Total Factor Productivity Growth, Structural Change ... - SSRN papers

Comparative Economic Studies, 2009, 51, (421–446) r 2009 ACES. All rights reserved. 0888-7233/09

www.palgrave-journals.com/ces/

Symposium Paper

Total Factor Productivity Growth, Structural Change and Convergence in the New Members of the European Union EL-HADJ M BAH1 & JOSEF C BRADA2,3 1

Department of Economics, University of Auckland, Private Bag 92019 Auckland, New Zealand. 2 Department of Economics, Arizona State University, Box 873806, Tempe, AZ, 852870-3806, USA. 3 Macedonian Academy of Sciences and Arts, Bul. Krste Misirkov 2, 1000 Skopje, Republic of Macedonia.

We estimate total factor productivity (TFP) growth in agriculture, industry and services in new European Union member countries and show how structural change contributes to growth. Because of the difficulties in measuring the capital stock of transition economies, we develop a model that estimates sectoral TFPs from data on sectoral employment and GDP per capita. Compared to Austria, new EU members have lower TFP levels, but their TFP growth is largely higher. Inter-sectoral movements of labour do not play a large role in aggregate TFP growth, and capital accumulation is an important component of convergence to EU levels of per capita GDP. Comparative Economic Studies (2009) 51, 421–446. doi:10.1057/ces.2009.8; published online 4 June 2009

Keywords: European Union, total factor productivity, structural change, economic convergence, economic growth, capital accumulation, productivity JEL Classifications: D24, E22, O11, O14, O41, O47

INTRODUCTION A key task for the transition economies that have recently joined the European Union (EU) is real convergence, the catching up with the per capita income levels of the older and more developed members. Although some observers have stressed that this process would require extensive investment

EM Bah & JC Brada TFP Growth and Structural Change

422

in physical and human capital (Blanchard, 1997, Buiter, 2000), the growth accounting literature suggests that these are not likely to be the decisive forces leading to convergence. This literature, from Solow (1957) to Prescott (1998) and Hall and Jones (1999), stresses that economic growth as well as inter-county differences in per capita income are largely because of changes, or differences, in total factor productivity, with the accumulation of physical and human capital playing only a subsidiary role.1 The EU’s own experts (European Union, 2006, p. 35) project that the new member countries’ per capita incomes will grow at about 4% per annum while those of the older and richer members will grow at 2%, leading to a convergence of per capita incomes in 2040. Moreover, the same estimates assume that over 50% of the growth in per capita incomes in both old and new members will be the result of total factor productivity growth rather than of factor accumulation, although, of course, total factor productivity (TFP) growth in the new member countries would thus be about twice as fast as in the old member countries. The ability of the new EU members to generate and sustain significant gains in TFP should not be taken for granted. The USSR and the countries of East Europe saw gains in TFP came to a virtual halt in the early 1980s, if not before then, a situation unprecedented among countries at such a level of development. The first to note the slowdown in TFP growth in the USSR was Kaplan (1968) who showed that, for any plausible parameter values of a Cobb-Douglas production function, Soviet TFP growth was falling towards zero.2 The results of similar research for East Europe by researchers in both the West and in the communist countries themselves came to more or less that same conclusion, namely that, by the start of the 1980s, the only sources 1 For example, Hall and Jones deconstruct the ‘the 35-fold difference in output per worker between the United States and Niger. Different capital intensities in the two countries contributed a factor of 1.5 to the income differences, while different levels of educational attainment contributed a factor of 3.1. The remaining difference – a factor of 7.7 – remains as the productivity residual’ (Hall and Jones, 1999, p. 83). Prescott (1998) reaches a similar conclusion without assuming a specific form of the production function. See also, Klenow and Rodriguez-Clare (1997), Hendricks (2002), Parente and Prescott (1994,2000) and Caselli (2005). 2 The literature on the decline in Soviet productivity growth took a unique direction because of Weitzman (1970), who attributed the output slowdown to a low elasticity of substitution between capital and labour, a finding that, of course, improved estimated TFP growth given the slow growth of the labour force and the rapid growth of the capital stock in the communist countries. Easterly and Fisher (1995) continued this insistence on the low elasticity of substitution explanation for Soviet growth retardation, but, in the face of the critique of their empirical work by Beare (2008), they were forced to accept the conclusion that ‘the rate of technical progress declined over the course of the history of the former Soviet Union’ (Easterly and Fisher, 2008, p. 147). Efforts to find similarly low elasticities of substitution between capital and labour in East Europe during the socialist era were generally not successful (see Brada, 1985) and evidence for the more recent period seems to confirm the general usefulness of the Cobb-Douglas specification (Gollin, 2002).

Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

423

of growth in East Europe were increases in capital and hours worked, with TFP growth non-existent or negative.3 It is possible that the forces that retarded TFP growth so severely during the late communist era have not been entirely eliminated by the process of transition. If the sources of TFP growth cannot be greatly influenced by short-term changes in policies, institutions or economic systems, then the transition economies would be consigned to being second-class members of the EU for a long time.4 Alternatively, if TFP does respond quickly to changes in policies, institutions and economic system, then the recent upsurge in TFP growth in the new EU members’ countries could be because of temporary factors, whose effects will soon wear off, a scenario developed in some detail by Van Ark (1999). Hall and Jones (1999) attribute high TFP levels to better institutions, but the new EU members have already undertaken many of the reforms needed to create functioning market economies and to meet the institutional and legal standards of EU membership. Thus, while some institutional improvements may still be possible, their pace, and thus that of TFP growth, may be much slower than it was in the past decade. In contrast to Hall and Jones, Frankel and Romer (1999) stress the role of openness to trade as the driver of TFP, but the new EU member countries have already undertaken as much opening up to international trade and investment as is likely to be feasible and further rapid growth of trade to GDP ratios does not seem likely. Finally, the transition economies, including the new EU members, may have been the beneficiaries of what we may call temporary Prescott–Granick effects that led to rapid, but short-lived, gains in TFP. Prescott (1998) and Parente and Prescott (2000) argue that TFP growth and levels are inversely related to the ability of incumbent workers to prevent the adoption new technologies, work rules and ways of organising production in order to protect the rents that they can earn using older technologies or ways of organising their work. In the case of the transition economies, the pattern of TFP growth can thus be understood in terms of David Granick’s (1989) description of the Soviet economy as a ‘job-rights economy’ in which workers had an explicit right to a particular job at a particular location. At the start of the Soviet experiment, unskilled workers were brought into factories from 3

See Brada (1989) for estimates of TFP and Ofer (1987) for a survey of the literature on this

topic. 4 The literature on the causes of TFP differentials is quite unsettled in this respect. Some authors, including Prescott (1998), take the view that a country’s TFP levels are subject to rather rapid change because of institutional and policy changes, while others, such as Hall and Jones (1999) and Frankel and Romer (1999) suggest that rather immutable factors such a legal origins, geography and so on are important determinants of TFP levels, which means that sustained rapid TFP growth in the East European countries would be difficult if not impossible.

Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

424

agriculture; they had no rents to preserve and no understanding of their job rights. As workers gained tenure at their places of work, they were also increasingly able to earn rents from operating the existing technology, and they thus had both the ability and the incentives to block the efficient introduction of new technologies and ways of working. As a result, new technology and ways of fully exploiting its productivity-enhancing characteristics could only be introduced into newly built and staffed factories but not into existing ones, thereby sharply reducing TFP growth. In the course of the transition, these job rights disappeared because open unemployment reduced workers’ bargaining power and because socialist-era laws providing these job rights were swept away, and the rents earned by the old industrial elite of the work force disappeared. Thus, in the Prescott– Grancik view, the current accelerated pace of TFP growth in the new EU member countries will continue only so long as workers continue to be unorganised and unable to exert pressure to slow changes in work rules and the fully effective introduction of new technologies. Because the work-place inflexibilities that the Prescott–Granick view considers important barriers to TFP growth are alleged to be the cause of slow productivity growth in the older EU member countries, fears that they will also spread to the new members are not unfounded. Consequently, the future pace of TFP growth in the new EU member countries is both uncertain and of great importance to their future well being. A second and related aspect of real convergence is structural convergence. The communist regimes in East Europe had followed a development strategy that favoured industry and agriculture at the expense of services. Thus, these countries entered the transition, and EU membership, with employment shares in industry and agriculture that were much larger than those found in market economies at similar levels of development and with service sectors that had much lower shares of employment than were to be found in comparable market economies (Gregory, 1970; Ofer, 1976). These disparities in employment carried over to the shares of these sectors in aggregate output as well. As these economies turned to the market to allocate resources, the service sector expanded dramatically while agriculture and industry lost employment share (European Union, 2006), although all of these economies continue to exhibit higher labour shares in agriculture and industry and lower shares in services than are to be found in the older EU member counties. Whether this structural difference is a legacy of communist policies or whether it simply reflects the fact that structural change in favour of services at the expense of agriculture and industry occurs with rising per capita incomes in nearly all market economies is unclear. In either case, structural change is more rapid in the new EU member countries so it remains Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

425

to be seen whether this faster and ongoing shift of resources between sectors is an important contributor to, or retardant of, aggregate TFP growth.5 In principle, it should be possible to undertake growth accounting exercises for the transition economies at the sectoral level, thus measuring TFP levels and their evolution over time, but in reality we face a fundamental problem in estimating the stock of capital. The transition from socialism to capitalism effectively destroyed a large, but unknown, part of the capital stock. Part of the destruction was physical; factories were abandoned and equipment was scrapped or thrown away. Another part of the destruction was what might be called ‘moral’, meaning that the huge changes in the structure of demand and the wholesale acquisition of new and more productive technologies from the West that occurred in the course of transition devalued, or accelerated the depreciation of, much of the communist-era capital stock (Campos and Coricelli, 2002). Studies of this phenomenon have produced estimates of surprisingly large declines in Russian and East European capital stock over the course of the transition. For example, Deliktas and Balcilar (2005) estimate that up to 50% of the communist-era capital stock was destroyed in the early transition.6 The various estimates of the excess destruction of capital stock differ in their magnitude as well as in the methodologies utilised to estimate the losses and in the assumptions driving the estimates. Moreover, given the logic of the argument for the destruction of capital, the amount destroyed in each country should depend, inter alia, on the communist-era openness of the country, on its industrial structure, and on the degree of its integration into the CMEA or Soviet economy.7 5

Stephan (2002) investigates the extent to which structural differences between the old and new EU members lead to differences in per capita incomes, but his analysis focuses on labour productivity differences rather than on differences in TFP. 6 See McKinsey Report (1999), Kushnirsky (2001) and Darvas and Simon (2000) for other, but similar, estimates. Izumov and Vahaly (2006, 2008) provide a survey of the issues and the literature on the transition-era capital stock as well as a methodologically consistent set of estimates of the ‘adjusted’ capital stocks of the Russian and former CIS economies. While we do not examine these economies in our paper, the gaps between official and adjusted estimates of the capital stock and their implications for TFP estimates shown by these studies are instructive. 7 A referee suggested that there was also destruction of communist-era human capital and that the value of East European human capital as measured by years of schooling may have been overstated. Thus, for example, Steffen and Stephan (2008) attribute much of the productivity differential between East and West to a human capital deficit, and Fo¨ldva´ri and Van Leeuwen (2009) claim that human capital accumulation had a positive effect on Hungarian GDP growth in the 1990s. Nevertheless, we refer the reader to the sources cited in Footnote 1 for compelling arguments why human capital accumulation is not likely to be a key driver of TFP growth. In this paper, we ignore any explicit accounting for human capital, and, in our results, inter-country differences in human capital are consequently reflected in differences in TFP. If we measure human capital by years of schooling, then the transition economies are only mildly behind the United States, which had 12.8 Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

426

Needless to say, wide divergences in these unofficial estimates of the capital stock lead to wide divergences in estimates of TFP growth and levels in the course of transition (Burda and Severgnini, 2008a). Absent plausible official estimates of sectoral and even aggregate capital stocks and the wide divergence in the unofficial estimates, as well as the lack of sectoral capital stock data, we propose to measure sectoral TFPs without recourse to capital stock data.8 Effectively, our approach substitutes readily available data on sectoral employment and aggregate GDP, the constraints on the interrelations between macroeconomic variables derived from a widely used model of economic growth and structural change, and model parameters obtained through calibration for generally unreliable or unavailable data on sectoral capital stocks in the transition economies. This substitution of easily obtainable data and a model and parameters that have proven their value in other applications seems to us to be a useful way to approach the questions that lie at the heart of this paper. A similar data limitation for developing countries has led researchers to develop indirect methods for estimating sectoral TFPs by making use of cross-section prices in a multi-sector growth model similar to the one we use to infer sectoral relative TFPs.9 In this paper we use a three-sector model developed by Bah (2008) to infer sectoral TFP time series for the new EU members. This kind of model also has been used by Rogerson (2008) to analyse labour market outcomes in Europe.10 The rest of the paper is organised as follows. Section ‘A three-sector model of structural transformation’ describes the model, characterises the competitive equilibrium and calibrates the model to the US economy. Section ‘Estimates of sectoral TFP in Austria and transition economics’ applies the model to Austria and to a sample of transition countries to demonstrate differences in sectoral TFP levels and their change relative to Austria, an EU member with a per capita output close to the (old member) EU average and with some similarities in size and location to a number of the transition economies. With a few exceptions, in all sectors Austrian TFP exceeds that of

