1 

 

Services in Developing Economies: The deindustrialization debate in perspective1 

 

Gisela Di Meglio*, Jorge Gallego**, Andrés Maroto**, Maria Savona*** 

 
* Department of Economic Analysis II and ICAE. Universidad Complutense de Madrid. 

Corresponding author. E-mail: gdimeglio@ccee.ucm.es 
** Department of Economic Analysis. Universidad Autónoma de Madrid. 
*** Science Policy Research Unit. University of Sussex. 

 

ABSTRACT  

 

The article adds to the debate around the ‘premature deindustrialization’ of developing 

countries by analyzing the contribution of services to aggregate productivity and output 

growth within a Kaldorian framework. We revisit Kaldor´s Growth Laws (KGL) and 

empirically test them for a number of economic activities, including four service 

branches, across twenty-nine developing economies in Asia, Latin-America and Sub-

Saharan Africa over three decades (1975-2005). Panel data estimations are 

complemented by a shift-share decomposition of labour productivity growth. We find 

support to the Kaldorian argument for both manufacturing and business services 

contribution to aggregate productivity growth. Conversely, other services slow down 

aggregate productivity and output growth. We suggest qualifying and repositioning the 

debate on premature deindustrialization within a broader reflection on the opportunities 

for development linked to structural change. We claim that these opportunities might 

include not only manufacturing sectors but also business services.  

 

 

INTRODUCTION AND BACKGROUND 
 

The role of structural change has been core to theories of economic development for 

several decades now. The structure of an economy matters for growth performance and 

development since sectors have different capabilities to achieve and to induce 

productivity gains, as well as to benefit from domestic and foreign demand growth 
                                                           
1 The final version of this paper is published in Development and Change, 2018, 49 (6). pp. 1495-1525. ISSN 0012-
155X. DOI: 10.1111/dech.12444. © 2018 International Institute of Social Studies. 



2 

 

(Cimoli et al., 2009; Thirlwall, 2013). Processes of structural transformation have been 

heterogeneous both across and within developing countries (Bah, 2011), leading to 

different contributions to economic performance (MacMillan and Rodrik, 2011).  

 

A renewed interest in the relationship between structural change and development has 

recently been emerging, particularly at a time when deindustrialization is occurring at 

low levels of per capita income in developing countries, unlike the path that has 

historically characterized developed countries (Bah, 2011;  Rodrik, 2016; Di Meglio, 

2017). Concerns of a ‘premature deindustrialization’ of developing countries (Dasgupta 

and Singh, 2005, 2006; Palma, 2005; Felipe et al., 2015; Rodrik, 2016) are consistent 

with a revamped debate on the role of industrial policy (Stiglitz and Lin, 2013; Storm, 

2015) and the recent plea for countries to ‘reindustrialize’ (Ciarli and Di Maio, 2014; 

Tregenna, 2011; Westkämper, 2014).  

 

Much of the scholarship nurturing concerns around premature deindustrialization 

implicitly assumes the seminal Kaldorian stylized facts, which depict the manufacturing 

sector as the main driver of labour productivity performance, growth and, ultimately, 

economic development (Kaldor, 1966; 1967; Thirlwall, 1983). The so-called Kaldor´s 

Growth Laws (henceforth, KGL) propose that industrialization induces growth of the 

output per worker as a result of two main mechanisms. First, productivity in 

manufacturing rises with the growth of manufacturing output due to the presence of 

increasing returns to scale (IRS) at the sectoral level. Second, output growth in 

manufacturing is shown to positively affect the rate of productivity growth in other 

sectors. Over time, several other studies have examined and confirmed the interpretation 

and the validity of the different KGL from a variety of perspectives. The evidence 

abounds on developed economies.2 However, it remains fairly fragmented on 

developing countries.  

 

The fundamental role played by manufacturing as a source of growth in developing 

countries is shown, among others, by Felipe (1998) across five Southeast Asian 

countries, by Cimoli et al., (2005) and Libanio (2006) across five and seven Latin-

American economies, respectively, and by Pacheco and Thirwall (2014) for eighty-nine 

developing economies. However, given the traditional picture of services as a structural 
                                                           
2 See Romero (2016) for a review. 



3 

 

burden for economies, the literature has mostly overlooked non-manufacturing sectors 

(Ocampo, 2005), whilst the few studies that focus on services mainly account for the 

aggregate sector. In this respect, Dasgupta and Singh (2005; 2006) have pioneered the 

argument that, although manufacturing continues to be critical for development, 

services can also be regarded as an additional engine of growth. In the same vein, Felipe 

et al. (2009) argued that in the case of Asia, notwithstanding the heterogeneity of the 

productive structure within this macro-region, services can also have productivity-

growth inducing effects through the exploitations of scale economies.  

 

The cross-sector performance of services is heterogeneous. Some service sectors, 

particularly knowledge-intensive and other business-related services,3 have challenged 

the ‘old myths’ of services as a structural burden (Gallouj and Savona, 2008) and have 

turned into important sources of productivity growth also in developing countries 

(Timmer and de Vries, 2007; 2009). Therefore, country-level processes of 

deindustrialization should be assessed by distinguishing the roles that different service 

branches may have in the economy.  

 

The aim of this article is to add to the recent debate on premature deindustrialization of 

developing countries, particularly to reposition the debate within a broader reflection on 

the opportunities for development linked to structural change. To do so, we build upon 

and extend the Kaldorian framework to include a number of service branches.  

The novelty of the empirical strategy is two-fold: first, we implement econometric tests 

of the KGL on a number of industries, including four service sub-sectors, across twenty-

nine developing countries from Asia, Latin-America and Sub-Saharan Africa over the 

last three decades (1975-2005). To our knowledge, no study has previously tested the 

validity of the Kaldorian stylized facts for services at this level of disaggregation. 

Second, panel data estimations are complemented with a decomposition of labour 

productivity growth, by means of a shift-share analysis, in order to study how resource 

(labour) reallocation during the process of structural change has affected economic and 

labour productivity growth.4 Overall, this research also fills a gap in the current 

                                                           
3 Knowledge-intensive services usually include: Computer and related activities (K72 ISIC code), Research and 
Development (K73), and Other business activities (K74) such as engineering, technical consultancy, legal aid and 
other business services.   
4 Our research interest lies exclusively on studying changes in the reallocation of labour across sectors. However, we 
are aware that structural change is a much broader concept encompassing many other transformations taking place in 
the economy, e.g.: in savings and investments rates, in urbanisation, in institutions, etc. (Matsuyama, 2009). 



4 

 

literature by revisiting the three KGL for agriculture, manufacturing, services and 

examining the role played by different service sub-sectors in economic performance 

across Asian, Latin America and African developing countries.  

 

We find support to the Kaldorian argument for manufacturing sectors across all the 

macro-regions object of our analysis. We also find robust empirical support of it for the 

branch of business services. However, other services including personal and informal 

ones represent a structural burden for aggregate productivity and output growth. 

Importantly, we qualify the heterogeneous patterns of structural change that have 

characterized the different macro-areas and have explained their heterogeneous 

economic performance.  

 

The manuscript is structured as follows. Section 2 revisits the classical Kaldorian 

framework. Section 3 presents the data and the empirical strategy. Section 4 discusses 

the econometric results. Section 5 summarizes the main findings, and Section 6 

concludes by identifying priorities for future research agenda.  

 

 

THE KALDORIAN FRAMEWORK 
 

In his seminal contributions, Nicholas Kaldor (1966, 1967) attempts to explain large 

growth rates differences across 12 OECD countries during 1953-1964 by adopting a 

sectoral approach where dualisms à la Lewis (Lewis, 1954) can be found.5 Kaldor 

argues that both the production and demand characteristics of each aggregate sector of 

the economy (agriculture, industry and services) matter for economic growth. In 

particular, the capital-intensive manufacturing sector shows greater potential for 

productivity growth than other sectors due to the presence of both static and dynamic 

economies of scale. Accordingly, the reallocation of labour from activities subject to 

diminishing returns of scale (i.e., agriculture) to more productive sectors (i.e., 

manufacturing) fosters productivity growth in both sectors and overall output 

expansion. Manufacturing also has greater potential than other sectors for releasing 

balance of payment constraints due to the higher tradability of manufactured products. 

Therefore, external demand growth of such products may spark a virtuous circle of 
                                                           
5 For a review on Kaldor’s contributions to development economics see Targetti (2005). 



5 

 

cumulative growth (Dixon and Thirlwall, 1975; Kaldor, 1970). Within this framework, 

Kaldor articulates a set of long-run relationships or empirical generalizations of growth 

of output, employment, and productivity at the sectoral level of the economy, which 

later became known as the Kaldor's Growth Laws (KGL). 

