS

total

agri-

culture

inDuStry

ServiceS

total

agri-

cult

ure

inDuStry

ServiceS

total

agri-

culturei

nDuStry

ServiceS

PRC

5

.90

2.3

9

8.2

2

5.7

9

9.2

3

5.1

4

10.0

5

11.9

3

9.5

3

4.2

3

12.1

0

8.9

1

8.7

6

3.2

8

9.7

7

9.3

6

India

2

.52

0.3

1

3.8

4

4.4

5

5.7

1

4.1

8

6.5

7

6.2

8

5.5

3

2.9

5

5.6

0

7.3

4

5.5

6

1.6

8

6.0

9

7.3

0

NIE

HongKong,

China

8

.73

6.9

7

3.6

6

4.6

2

.

.

.

Korea

7

.95

3.6

9

13.5

7

6.3

2

7.3

2

1.5

8

8.9

8

7.2

3

5.9

6

0.9

6

6.4

1

5.7

6

5.2

3

0.0

9

6.9

5

4.1

7

Singapore

8

.40

2.7

5

9.1

3

8.3

8

7.1

8

3.7

4

6.1

7

7.7

5

7.3

2

3.4

9

8.1

3

7.1

4

3.9

8

1.4

4

2.8

2

4.6

6

ASEAN-4

Indonesia

7

.48

4.1

5

9.8

0

8.4

1

6.1

4

4.0

1

7.0

8

6.6

2

4.5

1

2.0

6

5.6

7

4.5

3

4.5

4

3.4

1

4.0

6

5.5

0

Malaysia

7

.56

4.9

6

8.2

9

8.6

3

5.6

5

3.4

6

6.4

4

6.0

1

6.8

8

0.1

1

8.0

6

7.7

4

5.0

8

3.6

7

5.5

2

4.8

6

Philippines

5

.81

3.9

3

7.8

9

5.0

2

1.8

9

1.5

0

0.5

8

3.4

1

2.7

1

1.4

5

2.4

0

3.6

3

4.2

4

2.9

1

3.6

5

5.1

9

Thailand

6

.80

4.0

6

9.4

0

6.6

3

7.0

2

4.3

9

8.6

7

6.8

2

4.9

4

0.4

2

6.9

1

4.5

0

4.9

7

3.3

9

6.1

2

4.1

3

ohersuhea

Aia

Cambodia

6.5

7

1.5

9

1

5.9

2

5.0

6

LaoPDR

3.8

6

6.1

8

4.8

6

11.2

4

5.4

4

5.7

8

3.5

6

9.8

2

5.8

9

VietNam

4.3

6

7.1

4

3.8

2

9.9

6

7.3

9

6.8

9

3.7

1

9.5

9

6.1

0

ohersuhAia

Bangladesh

0

.82

0.3

1

1.2

6

2.1

5

3.1

7

1.5

8

4.4

1

3.7

2

4.6

9

3.3

2

6.7

5

4.1

5

5.2

8

3.5

5

6.8

3

5.3

5

Bhutan

7.1

3

6.2

1

3.3

3

8.2

8

7.2

5

6.2

4

.

.

.

Maldives

6.4

8

.

.

.

Nepal

2

.52

0.7

2

3.9

5

3.4

7

7.3

0

3.4

6

4.7

5

2.4

2

6.9

7

5.9

7

3.4

3

3.7

3

2.3

3

3.1

6

Pakistan

4

.04

1.8

4

5.4

0

6.0

5

6.6

2

4.2

4

7.8

6

6.5

2

3.8

8

4.0

5

4.5

6

4.4

3

4.0

5

2.0

4

4.6

7

4.5

7

SriLanka

4

.15

2.3

6

3.3

9

5.0

0

4.0

4

2.2

5

4.1

1

5.1

2

5.1

2

2.4

9

6.5

4

5.4

3

3.8

8

0.3

5

3.2

7

5.4

1

continuednextpage.