years of schooling in 2000, while Poland, for example, had 12.0 years in 2001, and the Czech Republic 10.4 in 2001. 8 Using a different methodology, Burda and Severgnini (2008b). also choose to estimate aggregate TFP in the transition economies without recourse to official capital stock data. They show that attempting to construct capital stock data using perpetual inventory methods leads to highly unreliable estimates of both stocks and TFP. 9 See Herrendorf and Valentinyi (2006) and Hsieh and Klenow (2007). 10 The decomposition of the economy into three sectors, agriculture, services and industry, and the emphasis on the growth effects of reallocating labor and capital among these sectors also links our model to the work of Dowrick (1989) and Dowrick and Nguyen (1989) on growth in OECD countries. Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

427

the new member countries, but there are important sectoral differences in TFP between the new EU member countries and Austria. Moreover, not all members’ TFPs are progressing in all sectors in a way that promotes catch-up with Austrian per capita income. Although structural change does not appear to be a serious barrier to growth in the new EU member countries, some of them are found to rely heavily on input growth rather than TFP improvements for their GDP growth. The last section draws out some policy implications of our findings.

A THREE-SECTOR MODEL OF STRUCTURAL TRANSFORMATION Below, we present a model developed by Bah (2008). The model is a closed economy growth model with three sectors: agriculture, industry and services. The key features that drive labour reallocation across sectors are as follows. First, preferences are non-homothetic. As income rises, the household shifts its demand from agricultural goods to industrial goods and services. Because the household produces just enough of the agricultural good for subsistence, resources are shifted away from that sector as its productivity rises. Second, the elasticity of substitution between manufacturing and services and the TFP growth differential determine labour reallocation between those two sectors. The model thus generates a process of structural transformation that was first described by Kuznets (1966). The model Preferences and endowments There is a representative household who lives forever, and we normalise its size to 1 for simplicity. In each period the household is endowed with one unit of time, and it is also endowed with initial capital stock at time 0 and the total land for the economy which we normalise to 1. The household supplies labour inelastically to the three sectors. The instantaneous utility is given by:  UðFt ; At Þ ¼

At if  if logðFt Þ þ A

 At oA  At  A

ð1Þ

where At is the agricultural good and Ft is a composite consumption good defined as a CES aggregate of the industrial good (Mt) and the services (St).  e1 e e1 e1 Ft ¼ lMt e þ ð1  lÞSt e

ð2Þ Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

428

Lifetime utility is given by: 1 X

bt UðFt ; At Þ

ð3Þ

t¼0

where b is the discount factor. This specification of preferences implies that the economy specialises in ¯ is reached. Moreover, the economy agriculture until the subsistence level A ¯. will never produce more of the agricultural good than A Technologies All three sectors use Cobb-Douglas production functions. The inputs for agriculture are labour (N) and land (L) while industry and services use labour and capital. The agricultural good is only used for consumption so the resource constraint is given by: a 1a At ¼ Aat Nat Lt

ð4Þ

Aat ¼ Aa ð1 þ gat Þt

ð5Þ

where

The TFP parameters Aa, gat are assumed to be country specific.11 The industrial sector’s output is used for consumption (Mt) in the composite good and investment (Xt). The industry sector resource constraint is: y 1y Nmt Mt þ Xt ¼ Amt Kmt

ð6Þ

Amt ¼ Am ð1 þ gmt Þt

ð7Þ

where

The law of motion of the aggregate capital stock (Kt) in the economy is given by Ktþ1 ¼ ð1  dÞKt þ Xt

ð8Þ

where d is the depreciation rate.

11 Because of the way in which Aat is calculated, the beginning and starting values of agriculture’s share of employment predicted by the model are by construction the same as the actual values, although the interim values are not. Thus, in Table 4 we present only the predicted values of agriculture’s share because the actual values are the same.

Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

429

The output of the service sector is only used for consumption through the composite good. Therefore, the service sector resource constraint is given by St ¼ Ast Ksty Nst1y

ð9Þ

Ast ¼ As ð1 þ gst Þt

ð10Þ

where

In the equations above, the TFP parameters (Am, gmt , As and gst) are also assumed to be country specific. We may expect that a country’s institutions and policies affect the productivity in each of these economic sectors. Equilibrium Given that there are no distortions in this economy, the competitive equilibrium allocations can be obtained by solving a social planner’s problem. The economy  Once specialises in the production of agriculture as long as Aa ð1 þ ga Þt oA. t  the economy begins the production of industrial goods and Aa ð1 þ gat Þ  A, services. This corresponds to the start of structural transformation, and we will solve for the competitive equilibrium from this point on. We sketch the solution of the model and refer the reader to Bah (2008) for the details. Below, we present the key equations that determine the equilibrium allocations from the social planner’s problem. Labour in agriculture is given by:   1a A Nat ¼ ð11Þ Aat Let Nt ¼ 1Nat be the total time that can be allocated between the industry and service sectors. The aggregate capital stock is given by the following dynamic equation: "  y  y1 # Kt Kt Ktþ1 ¼Amt Nt þ ð1  dÞKt  b 1  d þ yAmt Nt Nt " # ð12Þ  y Kt1 Amt1 Nt1 þ ð1  dÞKt1  Kt Nt1 Once capital is known, the quantity of labour used in the service sector is given by: Nst ¼

 y

Amt NKtt

Ct

 l e  Ast 1e 1 þ 1l Amt

ð13Þ

Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

430

where Ct is the non-agriculture aggregate expenditure and is given by  y Kt Ct ¼ Amt Nt þ ð1  dÞKt  Ktþ1 Nt