 

The first law states that the faster the growth of manufacturing output (𝑞 ) in an 

economy, the faster the growth of gross domestic product (𝑞 ). Kaldor fundamentally 

defines a causal relationship running from sectoral to aggregate growth and more 

precisely from manufacturing growth to the growth rate of GDP per worker (Ros, 2000) 

as shown in Kaldor´s second and third law. The first law can be posited then as:                                                                                            

 

 𝑞 = 𝑓 (𝑞 )         𝑓′ > 0   

 

According to Thirlwall (2013), there are two additional regressions that overcome the 

problem of spurious correlation that is evidently present in the former specification6:  

 𝑞 = 𝑓 (𝑞 − 𝑞 )         𝑓′ > 0  

 𝑞 = 𝑓 (𝑞 )                       𝑓′ > 0   

 

In the first equation the growth of GDP (𝑞 ) is regressed on the excess of the growth 

of manufacturing production (𝑞 ) relative to the growth on non-manufacturing 

production (𝑞 ). In the second equation the growth of non-manufacturing output is 

regressed on the growth of manufacturing output, the estimated coefficient indicating 

the strength and size of the impact of manufacturing sector growth on the rest of the 

economy.  

 

Accordingly, the linear specifications for examining Kaldor´s First Law at sectoral level 

are: 

 

𝑞 = 𝛼 + 𝛽 𝑞 + 𝜀                  (Equation I-A) 

𝑞 = 𝛼 + 𝛽 (𝑞 − 𝑞 ) + 𝜀               (Equation I-B) 

𝑞 = 𝛼 + 𝛽 𝑞 + 𝜀                 (Equation I-C) 

 

                                                           
6 Since total output growth is the weighted sum of sectoral output growth. 



6 

 

where j, i, t stand for sector, country and time, respectively, and εjit is assumed to be 

normally distributed. 𝑞  is total output growth (the growth of total value added in 

constant prices) and 𝑞  refers to the sectoral output growth (the growth of sectoral 

value added in constant prices).  

 

The second law states that the faster the growth of manufacturing output (𝑞 ), the 

faster will be the growth of productivity in manufacturing (𝑝 ) as a result of increasing 

returns to scale (IRS). This first mechanism explaining causality from manufacturing 

growth to GDP per worker growth is known as the Verdoorn´s law. This law can be 

interpreted from the perspective of employment growth in manufacturing (𝑒 ): the 

higher the scale economies of the sector, the lower the employment elasticity with 

respect to output, since productivity increases as a result of output expansion. This 

means that output expansion induces a less than proportional employment creation that 

causes productivity gains. 

 

Kaldor (1966) specified the Verdoorn relation in terms of a linear regression model: 

𝑒 = 𝛽 + 𝛽 𝑞  with 𝛽 > 0, being 𝛽  an indicator of IRS (the Verdoorn coefficient). 

However, due to the productivity identity, this can be expressed as: 𝑒 = −𝛽 + (1 −

𝛽 )𝑞 , with 10 1   , being )1( 1  the elasticity of employment with respect to output 

growth. Then, the specification of Kaldor´s Second Law can be expressed by:  

 

𝑒 = 𝛼 + 𝛽 𝑞 + 𝜀   (Equation II) 

 

where j, i, t stand for sector, country and time, respectively, and εjit is assumed to be 

normally distributed. 𝑒  is the sectoral employment growth and 𝑞  is the sectoral 

output growth. 

 

Finally, the third law states that a positive causal relationship exists between the 

expansion of the manufacturing sector and the growth of labour productivity outside of 

the manufacturing sector. This is explained by the presence of diminishing returns in 

agriculture and services, which supply labour to industry. This represents the second 

mechanism explaining causality from manufacturing to labour productivity growth. The 

reallocation of labour from agriculture to manufacturing releases surplus labour from 



7 

 

the non-dynamic sectors of the economy and, as a result, overall productivity growth 

increases.   

 

Most empirical studies focusing on developing economies (i.e., Dasgupta and Singh, 

2005, 2006; Wells and Thrilwall, 2003) estimate this law by regressing overall 

productivity growth (𝑝) on the growth of non-manufacturing employment (𝑒 ), 

controlling for the growth of manufacturing output (𝑞 ), which, according to the 

Verdoorn´s law, induces productivity growth. Accordingly, the linear specification can 

be written as follows: 

 

 𝑝 = 𝛽 + 𝛽 𝑒 + 𝛽 𝑞 , with  𝛽 < 0; 𝛽 > 0     (Equation III) 

 

Table 1 reviews and summarizes a number of KGL- related works for developing 

economies and classifies their different approaches on the basis of their country sample 

(N), their time horizon (T), the equations estimated in the empirical exercise, and the 

level of sectoral disaggregation adopted. As mentioned in the introductive section, only 

a few studies have drawn attention to the (aggregated) services sector in their 

estimations. A notable exception is Pieper (2003) who examines the Verdoorn´s Law 

(Kaldor´s Second Law) across 30 developing countries covering nine sectors.  

 

 

Table 1 here 

 

 

DATA AND METHODOLOGY 

 

The main sources of data used here are the Groningen Growth and Development Centre 

(GGDC) 10-Sector Database (Timmer and de Vries, 2007) and the Africa Sector 

Database (de Vries et al., 2013). These are the first databases that provide long-term 

series of value added (in current and constant prices) and employment for developing 

economies. On the one hand, these sources compute employment levels using 

population censuses information that tends to have a more complete coverage of 

informality (McMillan et al., 2014) particularly relevant in the analysis of developing 

countries. On the other hand, value added information is gathered within the framework 



8 

 

of the System of National Accounts, which makes the coverage of the informal sector 

by value added data to vary across countries and to depend on the quality of the national 

sources.  

 

With regard to the time span, the year 1975 was chosen as a starting point because of 

data availability for all countries in the sample. Moreover, following Pieper (2003) and 

León-Ledesma (2000), we use a moving average of value added (at constant prices), 

employment, and productivity growth rates (taking 5-years period averages) to smooth 

out short-term fluctuations present in the annual data (T=6). As detailed in Table 2, 

available time series allow us to disentangle the different role played by the three main 

aggregated sectors (j=1, 2 and 3) as well as a range of service sub-sectors (j= 4, 5, 6 and 

7). Therefore, the analysis is performed for seven different activities, namely: 1) 

Agriculture; 2) Manufacturing; 3) Services; 4) Trade services; 5) Transport services; 6) 

Business services; and 7) Public services.  

 

 

Table 2 here 

 

 

The empirical strategy followed in this research is two-fold. First, we use panel data 

analysis to estimate Kaldor´s first and second laws (Equations I-A, I-B, I-C and II). 

Regressions are performed by sector panel both for the whole sample of 29 developing 

countries and also for the three different regions, including nine countries from Latin-

America, nine countries from Asia and eleven countries from Africa.7 Accordingly, in 

the overall econometric analysis every sector panel ends up having 174 observations 

based on five-year growth rates. When dealing with the three regional sub-samples 

individually, every sector panel for the Asian and Latin-American countries included in 

the sample presents 54 observations (N=9 and T=6), while for the Sub-Saharan African 

economies each sector panel includes 66 observations (N=11 and T=6). Outliers are 

detected and treated using one dummy variable for each. Fixed country effects are 

added in order to deal with omitted heterogeneity. Equations are estimated by Ordinary 

                                                           
7 Latin-American countries (N=9): Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Mexico, Peru and 
Venezuela; Asian countries (N=9): Hong Kong (China), India, Indonesia,  Rep. of Korea, Malaysia, Philippines, 
Singapore, Taiwan and Thailand; African countries (N=11): Botswana, Ethiopia, Ghana, Kenya, Malawi, Mauritius, 
Nigeria, Senegal, South Africa, Tanzania and Zambia. 



9 

 

Least Squares (OLS) with Panel Corrected Standard Error Estimations (PCSE) 

accounting for group-wise heteroskedasticity, cross sectional dependence, and 

autocorrelation in disturbances within panels. 

 

Second, growth accounting can be a useful approach for analysing the relationships 

underlying Kaldor’s Third Law, in order to overcome the matter of spurious correlation 

and identification problems present in Equation III. Accordingly, following Felipe et al. 

(2009) and McMillan and Rodrik (2011), we perform a conventional8 shift-share 

decomposition analysis that allows us to decompose aggregate productivity growth in 

terms of (within sectors) differential growth of labour productivity and the reallocation 

of labour between industries (see Syrquin, 1984, for an overview). 

The decomposition was pioneered by Fabricant (1942) but later users of this method 

focused more on the labour productivity. Let π denote the labour productivity level, 

subscript j denote sectoral branches (j = 1,…,n with n the number of branches), sj the 

share of branch j in total employment and superscripts 0 and T the beginning and end of 

the period (0,T).  