appeNDix


40/55

0 November 2007


economy

170

10

10

000

0

total

agri-

culture

inDuStry

ServiceS

total

agri-

culture

inDuStry

ServiceS

total

agri-

cult

ure

inDuStry

ServiceS

total

agri-

culturei

nDuStry

ServiceS

CenralandWe

Aia

Armenia

9.4

0

3.6

5

1

0.1

9

12.2

3

Azerbaijan

10.0

5

7.6

6

1

0.4

8

8.7

1

Kazakhstan

9.8

3

3.5

8

1

1.7

9

9.5

1

Kiribati

2

.60

6.2

3

7.1

5

9.0

6

3.8

0

5.1

5

1.5

2

8.4

4

5.3

7

1.7

3

Kyrgyz

6.1

8

7.1

8

10.1

4

5.3

8

4.0

5

0.7

0

11.1

3

2.0

8

4.7

7

3.9

2

4.0

3

6.2

8

Mongolia

1.3

8

2.6

4

0.1

4

1.2

9

4.2

7

-6.0

0

7.4

2

6.3

9

Tajikistan

10.5

1

7.9

8

10.2

3

4.6

0

9.2

9

Turkmenistan

4.3

1

3.9

2

5.0

6

4.0

0

.

Uzbekistan

0.3

7

1.0

5

3.0

7

1.3

6

4.5

5

5.8

4

3.2

4

4.5

1

Vanuatu

1.3

0

0.5

8

1.2

6

1.6

6

3.7

2

1.9

7

0.9

3

4.7

5

0.0

8

PacifcIland

Fiji

5

.52

3.3

5

3.9

3

7.5

0

0.6

5

2.0

7

0.1

9

0.9

9

2.8

4

0.0

4

4.1

2

3.4

3

2.2

2

Marshall_

Islands

0.6

8

1.3

5

Micronesia

1.0

6

2.1

0

PapuaNew

Guinea

3

.03

2.7

1

1.3

5

2.3

3

4.4

8

3.2

4

7.7

0

1.5

4

0.1

7

Palau

1.8

1

Samoa

-0.3

1

1.3

4

3.1

6

4.8

3

5.6

7

4.8

5

SolomonIslands

5

.29

5.6

5

3.1

4

3.2

9

Timor-Leste

3.3

7

1.6

0

-0.5

1

5.6

4

Tonga

0.4

8

2.1

8

2.7

4

2.0

4

1.7

7

3.0

2

2.2

5

3.1

6

TOTAL

5

.60

2.0

5

8.0

7

5.8

5

6.9

0

4.0

9

7.9

0

7.7

6

6.6

2

3.1

2

8.5

3

6.8

0

6.5

8

2.7

0

7.8

9

6.9

2

meansdatanotavailable.

Sourceobasicdata:WorldBank,

WorldDevelopmentIndicatorsonlined

atabase,

downloaded4August2006.

appenDix

table4.continued


41/55


REFERENCEs

Abramovitz, M. 1986. Catching-Up, Forging Ahead and Falling Behind.Journal o Economic History46:386

406.. 1993. The Origins o the Post-War Catch-Up and Convergence Boom. In J. Fagerberg, B. Verspagen,

and N. von Tunzelmann, eds., The Dynamics o Technology, Trade and Growth. Aldershot: Edward Elgar.ADB. 2007.Asian Development Outlook 2007. Asian Development Bank, Manila.

Aghion, P., and P. Howitt. 1992. A Model o Growth through Creative Destruction. Econometrica 60:32351.

Anant, T. C. A., R. Hasan, P. Mohapatra, R. Nagaraj, and S.K. Sasikumar. 2006. Labor Markets in India: Issuesand Perspectives. In J. Felipe and R. Hasan, eds., Labor Markets in Asia: Issues and Perspectives. London:

Palgrave Macmillan or the Asian Development Bank.Banister, J. 2005. Manuacturing Employment in China. Monthly Labor Review(July):11-29

Baumol, W. J., S. A. Batey Blackman, and E. Wol. 1985. Unbalanced Growth Revisited: Asymptotic Stagnancyand New Evidence. The American Economic Review75:4(September):80617.

. 1989. Productivity and American Leadership: The Long View. Cambridge: MIT Press.Fagerberg, J. 2000. Technological Progress, Structural Change and Productivity Growth: A Comparative Study.

Structural Change and Economic Dynamics 11:393411.Forni M., M. Hallin, M. Lippi, and L. Reichlin. 2001. The Generalized Factor Model: Identifcation and

Estimation. The Review o Economics and Statistics 82(3):54054.Gerschekron, A. 1962. Economic Backwardness i

Documents

Sectoral Engines of Growth in Developing Asia: Stylized Facts and Implications