ð14Þ

The other equilibrium allocations can be easily derived. Calibration to the US Economy The model is calibrated to match the US economy from 1950 to 2000.12 There are 13 parameters to calibrate. The productivity levels Ai(i ¼ a, m, s) are normalised to 1 in 1950. This corresponds to choosing units. Following the literature, labour’s share in agriculture (a) is set to 0.7 and capital’s share in industry and services (y) is set to 0.3. The TFP growth rates for industry and services (gm, gs)the discount rate b and d are jointly calibrated to match four averages in the data from 1950 to 2000: average growth rate of GDP per capita, average growth rate of the price of the service good relative to the industrial good, average investment-tooutput ratio and average capital-to-output ratio. The growth rate of agricultural TFP (gat) is chosen such that the model matches the agricultural shares of hours worked in the United States. We assume that the growth rate varies each decade starting in 1950. The agricultural subsistence level is equal to the agricultural production in every period after the start of structural transformation. Given that the agricultural TFP is normalised to 1 in 1950, the subsistence level can be easily computed using the shares of hours in agriculture. The last two parameters to calibrate are the elasticity of substitution between the industrial good and services (e) and the weight of the industrial good in the production of the composite good (l). These two parameters determine the labour reallocation between the industrial and service sectors. We choose values of e and l to minimise the quadratic norm of the difference between the predicted and actual industrial employment shares from 1950 to 2000. Table 1 summarises the calibrated parameter values. 12 For details, see Bah (2008). Because we will implicitly test the appropriateness of the US-derived parameters for our sample of countries later in the paper, we ask the reader to defer concerns about the validity of the United States as a benchmark. A compelling reason for using the United States for the calibration exercise is the longer and more stable time series on variables needed for calibration. Whether the United States is the most appropriate country for deriving parameters for our sample of countries cannot be answered in any definitive way, but we do note that we are able to replicate the time paths of sectoral employment and per capita income in all these countries using the US-based parameters. The data sources are explained in Appendix.

Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

431 Table 1: Calibrated Parameters Parameter

Aa

Am

As

A¯

a

b

d

e

gm

gs

l

y

Value

1

1

1

0.24

0.7

0.975

0.05

0.30

0.019

0.009

0.02

0.3

ESTIMATES OF SECTORAL TFP IN AUSTRIA AND SELECTED TRANSITION ECONOMIES We use the parameters derived from our calibration exercise and data on sectoral labour shares and GDP per capita to estimate the sectoral TFPs for Austria and for nine transition economies: Bulgaria, the Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, the Slovak Republic and Slovenia, that have joined the EU.13 Because of unavailability of consistent data on hours worked by sector, our analysis covers only the period 1995–2005. Although our model is one of sectoral change, which is a longterm phenomenon, as we shall see, the sectoral changes in employment in the transition economies were quite significant even over this shorter period. We use Austria as a standard for comparison because in terms of population and land area it falls within the range of the transition economies in our sample, and it also is close geographically to a number of them. Moreover, it can be seen as a typical old EU member country in terms of its per capita income (Table 2). Austria’s per capita income in 1997 and 2005 exceeded that of the transition countries by a palpable amount.14 Austria is also slightly above the average per capita GDP of the old EU member countries, but its position relative to other ‘old’ EU members changed very little between 1995 and 2005. The transition economies have gained appreciably in their standing vis a vis the EU 15 average, although the gains differ considerably across countries. In the application of the calibrated model to Austria and the transition countries, we assume all the parameters are the same across countries except the series for sectoral TFP. We use the model to find sectoral TFP series such that, within the framework of the model, we can best replicate the paths of 13 The sectoral employment shares for Romania proved somewhat problematic, and thus we dropped that country from our analysis even though it, too, is now an EU member. 14 The new member countries may have larger informal sectors, which would somewhat close the gap between their per capita incomes and that of Austria. However, official estimates of GDP in some of these countries adjust for the informal sector, and, in any event, Austria, has a non-trivial informal sector as well.

Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

432 Table 2: Per Capita Incomes as Percentage of EU-15 Average Country

1997

2005

Austria Czech Republic Estonia Latvia Lithuania Hungary Poland Slovak Republic Slovenia

112.9 61.9 35.0 29.8 33.3 45.5 40.1 42.3 64.5

113.3 67.8 51.7 43.1 47.1 57.2 46.0 50.1 75.0

Source: EU (2006)

GDP per capita and the sectoral employment shares. For agricultural TFP, we use the fact the subsistence level is assumed to be the same in every country. Therefore, for any other country, we can use the US employment shares and calculated agricultural TFP to deduce that country’s agricultural TFP. We calculate the agricultural TFPs in 1995, 2000 and 2005 and then assume constant growth between those dates. For the TFP series in industry and services and the initial capital stock, we match GDP per capita relative to the United States in 1995, the average GDP per capita growth and labour reallocation from industry between 1995 and 2005. Sectoral TFP in Austria Figure 1 shows the results of our simulation of Austrian per capita GDP growth, sectoral employment and sectoral TFP. The first panel shows per capita GDP, with 1995 normalised to one. The simulated and actual data for GDP per capita are close to each other as are the sectoral employment shares reported in Panel 2.15 The shares of agriculture and industry in employment have fallen while services employment’s share has increased. Overall, structural change in Austria has been relatively slow over the period analysed. The last panel in Figure 1 shows that Austrian total factor productivity in industry relative to US industrial TFP in 1950 was around 2.1 in 1995 and close to 2.4 in 2005. Agriculture and services TFPs in Austria in 15

That the model is able to match Austrian sectoral employment and per capita GDP over the sample period is a strong, but not absolute, verification of the validity of the US-based parameters for Austria. Because of space constraints we do not provide analogous Figures for the transition countries but summarize the results in tabular form. These Figures are available from the authors. Moreover, the model fits the actual GDP per capita and sectoral employment shares of the transition countries as well. If the model could not match the sample countries’ dynamics, it would mean that the parameters obtained from the US-based parameterization were inapplicable. Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

433

Figure 1: Actual and model predictions of per capita GDP, employment shares and sectoral TFP for Austria.

1995 were between 1.4 and 1.5 times the corresponding 1950 level in the United States. TFP growth in Austrian industry and agriculture was relatively high, but TFP growth in the Austrian services sector was very slow. The ratios of Austrian TFPs to US TFPs should not be taken as indications of the relative productivity in the three sectors of the Austrian economy. While Austrian industry’s TFP relative to the 1950 US level is higher than the ratio of Austria’s agricultural TFP to the US level, many studies of US productivity suggest that, in 1950, TFP in US agriculture was higher than was TFP in industry. Thus, Austria’s greater gains in industrial TFP vis a vis the United States may not have offset the 1950 advantage of US agricultural TFP over US industry’s TFP. As a result we are not able to infer from Figure 1 whether the expansion of one of Austria’s three sectors at the expense of the other two tends to raise or lower aggregate TFP growth. By observing the growth rates of TFP in the three sectors, we can, however, determine whether such inter-sectoral shifts in resources promote aggregate TFP growth by moving more resources into sectors that enjoy faster TFP growth over time. In the case of Austria, the movement of resources from agriculture and industry to services can be seen as a drag on growth in the sense that resources were moved from sectors with high rates of productivity growth to Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

434

one with low TFP growth.16 Given the slow pace of labour reallocation in Austria, the effect on aggregate growth is likely to be negligible.