Formally, the decomposition analysis is written as follows: 

 

      0 0 0 0 0 0
1 1 10

0 0 0 0

N N N

j jT j jT j jT j j jT j
j j jT

s s s s s    
 
   

  

   


   
  

   (Equation IV) 

 

ISESCEISEDSESSE   

 

According to equation IV, aggregate productivity growth can be decomposed into intra-

sectoral productivity growth (ISE, the last term on the right-hand side) and the effects 

of structural change (SCE) which consist of a static shift effect (SSE, the first term) and 

                                                           
8 The use of this technique needs some qualifications (Timmer and Szirmai, 2000). First of all, the shift-share analysis 
is supply-side oriented and assesses the effects of structural change on productivity growth. Changes in demand are 
taken as exogenously determined. What the shift-share shows is that the shifts in inputs that have taken place, 
whether or not driven by developments on the demand side, are (or are not) important in quantitative terms for 
aggregate productivity growth. This technique is based on several assumptions. The violation of any of these 
assumptions might result in an under- or overestimation of the contribution of structural change to productivity 
growth. The problematic assumptions involve the aggregate level of the analysis, the assumption of marginal 
productivity equal to average productivity, the assumption of input homogeneity, the incidence of spillovers and the 
causal links between growth of output and productivity.  



10 

 

a dynamic shift effect (DSE, the second term). While the static effect measures 

productivity growth caused by a shift of labour towards sectors with a higher labour 

productivity level at the beginning of the period, the dynamic effect captures shifts 

towards more dynamic sectors, i.e. those with higher labour productivity growth rates. 

This interaction effect arises because of the use of a discrete fixed weight 

decomposition. We retain this term because it can provide an interesting economic 

interpretation for our analysis. As sectors differ not only in terms of productivity levels, 

but also in terms of productivity growth rates, resource reallocation has both static and 

dynamic effects and a distinction between the two is relevant.  

 

The empirical analysis tackles two hypotheses related to structural effects. First, the 

structural bonus hypothesis postulates a positive relationship between structural change 

and economic growth as economies upgrade from low to higher productivity industries 

(SCE > 0). Secondly, the DSE can be used to test the structural burden hypothesis 

(DSE < 0). This hypothesis states that as labour reallocates into sectors with (generally) 

lower productivity growth, productivity growth of the economy will decline. This 

hypothesis is related to the existence (or not) of increasing returns to scale within the 

sectors under consideration. If DSE > 0 increasing returns to scale (IRS) exist 

(industries which absorb more resources are those with increasing productivity 

levels).These results complement to some extent those obtained from the estimation of 

the second KGL. 

 

FINDINGS 
 

Kaldor´s First Law 

 
Panel data estimations for the whole sample of developing economies are shown in 

Table 3. In every sector under analysis the estimated coefficients are significant and 

follow the expected sign in Equation I-A and I-C. However, when Equation I-B is 

estimated, the regression coefficient is significant but negative in both agriculture and 

public services. In fact, only two sectors fulfill Equation I-B: manufacturing and 

business services.  Consequently, as in Dasgupta and Singh (2005 and 2006), we 

confirm Kaldor´s First Law across developing countries: manufacturing is an engine of 

output growth. Our results also show that one of the service sub-sectors, namely 



11 

 

business services, seems to behave in a similar way.  This sector embraces a set of 

different activities including standardized, most likely low skilled services (like real 

estate, cleaning or security) as well as customized human-capital intensive services such 

as R&D, computer-related activities, consultancy, engineering, advertising, etc. 

Business services play an essential role in the creation and diffusion of knowledge, new 

technologies and non-technological modes of innovation (Ciarli et al., 2012; Crespi et 

al., 2014; Gallouj and Savona, 2008; Guerrieri and Meliciani, 2005). Moreover, these 

activities have achieved productivity improvements that have outperformed 

manufacturing sectors’ ones (i.e., Maroto-Sánchez and Cuadrado-Roura, 2009; Timmer 

and de Vries, 2007 and 2009; United Nations, 2010). Therefore, our results may reflect 

changes in inter-industry linkages of developing economies as result of the increased 

use of business services as intermediate catering for manufacturing demand for more 

specialized functions.   

 

Table 3 here 

 

 

Kaldor’s First Law is also confirmed in the three macro regions under analysis, as 

shown in Appendix A. Manufacturing satisfies Equation I-A as well as both side tests. 

This finding is in line with those by Wells and Thirlwall (2003) for 45 African 

economies, Libanio (2006) for 7 Latin-American economies, and Felipe et al. (2009) for 

17 Asian countries. Moreover, no relationship between the expansion of agriculture and 

overall growth is found in Asia, whereas in Latin-America and Africa estimates for 

Equation I-B are significant but negative, as in the total country sample. This result is 

also found for the case of public services in the three different macro regions. 

 

Overall, business services have behaved similarly to manufacturing in Asia and Latin-

America, although no evidence is found across African economies. This may be related 

to the undersized manufacturing basis attained by this latter region, which hampers the 

development of many business-related services. Deindustrialization in Africa is 

characterized by a declining diversity and sophistication of the region’s manufacturing 

sectors (Page, 2011, McMillan et al., 2014). Indeed, when the side test is performed 



12 

 

relating overall growth to the excess of business services output growth (over the non-

business services growth), the regression coefficient is significant and negative.  

 

Kaldor´s Second Law  

 

Table 4 reports panel data estimations for Equation II and provides one-tailed test 

hyphotesis of constant returns to scale (CRS, β1j =1) and increasing returns to scale 

(IRS, β1j < 1) at the sectoral level. All sectors, with the exception of agriculture and 

trade, show employment elasticities with respect to output growth that are significant. 

Moreover, in five out of seven sectors we can reject the CRS hypothesis at the 5 per 

cent confidence interval for one-tailed tests. In those sectors, estimates of the 

employment elasticity with respect to output growth are significantly less than unity. 

 

 

Table 4 here 

 

 

Kaldor´s second law is therefore confirmed for the whole sample of developing 

economies: there are IRS in manufacturing activities. The estimated elasticity is close to 

0.5, a result in line with the traditional estimates of this effect (Felipe et al., 2009). 

Besides, Table 4 shows lower employment elasticities (i.e. higher degree of induced 

productivity growth) in services sectors in comparison with manufacturing. Despite 

surprising, this evidence of strong IRS in services is in line with the findings in Pieper 

(2003). In the linear specification of the law including country effects, the author finds 

that estimated employment elasticities are lower in the service sub-sectors in 

comparison with manufacturing. This finding is highly relevant, as it points out that: i) 

services may be subject to increasing returns, and ii) the same Kaldorian mechanisms 

which make manufacturing the engine of growth may also apply to service sectors. 

Unfortunately, we have found no further studies in the Kaldorian tradition to compare 

our findings for disaggregated service subsectors across developing economies. There is 

still a substantial gap in the Kaldor-Verdoorn related literature with regard to the full 

understanding of the specific factors behind differences in the magnitude of returns to 

scale across sectors, countries and over time (Romero, 2016).  

 



13 

 

When fitting Equation II to the data on Asia and Africa, evidence of IRS in 

manufacturing is also found (Appendix B). This is in line with Felipe et al. (2009) and 

Wells and Thirlwall (2003). In contrast with Libanio (2006), no relationship between 

employment growth and output growth in manufacturing is found in Latin America.9  

 

On the one hand, our results may be reflecting the increasing share in world total 

manufacturing output and the rapid technological upgrade of Asian manufacturers. 

Felipe et al. (2015) show that, in general, Asian countries have had larger manufacturing 

employment shares during 1970-2010 than their African or Latin Americans 

counterparts. Furthermore, during the past four decades, Asia has experienced faster 

capital deepening and higher Total Factor Productivity (TFP) growth than other 

developing economies, including Latin-America (Jaumotte and Spatafora, 2007). At 

sectoral level, productivity growth in industry and in services was higher than in other 

regions and, within the manufacturing sector, there has been a shift towards more skill-

intensive sectors with higher productivity levels and growth. As suggested by Felipe et 

al. (2009), manufacturing output in a number of Asian economies (i.e., South Korea, 

Malasya, Taiwan and Singapore) has shifted into more technology- and scale-intensive 

subsectors. This has been supported by strong institutional quality, trade openness, and 

financial sector development. Thus, notwithstanding the high heterogeneity within Asia, 

these facts helped to promote the catching-up with advanced economies (Jaumotte and 

Spatafora, 2007).  

 

On the other hand, technology- and knowledge-intensive industries have lost ground 

over the past decades in many Latin-American countries, which have experienced a 

relative decline in productivity growth (Pagés, 2010), combined with a relative large 

share of non-skilled intensive sectors within manufacturing (Jaumotte and Spatafora, 

2007). The manufacturing output has largely shrunken, in favour of natural-resource 

processing industries (i.e., tobacco, coal, paper, and petrol) (Cimoli et al., 2005). There 

is evidence that points to a strong shift towards processing industries related to 

commodities for highly competitive world markets (Cimoli and Katz, 2003), 

accompanied by domestic sources of technology change and productivity growth 

rapidly decreasing. 

 
                                                           
9 Libanio (2006) estimates Equation II but also controlling for the growth of capital.  



14 

 

More importantly, our findings induce to question the traditional role posed to services 

as unlikely drivers of productivity growth in developing economies. Following the 

aggregated picture, IRS are found in total services in both Asia and Africa. As in 

manufacturing, no relationship is found between employment growth and output growth 

in services for the Latin-American economies. According to Wells and Thirlwall 

(2003), no economic meaning can be attached to this kind of result except that 

employment growth seems to be independent of output growth. It is important to point 

out that business services show IRS in both Asian and Latin-American countries.10 

Relatedly, Timmer and de Vries (2009) suggest that market services (including trade, 

financial, business services and communications) have been important contributors to 

growth and development in Asia and Latin-America from 1950 to 2005. These authors 

find that productivity gains within manufacturing and market services are key drivers 

for growth. 

 

The regression exercise on Kaldor´s second law indicates the capability of sectors for 

generating induced productivity growth. As this is still the object of a number of 

empirical controversies –as will be discussed later on–, the next section provides a 

direct measurement of the contributions of structural change to productivity growth by 

means of a dynamic shift-share analysis.  

 

 

Kaldor´s Third Law 

 

The results of productivity growth decomposition accounting for Kaldor´s third law are 

shown in Table 5, broken down into sectoral contributions.11 In line with Equation IV, 

the sum of the structural effects (SCE = SSE + DSE) and the intra-sectoral productivity 

growth effect (ISE) is equal to the average growth rate of labour productivity in the 

corresponding aggregate (first cell). This is how the data sums up horizontally. 

Vertically, for each of the three components, the contributions of each sector also add 

up to the corresponding figure in the first line of each sub-table. As additional 

information, the number in brackets shows the average growth of labour productivity 

                                                           
10 In African countries, the CRS hypothesis cannot be rejected at the 5 per cent confidence level. 
11 The category ’Other industry’ (comprising mining and extracting activities, construction, and energy) is also 
included in this section in order to obtain valid results of shift-share technique.   
 



15 

 

within individual sectors, and does not add up either in the horizontal or in the vertical 

dimensions. The figures allow us to identify whether there are any regular patterns of 

differential productivity growth across industries. 

 

 

Table 5 here 

 

Results show that the moderate labour productivity growth (0.68 per cent) of 

developing countries in the period under analysis is largely explained by the intra-

sectoral effect (ISE). In particular, almost three quarters of the total productivity growth 

correspond to such component while structural change (SCE) only accounts for the 23.7 

per cent. This finding is consistent with the one shown by McMillan and Rodrik (2011) 

for a quite similar period (1990-2005). As it is often the case in the relevant literature 

(see Peneder, 2003 and Maroto and Cuadrado, 2009, for a review), the structural 

components (SCE) seem to be generally dominated by the within effects of productivity 

growth (ISE). 

 

Results also show that the structural bonus hypothesis is rejected for agriculture (SCE = 

-0.127) and is scarcely supported for manufacturing sector (+0.005).. However, it 

appears to play a more important role in services (SCE = +0.363). In particular, 

business services and trade account for more than the 70 per cent of the structural 

effects within services. This suggests that labour has reallocated from primary activities 

into manufacturing and, mainly, services. Nevertheless, industries absorbing resources 

have lost dynamism during 1975–2005 as the structural burden emerges (DSE<0)in 

both manufacturing and services – and, particularly, in trade.12 

 

Figures from Table 5 hide significant differences across regions. Appendix C shows 

that Asia has the highest labour productivity growth during 1975–2005 (1.65 per cent) 

which is mostly explained by the intra-sectoral component (82.5 per cent). Moreover, 

the structural bonus (SCE > 0) is found in all sectors with the exception of agriculture. 

Services show enough place to productivity gains through factor reallocation effects 

                                                           
12 The size of the interaction effect will of course depend on the length of the period under consideration because it 
vanishes when the length approaches to 0. Our results on the Table 5 are robust to the various ways in which the 
structural decomposition formula can be applied. Using annual data instead of data at the beginning- and at the end-
year of a period (as shown in the Table 5) the results are quite similar. 



16 

 

(SCE = 0.617) mainly because business services accounts for a large part of such 

structural bonus (40 per cent). The DSE is negative but rather small in comparison with 

the other regions under study with dynamic productivity gains emerging from both 

manufacturing (+0.009) and services industries (+0.174). This means that Asian 

economies – especially Indonesia, Thailand and Taiwan – seem not to have reached the 

structural burden yet in these activities. 

 

The situation in Latin America is quite the opposite. This group of countries shows an 

extremely poor productivity performance during the three decades analyzed (0.008 per 

cent). Notably, this is the only developing region in which static productivity losses are 

observed in manufacturing. The intra-sectoral productivity growth only accounts for 

one quarter while structural effects account for the other three quarters. The structural 

burden is comparatively important for the case of Latin America (with dynamic effects 

quantitatively similar to static effects), showing a negative DSE in all sectors under 

study. Finally, the African region follows, to a certain extent, the pattern found for the 

whole set of developing economies but with poorer developments. Labour productivity 

growth (0.40 per cent) is mostly explained by the within component (54.5 per cent).  

 

Overall, we find that both manufacturing and services, mainly business services, show 

structural productivity gains for the whole set of developing countries. However, the 

quantitative gains of the structural bonus differ across regions. In Asia, the structural 

effects are quantitatively larger (+0.289) than the case of poor performing Latin 

America (+0.006) or Africa (+0.184). Additionally, Asia shows increasing returns to 

scale both in manufacturing and business services (in the same way we observed in the 

results shown in Table 4 for the overall sample) as the DSE is positive for the whole 

period. On the contrary, Latin America and Africa show negative DSE – both in 

manufacturing and in most service branches.  

 

In this respect, MacMillan and Rodrik (2011) point out that Asian countries have 

experienced growth-enhancing structural change during 1990-2005, whereas in Africa 

and Latin-America growth-reducing structural change has prevailed, indicating that 

labour has shifted from high-productivity sectors (i.e., manufacturing) to less productive 



17 

 

activities (i.e., personal services, informality or even unemployment).13 Both low-

income countries of Sub-Saharan Africa and middle-income economies of Latin-

America have been intensely hit by deindustrialization, while Asian regions have been 

insulated from this trend (Rodrik, 2016). This kind of ‘wrong’ structural transformation 

is suggested to be related to the presence of large endowments of natural resources 

(which do not generate much employment unlike manufacturing industries and 

business-related services), the overvaluation of currencies (which have a negative effect 

on tradable modern sectors), and the reduced flexibility of labour markets (which 

hampers the flow of labour across firms and sectors). 

 

CONCLUDING REMARKS  
 

This article has contributed to a recently revamped debate around the threat of 

‘premature deindustrialization’ in emerging economies, originally put forward by 

Dasgupta and Singh (2005, 2006) and Palma (2005) and recently reprised by Bah 

(2011), Felipe and Mehta (2016) and Rodrik (2016). Concerns around 

deindustrialization are consistent with the narrative, present in academic and policy 

circles, of ‘industrial policy is back’14, that suggests to go back to public active 

industrial policies to support resurgence of manufacturing sectors.  

 

We have argued that within this context it is important to qualify and give empirical 

content to the effects of premature deindustrialization, not least to properly ground 

industrial policies in developing countries. We therefore identified a few empirical 

stylized facts on the contribution of a set of different service branches – as well as the 

manufacturing and primary sectors – to aggregate growth and productivity performance 

in twenty-nine developing countries over the past decades. To do so, we have reprised 

the classical Kaldorian framework and devised an original empirical strategy, that has 

included estimations of the Kaldor-Verdoorn Laws, complemented by a shift-share 

analysis.  

 

                                                           
13 Recently, MacMillan et al. (2014) decompose their results for the period 1990-2000 and 2000-onwards, and find 
that structural change has been also growth-enhancing for the case of Africa in the latter period as result of small 
expansions in different manufacturing sub-sectors. 
14 See the recent Stiglitz (2016), the Juncker Plan in Europe, the Made in China 2025 programme, the Indian National 
Manufacturing Policy, and the new industrial strategy policies around the globe. 



18 

 

Our findings suggest that the manufacturing sector has indeed been an engine of growth 

during the past three decades across Asian, Latin-American and African countries, as 

the KGL remain valid for developing economies. More importantly, the evidence 

suggests that within the heterogeneous service sector, business services represent an 

additional engine of growth, as they contribute to aggregate productivity by mean of the 

same Kaldorian mechanisms that have traditionally been at work in manufacturing 

industries. 

 

Indeed, much of the attention devoted by the literature to business services, and 

knowledge-intensive services in particular, has focused on their capabilities to embody, 

process, accumulate and disseminate both codified and tacit information and knowledge 

to other firms and sectors. Such a role is grounded in their high share of skilled human 

capital, their contribution to learning processes and knowledge accumulation, and their 

role as co-producers of innovation (Gallego and Maroto, 2015) – e.g., by facilitating 

knowledge transfer coming from foreign firms locating in developing countries. 

Additionally, an important number of technology-intensive manufacturing sectors 

represent a pool of demand for these knowledge-based business services (Guerrieri and 

Meliciani, 2005), which points to the importance of (forward and backward) inter-

industry linkages between business services and the manufacturing sector, and with 

their use of knowledge and technology (Ciarli et al., 2012; Meliciani and Savona, 2015; 

López-Gonzalez et al., 2015). In this respect, a core manufacturing sector may be 

critical for growth not only per se, but also as it is able to promote the emergence of 

backward and forward linked sectors that Hirschman (1958) would label as ‘high 

development’ inducive (López-Gonzalez et al., 2015), with business services fitting this 

category.   