Sectoral TFP in transition economies In Tables 3 and 4 we summarise the simulation results for the transition economies over the period 1995–2005. We first briefly discuss cross-country similarities and differences in the results and then discuss how sectoral TFP levels and trends influence the convergence of the transition economies to EU levels of per capita GDP. Next we provide an international comparison of sectoral TFPs to investigate whether the aggregate TFP lag implied by the transition economies’ lower per capita GDP is because of lower TFPs in all sectors of the economy or whether their lower aggregate TFPs are the result of particularly poor productivity in particular sectors of their economies. Table 3 shows that the model was able to generate GDP per capita growth rates that closely reflect actual GDP per capita growth in the transition countries. Table 4 shows the sectoral employment shares projected by the model for the transition economies as well as their actual employment shares. We thus note that the model is able to generate both per capita GDP growth rates and changes in sectoral employment that closely approximate the actual changes experienced by these countries. This, like the close tracking of Austrian growth and structural change, suggests the validity of our parameterisation on the basis of US data. All the transition countries underwent a similar change in structure that involved the movement of labour out of agriculture and industry and into services.17 In this sense, structural change in the new EU members mirrors that taking place in the old EU members, even if, looking at the current sectoral distribution of labour, the new members lag behind the older ones by having higher shares of employment in agriculture and industry and a lower share of services employment. On the other hand, the pace of structural change is faster in the transition countries than it is in Austria. There are significant differences among the nine transition economies in terms of the TFPs of their sectors relative to 1950 United States sectoral TFP levels. Table 5 ranks the sectors of the new EU member countries relative to 16

A referee noted that the low growth of productivity in services may be an artifact of problems in measuring the output of services and that the services sector may be a catalyst for productivity growth in the other two sectors. This, of course, would apply to the new members and to Austria as well. 17 A number of countries show a small reversal in that, at the end of our period of observation, industry slightly gains labor share at the expense of services. This may be related to large inflows of FDI and the emergence of East Europe as a sourcing point for manufactured goods exports to the EU. Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

435 Table 3: Actual and Predicted per capita GDP Growth 1995–2005 Country

Model

Actual

Austria Bulgaria Czech Republic Estonia Hungary Latvia Lithuania Poland Slovak Republic Slovenia

19.71 49.32 27.46 107.00 57.73 116.59 89.15 48.74 41.56 45.51

20.22 49.01 27.44 106.95 57.71 116.66 88.98 48.77 41.34 45.46

Note: The numbers are in percent.

Table 4: Actual and Predicted Sectoral Shares in Employment Predicted by the Model

Austria Bulgaria Czech Rep Estonia Hungary Latvia Lithuania Poland Slovak Rep Slovenia

Actual Data

Agriculture

Industry

Services

Industry

Services

1995

2005

1995

2005

1995

2005

1995

2005

1995

2005

7.25 25.14 6.37 10.16 8.38 19.32 23.48 21.77 9.49 11.19

5.08 23.73 4.00 5.28 4.96 12.44 14.71 17.40 4.64 8.84

30.68 30.53 40.79 33.58 34.61 27.69 27.93 32.67 39.55 40.91

30.34 28.15 38.42 33.59 34.62 26.88 28.27 29.71 38.00 36.32

62.07 44.33 52.84 56.26 57.01 52.99 48.59 45.56 50.96 47.90

64.58 48.12 57.58 61.13 60.42 60.68 57.02 52.89 57.36 54.84

31.57 33.31 41.64 33.48 33.13 26.81 27.83 32.41 39.14 42.15

27.87 24.68 39.03 33.41 33.16 26.31 28.26 28.52 38.40 36.04

61.04 41.59 51.89 56.35 58.39 53.75 48.69 45.78 51.36 46.52

66.87 51.58 56.95 61.31 61.87 61.10 57.01 54.07 56.81 54.17

Note: Actual and predicted labor shares for agriculture are the same for the beginning and ending year, although not for other years in the simulation.

US 1950 levels. The situation of industry is least ambiguous; it ranks as best or second best vis a vis the United States in all transition countries, and it is never the last sector in rank. For a sector that produces tradables and thus faces international competition, a sector receiving large amounts of FDI, and a sector where technology transfer by multinational firms is routine and relatively easy, such high rankings are not surprising. The rankings for the service sector are also quite consistent, but poor. Services is never the best sector relative to US 1950 TFP, and most often it is the sector that lags all others relative to US TFP. The long pre-transition neglect of services, the clear Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

436 Table 5: Rankings of Sectoral TFPs Relative to the US -1950 Country

Sector Ranking

Bulgaria Czech Republic Estonia Hungary Latvia Lithuania Poland Slovak Republic Slovenia

Industry>Services>Agriculture Agriculture>Industry>Services Agriculture>Industry>Services Agriculture>Industry>Services Industry>Agriculture>Services Industry>Services>Agriculture Industry>Services>Agriculture Agriculture>Industry>Services Industry>Services>Agriculture

shortages that existed in the provision of retail and other such ‘low productivity’ services after the fall of communism and the subsequent rapid expansion of those sectors, and slowness in developing a modern services sector help to account for this poor productivity picture. A weak internationalisation of the services sector may also be a factor. The relative position of agriculture is the most variable of the three sectors. In some countries it comes closest to US TFP levels, but in other counties it shows the biggest gap. This may reflect cross-country differences in the productivity of the agrarian sector because of variations in the effectiveness and extent of agricultural reforms, to the dissolution of collective agriculture, to the effectiveness of land distribution and reform and so on. Clearly, the poor productivity performance of services, the sector that shows the greatest employment gains, should be a policy concern for transition-economy governments. Structural change and aggregate TFP growth In this subsection we estimate the loss in GDP that results from the structural transformation process. This question is motivated by the fact that in all of our countries labour moves among sectors at a faster pace than it does in the old EU member countries. We note that the process of structural change is a key feature accompanying development. To determine whether structural change, either by moving labour from low to high TFP sectors or vice versa has an important impact on aggregate growth, we use the model to compute GDP per capita using the capital stock and TFP series estimated in the foregoing section. However, instead of using the corresponding labour shares from the model, we use the sectoral employment shares given by the data for 1995.18 Table 6 summarises the loss 18 It is important to note that, in the model, labour reallocation across sectors results from differences in sectoral TFP growth rates and the preference specifications.

Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

437 Table 6: Loss of GDP Per Capita Due to Structural Transformation (as a percnt of 1995 GDP Per Capita) Country Austria Bulgaria Czech Republic Estonia Hungary Latvia Lithuania Poland Slovak Republic Slovenia

Percentage Loss 1.28 0.95 2.29 3.83 2.31 4.47 6.55 2.74 4.44 3.16

of GDP per capita growth from 1995. The largest loss in potential per capita GDP was in Lithuania, whose per capita GDP in 2005 was 6.55% of 1995 per capita GDP less that it would have been with no structural change. This means that annual growth of per capita GDP was around a half a percent slower than it would have been with no structural change. This is not a trivial amount, even when judged against the almost doubling of per capita GDP between 1995 and 2005, but for the other countries the effect of structural change on growth is negligible. Consequently, we can conclude that past and future structural change, even at the accelerated pace seen in the transition economies over the past decade, has a relatively minor impact on the new EU members’ ability to catch up with the older EU countries in terms of per capita GDP. Sectoral TFP in comparative perspective Our results show that aggregate TFPs have risen in all of the transition economies over the sample period, 1995–2005. However, per capita GDP convergence between the transition economies and the older EU member countries will require the sectoral TFPs of the transition economies to grow closer to the levels of the old member countries. To the extent that some transition economies’ sectors lag behind in TFP growth, their other sectors will have to achieve even faster TFP growth to assure convergence. Figure 2 shows the agricultural TFPs of the transition economies relative to Austria’s agricultural TFP, which is normalised to one in each year. One transition economy, the Czech Republic, has higher TFP in agriculture than does Austria for the entire sample period and its agricultural TFP also grew faster than did Austria’s. As a result, by the end of our sample period, Czech TFP in agriculture was nearly 20% higher than Austria’s. This is not a surprising result because, while the two countries share a similar continental Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

438

Figure 2: Agricultural TFP Relative to Austria.

climate and grow similar crops using similar technologies, the quality of Czech land is higher because of a more favourable topography. Moreover, the Czech Republic has experienced a drastic dismantling of socialist-era collectives and the outflow of part-time and low-productivity labour from agriculture. Three other transition economies, the Slovak Republic, Hungary and Estonia also show rapid convergence to Austrian TFP levels, with the former two countries surpassing Austria by the end of the sample period and Estonia’s TFP is nearly equal to that of Austria. These four countries’ agrarian sectors thus already operate at productivity levels that are comparable to that of an old EU member with an above (old EU) average per capita income. This suggests that concerns that CAP funding would be required to support relatively inefficient agrarian sectors in these countries are exaggerated. For the other transition economies the picture is less favourable. Slovenia started the period with TFP in agriculture at about three-fourths of Austria’s, but its TFP growth failed to match Austria’s over the sample period so that the TFP gap between the two countries widened. The two Baltic Republics, Latvia and Lithuania have TFP about one half of Austria’s and they made only modest progress in closing this productivity gap between 1995 and 2005. Poland and Bulgaria fell farther behind Austria, and both have TFPs in agriculture that are less than one-half of Austria’s. For Poland, its many small and inefficient private farms are a likely source of that country’s poor agricultural TFP showing. In these countries, lagging agrarian productivity may require additional structural support from the EU if the CAP is not to be burdened with inefficient farm sectors. Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

439

Figure 3: Industrial TFP Relative to Austria.

Figure 3 provides similar comparisons of the transition countries’ TFPs in industry to that of Austria. None of the transition economies has an industrial TFP that matches that of Austria. Nevertheless, four transition countries, Estonia, Latvia, Lithuania and Hungary showed large gains in TFP over the sample period, ending the period with TFPs that are from two-thirds to threefourths of Austria’s. Poland experienced a more gradual convergence to Austria’s TFP levels, while the Czech Republic and Slovenia had relatively high levels of TFP, but failed to keep up with TFP growth in Austria over the sample period. The Slovak Republic and Bulgaria had TFPs in industry that were about one-half of Austria’s.19 Because industry continues to be a major part of the new members’ economies, poor performance vis a vis Austria is a significant hindrance to catch-up. The TFPs for services are reported in Figure 4. None of the transition economies matches Austria’s services TFP although Slovenia, the Czech Republic and the Slovak Republic are near, but closing the gap only very slowly. The three Baltic Republics and Hungary, while staring at relatively low TFP levels, all made significant improvements in service sector TFP over the sample period. In contrast, Poland and Bulgaria had low TFPs and failed to make much headway in catching up with the other economies. This analysis of relative TFP performance yields several conclusions. The first is that TFP in agriculture is not a major barrier to catch-up for many new 19

The divergence in industrial TFP performance in the transition economies is somewhat surprising. All received massive inflows of FDI, investment rates have not differed much, and general reform measures have been quite similar as well. Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

440

Figure 4: Service TFP Relative to Austria.

members because they already have achieved relatively high productivity levels. Moreover, agriculture plays a diminishing role in aggregate economic activity. Second, some countries, such as Bulgaria, have significant problems in achieving acceptable rates of TFP growth in industry and services, and, given the growing share of these sectors in aggregate output, this poor performance is a real barrier to these countries’ efforts to catch up to the average EU per capita income. Conversely, some of the transition economies, such as the Baltic States and Hungary, are making good progress in raising TFP levels in both industry and services, and these countries thus should also experience high aggregate TFP growth that will facilitate their convergence to EU living standards. Sources of economic growth in new EU member countries So far, we have shown that the new EU member countries have not had their growth severely hampered by structural change, and that they differ considerably among themselves with respect to which of their sectors’ TFP levels are closest to those of Austria and which are contributing the most to aggregate TFP growth. In this section we return to the question of aggregate TFP growth and catch-up for the new EU member countries. In the introduction, we noted that the greatest obstacle to measuring aggregate TFP growth in the transition countries lies in estimating the capital stock. We therefore use Equation 12 to compute the aggregate capital stock for our sample of transition economies. If there were a major reduction in the starting capital stock of transition economies for the reasons discussed at the Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

441 Table 7: Estimates of Aggregate Capital Stock, 1995=100

Austria Bulgaria Czech Republic Estonia Hungary Latvia Lithuania Poland Slovak Republic Slovenia