 

Within the developing world, Asia emerges as the only macro-region where both 

manufacturing and business services consistently and systematically behave as dynamic 

sectors in the Kaldorian sense. This might be due to the emergence and diffusion of 

Global Value Chains (GVCs) across countries that had prior exhibited lower 

productivity and are now catching up, mostly in Asia (Felipe and Mehta, 2016). The 

specialization of the region in export-oriented manufacturing with strong inter-industry 

linkages allow for the development of high-tech business service sectors, which 

stimulate productivity and growth. The increased tradability of (manufacturing-linked) 



19 

 

business services within GVCs have turned these activities into major players in the 

current wave of the globalization process (Gallego et al., 2013), opening up new 

opportunities for growth in developing economies (Gereffi and Fernández, 2010a; 

2010b; Hernández et al., 2014; López-Gonzalez et al., 2015). In support of this 

evidence, some modern tradable services (e.g., IT-related services) have notably 

expanded in Asian countries (e.g., India, Philippines) and, as argued by Dasgupta and 

Singh (2005), may also lead to the expansion of manufacturing, rather than the other 

way around.  

 

Conversely, and expectedly, we have found empirical support to the argument put 

forward by MacMillan and Rodrik (2011), McMillan et al. (2014) and Rodrik (2016), 

namely, that a shift to low tech, personal and informal services as a result of a loss of 

industrial core might instead lead to a Kaldorian-like productivity and growth 

slowdown. In particular, Latin America and Africa have overall followed a different 

path of structural change and growth.  As argued by Dasgupta and Singh (2006), a 

pathological kind of deindustrialization has occurred within both regions during the 80s 

and the 90s as many countries specialized according to their static comparative 

advantage  –  namely in resource-based industries, simple processing and/or labour 

intensive products with little prospect for upgrade (Shafeaeddin, 2005). Long– run 

dynamic comparative advantages –that require the creation and diffusion of 

technological capabilities and innovations, and depend on strong linkages between firms 

and knowledge flows (Ocampo, 2005) – were instead disregarded. In particular, our 

results show that neither static nor dynamic productivity gains in manufacturing has 

been achieved in the case of Latin America. The deindustrialization experienced in this 

area has most likely hampered the emergence and development of advanced business-

related services (Di Meglio, 2017) and might currently represent a case of 

‘specialisation trap’.  

 

Overall, our findings show that the debate around ‘pre-mature deindustrialization’ in 

developing countries can be put in perspective, as the within- and across sector 

productivity performance of services is very heterogeneous. What can be inferred by the 

wealth of results discussed above is that there are different types of deindustrialization, 

not all of which represent a structural burden for (developing) economies. Namely 

business services might support structural transformation of core manufacturing bases, 



20 

 

as their productivity performance show similar dynamics across several macro areas of 

the world. Our findings show that the most dynamic sectors remain the manufacturing 

one and those that are tightly linked to these, such as business services. However, they 

also highlight that what matters, beyond the sectoral boundaries, is the ability to create 

value added, which is not necessarily linked to a critical mass of manufacturing, rather 

to its potential for technological upgrading, knowledge accumulation and increased 

tradability.  

 

The challenge for development policy is to induce and facilitate processes of structural 

transformation based on sectoral and technological upgrading, which might lead to a 

‘qualified’ premature deindustrialization. Although with the data at our disposal we 

cannot directly empirically support this, we might still claim that it is not the 

‘premature’ nature of deindustrialization that might represent a concern, but the 

‘direction’ of it. 

 

DIRECTIONS FOR A RESEARCH AGENDA 
 

A number of conceptual, methodological and empirical issues around the topic of pre-

mature deindustrialization remain at stake. Below we identify areas for a research 

agenda on services in developing countries, which might build upon the findings of this 

manuscript.  

 

When the Kaldorian framework emerged, there was more of a clear-cut distinction 

among sectors in an economy. At present, the distinction between many service and 

manufacturing industries is more debatable since their boundaries have become more 

blurred over time, and the manufacturing-service interface evolved, which makes the 

traditional sectoral classification unsustainable (Daniels and Bryson, 2002). Future 

research should focus on sectoral structural change from the perspective of the historical 

and geographical processes of knowledge dynamics (see among others, Ciarli et al., 

2012) that has occurred in developing countries. Relatedly, it would be important to go 

more in-depth into the micro-economic level and look at the extent of the co-production 

relationship across service and other sectors’ firms (Savona and Steinmueller, 2013; 

Gallego and Maroto, 2015). 

 



21 

 

A few methodological and empirical issues still need to be addressed and offer a 

promising future research agenda. First, the estimation of KGL remains an econometric 

challenge. There is room for further improvements on the use of additional variables, 

dynamic and non-linear techniques. In particular, more work is needed to reconcile all 

the specifications of the Verdoorn’s law (Romero, 2016). Indeed, an extensive debate in 

the literature has focused on the fact that the specification used by Kaldor does not 

control for the contribution of growth of capital stock. Second, assuming, as Kaldor did, 

that the KLG is based on a technical progress function, excluding this variable from 

estimations is likely to provide a biased coefficient of returns to scale, except if a 

constant capital/output ratio is assumed (McCombie, 1982). However, in the case of 

developing economies, it is difficult to find reliable and consistent data of capital stocks 

at the sectoral level, as also noted by Jaumotte and Spatafora (2007) and Wells and 

Thirlwall (2003). Lack of data makes it difficult to account for capital stock growth in 

the estimation of Kaldor´s Second Growth Law and severely limits international total 

factor productivity comparisons.  

 

A further issue related to the econometric estimation of the Verdoorn´s law is the 

possibility of simultaneous equation bias due to the potential endogeneity of the 

regressors. To overcome this, a variety of econometric techniques have been applied: 

simultaneous equation; instrumental variables and Granger causality tests. More 

recently, Felipe et al. (2009) propose using a semi-parametric technique. However, as 

discussed in McCombie et al. (2002), these procedures still suffer from limitations, and 

the controversy is yet not solved.  

 

The diffusion of technical progress poses an additional challenge in the empirical 

estimation of the Verdoorn law. If the former varies across countries, manufacturing 

productivity increase in ‘laggards’ countries may reflect the transfer of technology from 

leading countries, rather than indigenous innovation leading to domestic increasing 

returns of scale. To overcome this issue, the Verdoorn-related literature suggest the use 

of additional variables to account for the level of technological development and of 

cross-regional data. However, currently available data still do not allow undertaking 

such strategy.  

 



22 

 

The Kaldorian framework relies on traditional productivity indicators. However, the 

accurate measurement of productivity in services is still an unresolved matter (Djellal 

and Gallouj, 2008; see also Grassano and Savona, 2014, for a review). As summarized 

in Maroto and Rubalcaba (2008) and in Di Meglio, (2013), measuring output and input 

in services has not gone much further since Griliches (1992). This is a well-known issue 

particularly with regard to public services, where the measurement of productivity 

remains inadequate and flawed by data deficiency (Di Meglio et al., 2015). Data 

collection and empirical evidence should be preceded in this case by a substantial 

theoretical effort.  

 

ACKNOWLEDGEMENTS 

 

We gratefully acknowledge funding through Ramón Areces Foundation, Spain. We 

thank two anonymous reviewers for SPRU Working Papers Series, and four anonymous 

reviewers of Development and Change. Usual disclaimers apply. Maria Savona 

gratefully acknowledges the European Union's Horizon 2020 research and innovation 

programme under grant agreement No. 649186 (Project ISIGrowth). 

 

 

 

  



23 

 

REFERENCES 

 

Bah, E. (2011) ‘Structural Transformation Paths across Countries’, Emerging Markets 

Finance and Trade 47(2): 5–19.  

Ciarli, T., Meliciani, V. , Savona, M. (2012) ‘Knowledge Dynamics, Structural Change and 

The Geography of Business Services’, Journal of Economic Surveys 26(3):  445–67.  

Ciarli, T., Di Maio, M. (2014) ‘Theoretical arguments for industrialization‐driven growth 

and economic development’, in F. A. S. T. Matambalya (ed) African Industrial 

Development and European Union Co‐operation. Prospects for a reengineered 

partnership. Routledge. 

Cimoli, M., Katz, J. (2003) ‘Structural reforms, technological gaps and economic 

development: a Latin American perspective’, Industrial and Corporate Change 12(2): 

387–411.  

Cimoli, M., Porcile, G., Primmi, A., Vergara, S. (2005) ‘Cambio Estructural, 

Heterogeneidad Productiva y Tecnología en América Latina’, in M. Cimoli (ed) 

Heterogeneidad estructural, asimetrías tecnológicas y crecimiento en América Latina, 

CEPAL, Naciones Unidas. Available at: http://mpra.ub.uni-muenchen.de/3832/ 

Cimoli, M., Dosi, G., Stiglitz J.E. (2009) Industrial Policy and Development: The Political 

Economy of Capabilities Accumulation. Oxford University Press. 