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

103 109 108 113 107 115 113 113 109 113

107 118 115 126 115 131 126 126 119 126

110 127 122 139 122 148 140 138 127 138

114 136 129 154 130 166 155 151 136 150

117 145 135 169 138 185 170 163 144 161

120 153 141 185 146 205 186 175 152 172

124 162 147 202 154 226 202 187 160 182

127 171 153 220 163 250 220 200 168 192

130 181 159 239 172 275 238 212 175 201

134 191 164 260 181 302 258 224 182 210

beginning of this paper, then the modelled stock of capita would exhibit faster growth from this lower starting point than would the official capital stock data. Moreover, if this difference in growth rates were significant, then a growth accounting exercise based on our estimates of the capital stock would show a faster growth rate of the capital stock and a correspondingly lower growth of TFP than would be obtained by undertaking the same exercise using capital stock series that did not adjust for the excess destruction of capital. Table 7 provides the aggregate capital stocks estimated from Equation 12 for each country for the period 1995–2005. All the transition economies experienced faster capital stock growth than did Austria, but there were also important differences among the transition countries themselves, with the Baltic Republics noteworthy for the rapid growth of their capital stocks. Table 8 then sets out the results of the growth accounting exercise based on the growth of GDP and of our estimates of the capital stock in the transition countries. As can be seen, GDP, the capital stock and TFP (except in the Czech Republic) grew more rapidly in the new member states than they did in Austria over our sample period. However, only in Estonia and Hungary did TFP growth contribute more than 50 percent of GDP growth, and only in these two countries did the contribution of TFP growth to GDP growth exceed that of Austria, while in Latvia and Lithuania, TFP growth accounted for about the same percentage of aggregate growth as it did in Austria. In the other transition economies, capital accumulation was the main driver of aggregate growth. The extent to which the transition-induced excess depreciation of the capital stock affects our perception of TFP growth of the transition economies is difficult to judge because of a lack of other estimates that are strictly comparable in terms of both time period covered and measurement of inputs. Comparative Economic Studies

442

Growth 1995–2000 (percent)

Austria

Bulgaria

Czech Rep

Estonia

Hungary

Latvia

Lithuania

Poland

Slovak Rep

Slovenia

GDP

19.71

49.32

27.46

107.00

57.73

116.59

89.15

48.74

41.56

45.51

Capital

33.66

90.68

64.01

159.65

80.79

202.09

157.75

124.01

82.45

110.42

9.62

22.12

8.26

59.10

33.50

55.96

41.82

11.54

16.83

12.39

TFP

Contributions of TFP growth to GDP growth as percentage of total growth Our Estimate 1995–2005 Rapacki and Pro´chniak (2009) 1995–2003

48.8

44.8

30.1

55.2

58.0

48.0

46.9

23.8

40.8

27.2

68.5

103.7

112.5

43.0

67.5

125.7

86.7

84.0

69.9

EM Bah & JC Brada TFP Growth and Structural Change

Comparative Economic Studies

Table 8: Contributions of Capital Accumulation and TFP Growth to GDP per capita growth 1995–2005

EM Bah & JC Brada TFP Growth and Structural Change

443

Perhaps the closest in comparability are estimates of the share of TFP in real GDP growth provided by Rapacki and Pro´chniak (2009) who calculated the annual contribution of TFP growth to real GDP growth for transition economies using total employment and perpetual inventory capital stock estimates based on gross investment for the period 1990–2003. We averaged Rapacki and Pro´chniak’s annual estimates over the 1995–2003 period, and these are also reported in Table 8. In their estimates, which do not account for excess depreciation of capital, TFP growth accounts for a significantly higher share of GDP growth, in some cases over 100 percent of it. Some of the difference between their estimates and ours may be the result of differences in measuring the labour input, but, as suggested at the start of this section, our estimates attribute a larger role to capital accumulation in the convergence process than do estimates that do not account for the transition-induced destruction of capital and the subsequent faster growth of the capital stock.

CONCLUSIONS AND POLICY IMPLICATIONS In this paper we have estimated the TFPs of the transition economies that have joined the EU and of a roughly comparable ‘old’ EU member, Austria, which has a per capita income that is higher than that of any of the new members. These differences in per capita income mirror differences in aggregate TFP, as well as in differences between sectoral TFPs, which on average are also lower than those of Austria. However, the TFP gaps between Austria and the new members we observed are not uniform across sectors of the economy. The TFP gap appears smallest in agriculture and greatest in industry or services depending on the country. Moreover, the new member countries differ in the relative TFP levels of the three sectors vis a vis Austria. A second finding is that the transition economies themselves should not be seen as a homogeneous group. There are great TFP differences between them, and perhaps more troubling, some of them are not improving productivity in industry or services or both, suggesting that catching up with the EU average may prove an impossible task for them. The structural changes taking place in the transition economies mirror, but appear to be faster than, those taking place in Austria, and, indeed, in virtually all EU member countries. The proportion of the labour force employed in agriculture is falling, as is that of those employed in industry, while services employ the largest and increasing share of the labour force. This structural change is not necessary favourable for the transition economies in the sense that the TFP gap between themselves and Austria is the smallest in agriculture and larger Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

444

in services and particularly in industry. The gap between Austria and the transition economies in services TFP is not large for some of the transition countries, but more troubling is that about one half of the transition economies are not catching up with Austria in this key sector. Thus, if there is to be convergence in per capita GDP between the old and the new members of the EU, then measures to improve productivity in services and, perhaps as well, in industry will be required. Also noteworthy is that the aggregate TFP growth of each of the transition economies is based on strikingly different TFP dynamics and levels in the three sectors. There is a tendency when discussing the growth and income levels of the transition economies to assume that reforms, liberalisation, and greater reliance on the market, as captured by broad indicators of economic reform, such as the EBRD reform index, rankings of ‘competitiveness’, etc influence the performance of all sectors of the economy in more or less the same way. However, our results suggest that this may not be the case. Countries with broadly similar reform policies and reform ‘scores’ have different sectoral TFP levels and dynamics. This suggests that our measures of reforms may be too broad or that more specific policies that often go unnoticed play a more important role in determining TFP levels and growth than is commonly thought. Finally, our results suggest that the new members do not differ from the old EU members in terms of the contribution of capital accumulation and TFP growth to the growth of GDP. Our estimates, which account for the transitionrelated destruction of capital in Eastern Europe, indicate that capital accumulation has played an important role in per capita income convergence, and thus future convergence also depends on continued high rates of capital accumulation. Acknowledgements We thank Athanasios Vamvakidis, Paul Wachtel and annonymous referees for helpful comments.