Crespi, G., Tacsir, E., Vargas, F. (2014) ‘Innovation and Productivity in Services: 

Empirical Evidence from Latin America’. Inter-American Development Bank Working 

Paper Nº 6592, August.  

Daniels, P.W., Bryson, J.R. (2002) ‘Manufacturing services or servicing manufacturing? 

New forms of production in advanced capitalist economies’, Urban Studies 39: 977–91.  

Dasgupta, S., Singh A. (2005) ‘Will Services Be the New Engine of Indian Economic 

Growth?’, Development and Change 36(6): 1035–58. 

Dasgupta, S., Singh A. (2006) ‘Manufacturing, Services and Premature De-industrialization 

in developing countries: a Kaldorian empirical analysis’, CBR Working paper Nº 327, 

University of Cambridge, June. 

Di Meglio, G. (2013), ‘The place of ServPPINs in the range of public-private collaboration 

arrangements for services provision’, in: Gallouj, F., Windrum, P. and Rubalcaba, L. 

(eds.), Public Private Innovation Networks in Services. The dynamics of Cooperation in 

Service Innovation, , pp. 59-87, Edward Elgar Publishing. 



24 

 

Di Meglio G., Stare M., Maroto A., Rubalcaba L. (2015), ‘Public services performance: an 

extended framework and empirical assessment across the enlarged EU’, Environment 

and Planning C: Government and Policy, 33(2), 321-341. 

Di Meglio, G. (2017), ‘Services and Growth in Developing Countries: a Kaldorian 

Analysis’, in N. Beerepoot et al (eds): "Globalisation and Services-driven Economic 

Growth", Routledge, pp. 38-54.Dixon, R., Thirlwall, A. P. (1975) ‘A model of regional 

growth rate differences on Kaldorian lines’, Oxford Economic Papers 27: 201–14. 

Djellal, F., Gallouj, F.  (2008) Measuring and improving productivity in services: issues, 

strategies and challenges, Cheltenham, Edward Elgar. 

Fabricant, S. (1942) Employment in Manufacturing 1899–1939. NBER, New York.  

Felipe, J. (1998) ‘The Role of the Manufacturing Sector in Southeast Asian Development: 

A Test of Kaldor’s First Law’, Journal of Post Keynesian Economics 20(3): 463–485. 

Felipe, J., León-Ledesma, M., Lanzafame, M., Estrada, G. (2009) ‘Sectoral Engines of 

Growth in Developing Asia: Stylised Facts and Implications’, Malaysian Journal of 

Economic Studies 46(2): 107–33. 

Felipe, J., Mehta, A., Rhee, C. (2015). ‘Manufacturing matters, but it’s the jobs that count’, 

Asian Development Bank Economics Working Paper Series No. 420. Available at 

SSRN: https://ssrn.com/abstract=2558904. 

Felipe, J., Mehta, A. (2016) ‘Deindustrialization? A global perspective’, Economics Letters 

149: 148–151 

Gallego, J., Rubalcaba, L., Hipp, C. (2013) ‘Services and organisational innovation: the 

right mix for value creation’, Management Decision 51(6): 1117–34. 

Gallego, J., Maroto, A. (2015) ‘The Specialization in Knowledge-Intensive Business 

Services (KIBS) across Europe: Permanent Co-Localization to Debate’, Regional 

Studies 49(4): 644–64.  

Gallouj, F., Savona, M. (2008) ‘Innovation in Services: A Review of the Debate and a 

Research Agenda’, Journal of Evolutionary Economics 19(2): 149–72.  

Gereffi, G., Fernandez-Stark, K.(2010a) ‘The Offshore Services Value Chains. Developing 

Countries and the Crisis’, World Bank Policy Research Working Papers Nº 5262. 

Gereffi, G., Fernandez-Stark, K.(2010b) ‘The Offshore Services Value Chains’, Centre on 

Globalisation, Governance and Competitiveness.  

  



25 

 

Grassano, N., Savona, M. (2014) ‘Productivity in services twenty years on. A review of 

conceptual and measurement issues and a way forward’. SPRU Working Paper Nº 

2014-01.  

Griliches, Z. (1992) Output Measurement in the Service Sector, NBER, Studies in Income 

and Wealth, Chicago: University of Chicago Press.  

Guerrieri, P., Meliciani, V. (2005) ‘Technology and International Competitiveness: The 

Interdependence between Manufacturing and Producer Services’, Structural Change 

and Economic Dynamics 16 (4): 489–502.  

Hernández, R. A., Mulder, N., Fernandez-Stark, K., Sauvé, P., López Giral, D., Muñoz 

Navia, F. (2014) Latin America's Emergence in Global Services. A New Driver of 

Structural Change in the Region?, Economic Commission for Latin America and the 

Caribbeans. 

Hirschman, A. O. (1958) The strategy of economic development. New Have: Yale 

University Press. 

Jaumotte, F., Spatafora, N. (2007) ‘Asia Rising: A Sectoral Perspective’, IMF Working 

Paper Nº WP/07/130. 

Kaldor, N. (1966) Causes of the Slow Rate of Economic Growth of the United Kingdom. 

Cambridge University Press: Cambridge. 

Kaldor, N. (1967) Strategic Factors in Economic Development. New York State School of 

Industrial and Labour Relations, Cornell University: Ithaca NY. 

Kaldor, N. (1970) ‘The Case for Regional Policies’, Scottish Journal of Political Economy 

17: 337–48. 

León-Ledesma, M. (2000) ‘Economic Growth and Verdoorn's Law in the Spanish Regions, 

1962-91’, International Review of Applied Economics 14(1): 55-69. 

Lewis, W. A. (1954) ‘Economic Development with Unlimited Supplies of Labor’, The 

Manchester School, Vol. 22, pp. 139–191. 

Libanio, G. (2006) ‘Manufacturing Industry and Economic Growth in Latin  America:  A 

Kaldorian  Approach’, Policy Innovations Paper, s.4. 

López-Gonzalez, J.,  Meliciani, V., Savona, M. (2015) ‘When Linder Meets Hirschman. 

Inter-Industry Linkages and GVCs in Services’, SPRU Working Paper Series Nº 2015-

20.  

Matsuyama, K. (2009) ‘Structural Change in an Interdependent World: A Global View of 

Manufacturing Decline’, Journal of the European Economic Association 7(2-3): 478–

86. 



26 

 

Maroto, A., Rubalcaba, L. (2008) ‘Services productivity revisited’, The Service Industries 

Journal 28(3): 337–353. 

Maroto, A., Cuadrado, J.R. (2009) ‘Is growth of services an obstacle to productivity 

growth? A comparative analysis’, Structural Change and Economic Dynamics 20(4): 

254–265. 

McCombie, J.S.L. (1982) ‘Economic growth, Kaldor´s law and the static-dynamic 

Verdoorn law paradox’, Applied Economics 14: 279–94. 

McCombie, J., Pugno, M., Soro, B. (2002) Productivity Growth and Economic 

Performance: Essays on Verdoorn´s Law.  Palgrave-Macmillan: London.  

Meliciani, V., Savona, M. (2015) ‘The Determinants of Regional Specialisation in Business 

Services. Agglomeration Economies, Vertical Linkages and Innovation’, The Journal of 

Economic Geography 15:387– 416.. 

McMillan, M., Rodrik, D. (2011) ‘Globalization, Structural Change, and Productivity 

Growth’. NBER working paper Nº 17143. 

McMillan, M. S., Rodrik, D., Verduzco-Gallo, I. (2014) ‘Globalization, structural change 

and productivity growth: with an update on Africa’, World Development 63:11–32. 

Ocampo, J. A. (2005) Beyond Reforms: Structural Dynamics and Macroeconomics 

Vulnerability, ECLAC, Stanford University Press and the World Bank: Palo Alto, CA.  

Pacheco, P., Thirlwall, A. P. (2014) ‘A new interpretation of Kaldor's first growth law for 

open developing economies’, Review of Keynesian Economics 3(July): 384–98. 

Page, J. (2011) ‘Should Africa Industrialize?’, World Institute for Development Economic 

Research, Working Paper Series UNU-WIDER Research Paper  Nº 2011/47. 

Pagés, C. (2010) The Age of Productivity: Transforming Economies from the Bottom Up, 

Washington, USA:  Inter-American Development Bank. 

Palma, G. (2005) ‘Four Sources of ‘Deindustrialization’ and a New Concept of Dutch 

Disease, in J. Ocampo (ed) Beyond Reforms: Structural Dynamics and Macroeconomic 

Vulnerability, Stanford University Press and the World Bank: Palo Alto, CA. 

Peneder, M (2003) ‘Industrial Structure and aggregate growth’, Structural Change and 

Economic Dynamics 14: 427–48. 

Pieper, U. (2003) ‘Sectoral Regularities of Productivity Growth in Developing Countries: A 

Kaldorian Interpretation’, Cambridge Journal of Economics 27(6): 831–50. 

Rodrik, D. (2016) ‘Premature deindustrialization’, Journal of Economic Growth 21(1): 1-

33. 