REFERENCES Bah, E.M. 2008: A three-sector model of structural transformation and economic development. Department of Economics Working Paper, The University of Auckland. Beare, B.K. 2008: The soviet economic decline revisited. Econ Journal Watch 5(2): 134–144. Blanchard, O.J. 1997: The economics of post-economic transition. Oxford and New York: Oxford University Press. Brada, J.C. 1985: The slowdown in soviet and east European growth. Osteuropa Wirtschaft 30: 116–128. Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

445 Brada, J.C. 1989: Technological progress and factor utilization in eastern European economic growth. Economica 56(3): 433–448. Buiter, W.H. 2000: From predation to accumulation? The second transition decade in Russia. Economics of Transition 8(3): 603–622. Burda, M.C. and Sevegnini, B. 2008a: TFP growth in old and new Europe. Paper presented at the 14th Dubrovnik Economic Conference, Dubrovnik. Burda, M.C. and Sevegnini, B. 2008b: Solow residuals without capital stocks. SFB 649 Discussion Paper. Campos, N.F. and Coricelli, F. 2002: Growth in transition: What we know, what we don’t, and what we should. Journal of Economic Literature 40(3): 793–836. Caselli, F. 2005: Accounting for cross-country income differences. In: Philippe Aghion and Steven Durlaf (eds). Handbook of Economic Growth. Amsterdam: Elsevier. Darvas, Z. and Simon, M. 2000: Capital stock and economic development in Hungary. Economics of Transition 8(1): 97–223. Deliktas, E. and Balcilar, M. 2005: A comparative analysis of productivity growth, catch-up, and convergence in transition economies. Emerging Markets Finance and Trade 41(1): 6–28. Dowrick, S. 1989: Sectoral change, catching up and slowing down: OECD post-war economic growth revisited. Economic Letters 31(4): 331–335. Dowrick, S. and Nguyen, D. 1989: OECD comparative growth 1950–85: Catch-up and convergence. American Economic Review 79(5): 1010–1030. Easterly, W. and Fisher, S. 1995: The soviet economic decline. World Bank Economic Review 9(3): 341–371. Easterly, W. and Fisher, S. 2008: A reply to: Brendan K. Beare. The soviet economic decline revisited. Econ Journal Watch 9(2): 135–144. European Union. 2006: Enlargement, two years after: An economic evaluation. Paper 24, European Union, Bureau of European Policy Advisers and the Directorate- General for Economic and Financial Affairs. Fo¨ldva´ri, P. and Van Leeuwen, B. 2009: Average years of education in Hungary: Annual estimates, 1920–2006. Eastern European Economics 47(2): 5–20. Frankel, J. and Romer, D. 1999: Does trade cause growth? The American Economic Review 89(3): 379–399. Gollin, D. 2002: Getting income shares right. Journal of Political Economy 110(2): 458–474. Granick, D. 1989: The job rights economy. Cambridge: Cambridge University Press. Gregory, P. 1970: Socialist and non-socialist industrialization patterns. London: Praeger. Hall, R.E. and Jones, C. 1999: Why do some countries produce so much more output per worker than others? Quarterly Journal of Economics 114(1): 83–116. Hendricks, L. 2002: How important is human capital for development? Evidence from immigrant earnings. The American Economic Review 92(1): 198–219. Herrendorf, B. and Valentinyi, A. 2006: Which sector make the poor countries so unproductive? Arizona State University Department of Economics Working Paper. Hsieh, C.T. and Klenow, P.. 2007: Relative prices and relative prosperity. American Economic Review 97(3): 562–585. Izumov, A. and Vahaly, J. 2006: New capital accumulation in transition economies: Implications for capital-labor and capital-output ratios. Economic Change 39(1): 63–83. Izumov, A. and Vahaly, J. 2008: Old capital versus new investment in post-soviet economies: Conceptual issues and estimates. Comparative Economic Studies 50(1): 63–83. Klenow, P.J. and Rodriguez-Clare, A. 1997: The neoclassical revival in growth economics: Has it gone too far? In: Ben Bernanke and Julio J. Rotem- berg (eds) NBER Macroeconomics Annual. MIT Press, 73–103. Kushnirsky, F. 2001: A modification of the production function for transition economies reflecting the role of institutional factors. Comparative Economic Studies 42(1): 1–30. Comparative Economic Studies

EM Bah & JC Brada TFP Growth and Structural Change

446 Kuznets, S. 1966: Modern economic growth. New Haven: Yale University Press. McKinsey Report. 1999: Unlocking economic growth in Russia. London: McKinsey Global Institute. Ofer, G. 1976: Industrial structure, urbanization, and the growth strategy of socialist countries. Quarterly Journal of Economics 90(2): 219–244. Ofer, G. 1987: Soviet economic growth: 1928–1985. Journal of Economic Literature 25(4): 1767–1833. Parente, S.L. and Prescott, E.C. 1994: Barriers to technology adoption and development. Journal of Political Economy 102(2): 298–321. Parente, S.L. and Prescott, E.C. 2000: Barriers to riches. Cambridge: MIT Press. Prescott, E.C. 1998: Needed: A theory of total factor productivity. International Economic Review 39(3): 525–551. Rapacki, R. and Pro´chniak, M. 2009: Economic growth accounting in twenty-seven transition countries, 1990–2003. Eastern European Economics 47(2): 69–112. Rogerson, R. 2008: Structural transformation and the determination of European labor market outcomes. Journal of Political Economy 116(2): 235–259. Solow, R. 1957: Technical change and the aggregate production function. The Review of Economics and Statistics 39(3): 312–320. Steffen, W. and Stephan, J. 2008: The role of human capital and managerial skills in explaining productivity gaps between east and west. Eastern European Economics 46(6): 5–24. Stephan, J. 2002: The productivity gap between east and west Europe: What role for sectoral structures during integration? Acta Oeconomica 52(3): 289–305. Van Ark, B. 1999: Economic growth and labour productivity in Europe: Half a century of east-west comparisons http://www.eco.rug.nl/ggdc. Weitzman, M.L. 1970: Soviet postwar economic growth and capital-labor substitution. American Economic Review 60(4): 676–692.

APPENDIX: DATA SOURCES For the United States, the data for GDP per capita, expressed in 1990 international Geary-Khamis dollars, is from the Historical Statistics for the World Economy: 1-2003 AD by Maddison. The shares of sectoral hours worked and the price of services relative to industry are from the Groningen 10-sector industry database. We obtained average capital-to-output ratio and average investment-to-output ratio from the NIPA tables. The price of services relative to industry is calculated using data from the Gronigen 10-sector industry database. The database shows the value-added of each sector in constant and current prices. The price of a sector is obtained by dividing the value-added in current prices by the value-added in constant prices. For Austria and the transition countries, GDP per capita in constant 2000 PPP dollars and the sectoral employment shares are from World Development Indicators Online Database.

Comparative Economic Studies