27 

 

Romero, J. (2016) ‘Economic Growth from a Kaldorian Perspective: Theory, Evidence and 

Agenda’, Brazilian Keynesian Review, 2 (2), pp.189–210. 

Ros, J. (2000) Development Theory and the Economics of Growth, The University of 

Michigan Press. 

Savona, M., Steinmueller, W.E. (2013) ‘Service output, innovation and productivity: A 

time-based conceptual framework’, Structural Change and Economic Dynamics 

27:118–32.  

Shafaeddin, S.M. (2005). ‘Trade Liberalization and Economic Reform in Developing 

Countries: Structural Change or Deindustrialization?’. UNCTAD Discussion Papers 

179, UNCTAD: Geneva. 

Stiglitz, J.E., Lin J.Y. (eds.) (2013) The Industrial Policy Revolution I. The Role of 

Government Beyond Ideology. Basingstoke and New York: Palgrave Macmillan. 

Stiglitz, J.E. (2016) ‘Industrial policy, learning, and development’. World Institute for 

Development Economics Research, WIDER Working Paper 2016. 

Storm, S. (2015)’Structural change’, Development and Change 46(4): 666–99. 

Syrquin, M.(1984) ‘Resource allocation and productivity growth’, in Syrquin, M., Taylor, 

L., Westphal, L.E. (eds.), Economic Structure Performance — Essays in Honor of 

Hollis B. Chenery, pp. 75–101. Academic Press, Orlando, FL,  

Targetti, F. (2005) ‘Nicholas Kaldor: Key Contributions to Development Economics’, 

Development and Change 36(6): 1185–99. 

Timmer, M., Szirmai, A. (2000) ‘Productivity growth in Asian manufacturing: the 

structural bonus hypothesis examined’, Structural Change and Economic Dynamics 

11(4): 371–392.  

Thirlwall, A.P. (1983) ‘Symposium on Kaldor´s Laws’, Journal of Post Keynesian 

Economics, 5(3).  

Thirlwall, A.P. (2013) Economic Growth in an Open Developing Economy: The Role of 

Structure and Demand. Edward Elgar, Cheltenham.  

Timmer, M., de Vries, G. (2007) ‘A Cross-country Database For Sectoral Employment and 

Productivity in Asia and Latin America, 1950-2005’. Research Memorandum GD-98, 

Groningen Growth and Development Centre. 

Timmer, M., de Vries, G. (2009) ‘Structural Change and Growth Accelerations in Asia and 

Latin America: A New Sectoral Data Set’, Cliometrica 3(2): 165–90. 



28 

 

Tregenna, F. (2011) ‘Manufacturing Productivity, Deindustrialization, and 

Reindustrialization’, World Institute for Development Economic Research, Working 

Paper Series UNU-WIDER Research Paper Nº 2011/57 

United Nations (2010) ‘Structural Change and Employment Creation in Developing 

Countries’, in UN (ed), Trade and Development report, New York and Geneva, pp. 99-

131. 

Vries, K. de, R. Gouma, G. J. de Vries, M. P. Timmer (2013). ‘The Africa Sector Database: 

Sources and Methods’. Mimeo Groningen Growth and Development Centre. 

Wells, H., Thirlwall, A.P. (2003) ‘Testing Kaldor's Growth Laws across the Countries of 

Africa’, African Development Review 15: 89–105.  

Westkämper, E. (2014) Towards the Re-Industrialization of Europe: A Concept for 

Manufacturing for 2030. Springer.  

 



29 

 

Table 1. KGL in developing economies: a summary of the literature 

Article N T 

Equations estimated 
Estimations for 

aggregated sectors 

(agriculture, 

manufacturing, 

services)?  

Estimations 

for 

disaggregated 

services?  
1st law 2nd law (Verdoorn's law) 3rd law 

Felipe (1998) 5 Asian countries 1967-1992 qnm = a2 + b2(qm) + ε2 N.E. N.E. No No 

Pieper (2003) 
30 developing 

countries 
1970-1990 N.E. em = a1 + b1(qm) + ε1 N.E. Yes Yes 

Wells and 

Thirwall (2003)  

45 African 

countries 
1980-1996 

qGDP = a1+ b1(qm) + ε1 

qGDP = a2 + b2(qm - qnm) + ε2 

qnm = a3+ b3(qm) + ε3 

em = a1 + b1(qm) + ε1 p = a1 + b1(qm) + c1(enm) + ε1 Yes No 

Cimoli et al 

(2005) 

5 Latin American 

countries 
1970-2000 N.E. em = a1 + b1(qm) + ε1 N.E. No No 

Dasgupta and 

Singh (2005) 

30 developing 

economies 

1980, 1990, 

2000 
logqGDP = a1+ b1(logqm) + ε1 log(p) = a1 +b1(logqm) + c1(logenm) + ε1 Yes No 

Dasgupta and 

Singh (2006) 

48 developing 

countries 
1990-2000 qGDP = a1+ b1(qm) + ε1 p = a1 + b1(qm) + c1(enm) + ε1 Yes No 

Libanio (2006) 
7 Latin American 

countries 
1985-2001 

qGDP = a1+ b1(qm) + ε1 

qGDP = a2+ b2(qm - qnm) + ε2 

qnm = a3+ b3(qm) + ε3 

em = a1 +b1(qm) + c1(km) + ε1 N.E. No No 

Felipe et al 

(2009) 
17 Asian countries 1980-2004 lnqnm = a3+ b3(lnqm) + ε3 lnem = a1 + b1(lnqm) + ε1 

Decomposition of labour 

productivity growth 
Yes No 

Pacheco and 

Thirwall (2014) 

89 developing 

countries 
1990-2011 

qGDP = a1 + b1(qEXP) + ε1 

qEXP = a2 + b2(qm) + ε2 

qGDP = a1 + a2b1 +b1b2(qm) + ε2 

N.E. N.E. No No 

Note: N.E. for not estimated.



30 

 

Table 2. Description of sectoral composition of GGDC database 

j ISIC Rev. 3.1 
code  

Sector name  ISIC Rev. 3.1 description 

1 A-B  Agriculture  Agriculture, Hunting and Forestry, Fishing 

2 D  Manufacturing  Manufacturing 

3 G-P Services Trade services, Transport services, Business services, Public 
services 

4 G-H Trade services Wholesale and Retail trade; repair of motor vehicles, 
motorcycles and personal and household goods, Hotels and 
Restaurants  

5 I Transport services Transport, Storage and Communications  

6 J-K Business services Financial Intermediation, Renting and Business Activities 
(excluding owner occupied rents) 

7 L-P Public services Public Administration and Defence, Education, Health and 
Social work, Other Community, Social and Personal service 
activities, Activities of Private Households  

Source: GGDC 10-sector database. 
Note: Following McMillan and Rodrik (2011) and McMillan et al. (2014), we aggregate value added and 
employment data for the “Government Services” sector (L-N) and the “Personal Services” sector (O-P) 
into a single “Public Services” sector. 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 



31 

 

Table 3. Panel data estimation of Kaldor´s First Law: all developing countries 

SECTOR 
EQUATION I-A EQUATION I-B EQUATION I-C 

1j / s.e.  β1j / s.e. 2j / s.e.  β2j / s.e. 3j / s.e.  β3j / s.e. 

j=manufacturing 
 

.0125* 
.0053 

.4682*** 
.0249 

.0173 

.0094 
.1461* 
.0612 

.0160* 
.0065 

.3626*** 
.0329 

R2 0.781 R2 0.4687 R2 0.637 
j=agriculture 
 
 

.0107 

.0088 
.2371*** 

.0531 
.0170* 
.0068 

-.3027*** 
.0477 

.0124 

.0092 
.1343* 
.0601 

R2 0.516 R2 0.605 R2 0.522 

j=services  
.0008 
.0031 

.8411*** 
.0346 

.0161 

.0090 
-.1328 
.0769 

.0002 

.0070 
.7170** 

.0682 
R2 0.895 R2 0.461 R2 0.664 

j=trade  
 

.0101** 
.0033 

.5414*** 
.0412 

.0158 

.0093 
.0464 
.0573 

.0122** 
.0036 

.4671*** 
.0458 

R2 0.785 R2 0.451 R2 0.705 

j=transport and 
communications 

.0007 

.0059 
.4264*** 

.0493 
0150 
.0095 

.0220 

.0752 
.00001 
.0062 

.3934*** 
.0518 

R2 0.703 R2 0.448 R2 0.649 

j=business 
services 

.0066 

.0084 
.3373*** 

.0263 
0137 
.0090 

.1536*** 
.0415 

.0076 

.0082 
.2694*** 

.0237 
R2 0.727 R2 0.527 R2 0.653 

j=public 
services 

.0094 

.0083 
.3833*** 

.0844 
.0157* 
.0063 

-.370*** 
.0455 

.0112 

.0107 
.2650** 

.1012 
R2 0.515 R2 0.599 R2 0.420 

N 174 174 174 
Note: OLS estimations with  fixed effects and PCSE accounting for groupwise heteroskedasticity, cross sectional dependence and 
serial correlation. Dummy coefficients estimates are available upon request. Legend: s.e. for standar deviation; * p<0.05; ** p<0.01; 
*** p<0.001 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 



32 

 

Table 4. Panel data estimation of Kaldor´s Second Law: all developing countries  

SECTOR 
EQUATION II 

β0j / (s.e.)  β1j /(s.e.)  Ho: β1j = 1 / p-value  Ho: β1j <1 / p-value 
j=manufacturing -0.0096 

0.0115  
0.5819*** 

0.0566  Reject Ho 

( 0.0000 ) 
Retain Ho 

(1.0000) 
R2 0.620 

j=agriculture 
 

-0.0120 
0.0066 

0.1278 
0.0748   

R2 0.566 
j=services 0.0209*** 

0.0054 
0.2118*** 

0.0545 Reject Ho 

( 0.0000 ) 
Retain Ho 

(1.0000) 
R2 0.342 

j=trade  0.0224*** 
0.0046 

0.0028 
0.0564   

R2 0.400 
j=transport and 
communications 

0.0098 
0.0124 

0.3803*** 
0.0728 Reject Ho 

( 0.0000 ) 
Retain Ho 

(1.0000) 
R2 0.4072 

j=business services 0.0231 
0.0141 

0.3107*** 
0.0463 Reject Ho 

( 0.0000 ) 
Retain Ho 

(1.0000) 
R2 0.541 

j=public services 0.0195*** 
0.0054 

0.3470*** 
0.0858 Reject Ho 

( 0.0000 ) 
Retain Ho 

(1.0000) R2 0.385 
N 174 
Note. OLS estimations with  fixed effects and PCSE accounting for groupwise heteroskedasticity, cross sectional dependence and 
serial correlation. Dummy coefficients estimates are available upon request. Legend: s.e. for standar deviation; * p<0.05; ** p<0.01; 
*** p<0.001 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 



33 

 

Table 5. Productivity growth decomposition 1975-2005: all developing countries 

SECTOR 
Labour 
productivity 
growth 

Structural 
Effect 
(SCE = 
SSE + 
DSE) 

Static 
Structural 
Effect (SSE) 

Dynamic 
Structural 
Effect (DSE) 

Intra-
sectoral 
Effect (ISE) 

TOTAL 
+0.680 
(100%) 

+0.161 
(23.7%) 

+0.435 
(64.0%) 

-0.274 
(-40.3%) 

+0.519 
(76.3%) 

   = = = 
Agriculture (0.939) -0.127 -0.062 -0.065 +0.117 

Manufacturing (1.087) +0.005 +0.019 -0.014 +0.155 
Other industry (1.854) -0.081 +0.055 -0.136 +0.150 

Services (0.280) +0.363 +0.422 -0.059 +0.097 
      
Trade (0.068) +0.129 +0.171 -0.042 +0.004 

Transport and 
communications 

(1.040) 
+0.039 

+0.027 +0.012 +0.051 

Business services (0.191) +0.127 +0.149 -0.022 +0.003 

Public services (0.374) +0.068 +0.075 -0.007 +0.038 

Sectoral contributions to each effect 
 (adding the TOTAL by columns) 

Agriculture  -78,9% -14,3% 23,7% 22,5% 
Manufacturing  3,1% 4,4% 5,1% 29,9% 
Other industry  -50,3% 12,6% 49,6% 28,9% 

Services  225,5% 97,0% 21,5% 18,7% 

Subsectoral contributions to each effect 
(adding the SERVICES by columns) 

Trade  35,5% 40,5% 71,2% 4,1% 
Transport and 
communications 

 
10,7% 6,4% -20,3% 52,6% 

Business services  35,0% 35,3% 37,3% 3,1% 
Public services  18,7% 17,8% 11,9% 39,2% 

Note: The percentage contribution of each effect to aggregate productivity growth appears between brackets. ‘Other industry’ 
includes: mining and extracting activities, construction and energy. 

 

 

 

 

 

 

 

 

 

 

 



34 

 

APPENDIX A 

 

Table A. Panel data estimation of Kaldor´s First Law 

 

Table A.1. Asia 

SECTOR (j) 
 

EQUATION I-A EQUATION I-B EQUATION I-C 

1j / s.e.  β1j / s.e. 2j / s.e.  β2j / s.e. 3j / s.e.  β3j / s.e. 

Manufacturing 
.0276*** 

.0039 
.4626*** 

.0523 
.0710** 
.00611 

.2400* 
.1007 

.0537*** 
.0043 

.3617*** 
.0590 

R2 0.715 R2 0.292 R2 0.520 

Agriculture 
.0712*** 

.0080 
.1622* 
.0805 

.0431* 
   .0166 

-.1708 
.1019 

0719*** 
.0080 

.162 
.0848 

R2 0.357 R2 0.285 R2 0.305 

 
Services 

.0021  

.0038 
.9223*** 

.0416 
.0683 
.00 

-.244 
.1358 

-.0036 
.0137 

.9095*** 
.0864  

R2 0.920 R2 0.245 R2 0.747 

Trade 
.012* 
.0056 

.6755*** 
.0444 

.0570*** 
.009 

.3763 
.209 

.0134 
.00762 

.6102*** 
.0551 

R2 0.859 R2 0.351 R2 0.784 

Transport and 
communications 

.019* 

.0077 
.548*** 
.0977 

.0610*** 
.0054 

-.2399** 
.0919 

.02143* 
.0084 

.5108*** 
.107 

R2 0.623 R2 0.352 R2 0.571 

Business 
services 

.0404*** 
.0051 

.3097*** 
.0267   

.058*** 
.0070 

.2999*** 
.04819 

.0408*** 
.0064 

.2578*** 
.0274 

R2 0.750 R2 0.521 R2 0.649 

Public services 
.0362* 
.0150 

.5048** 
.1929 

-.5147*** 
.1132 

-.5147*** 
.1132 

.044*   
.01876 

.397 
.2341 

R2 0.362 R2 0.504 R2 0.277 
N 54 54 54 

 

Table A.2. Latin-America  

SECTOR (j) 
EQUATION I-A EQUATION I-B EQUATION I-C 

1j / s.e.  β1j / s.e. 2j / s.e.  β2j / s.e. 3j / s.e.  β3j / s.e. 

Manufacturing 
.0114** 
.0043 

.6353*** 
.0551 

.020* 

.0097 
.3848*** 

.1323 
.0148*** 

.0054 
.5564*** 

.0653 
R2 0.823 R2 0.422 R2 0.723 

Agriculture 
.0031 
.0090 

.612*** 
.2121 

.0190*** 
.0051 

-.6878*** 
.0857 

-.0014 
.0089 

.821*** 
.196 

R2 0.323 R2 0.659 R2 0.438 

Services  
-.0002 
.00288 

.8998*** 
.0454 

.0156 

.0083 
-.0420 
.1583 

-.0006 
.00690 

.7681*** 
.0918 

R2 0.916 R2 0.217 R2 0.6538 

Trade  
.0101** 
.0033 

.5420*** 
.0585 

.01674* 
.00833 

.1860 

.1124 
.01245** 

.0038 
.4503*** 

.0700 
R2 0.720 R2 0.239 R2 0.571 

Transport and 
communications 

-.0062 
.0046 

.6304*** 
.0509 

.01247 
.0098 

.1466 

.1278 
-.0004 
.0034 

.6007*** 
.0637 

R2 0.796 R2 0.207 R2 0.756 

Business 
services 

.0068 

.0083 
.3292*** 

.0452 
.0137 
.0083 

.1277* 
.0561 

.0088 

.0085 
.2349*** 

.0479 
R2 0.588 R2 0.381 R2 0.4017 

Public services 
.0055 
.0076 

.6264*** 
.1192 

.0157** 
.0051 

-.5235*** 
.0758 

.0069 

.0101 
.5352*** 

.1517 
R2 0.476 R2 0.522 R2 0.358 

N 54 54 54 

 

 



35 

 

Table A.3. Africa 

SECTOR (j) 
EQUATION I-A EQUATION I-B EQUATION I-C 

1j / s.e.  β1j / s.e. 2j / s.e.  β2j / s.e. 3j / s.e.  β3j / s.e. 

Manufacturing 
0.0435*** 

0.00798 
0.327*** 

0.0547 
0.0571*** 

0.00992 
0.1364* 
0.06506 

0.0479*** 
0.0084 

0.2583*** 
0.0567 

R2 0.626 R2 0.586 R2 0.575 

Agriculture 
0.06175*** 

0.0105 
0.2370*** 

0.0636 
0.0553 

0.01314 
-0.150*** 

0.0575 
0.0658*** 

0.0124 
0.0103 
0.0612 

R2 0.515 R2 0.397 R2 0.495 

Services  
0.0041 
0.0107 

0.7074*** 
0.0818 

0.04936*** 
0.0038 

-0.2532*** 
0.0769 

0.0098 
0.0166 

0.4857** 
0.1291 

R2 0.729 R2 0.542 R2 0.437 

Trade  
0.0128* 

0.0064664 
0.4545*** 

0.0474 
0.0629*** 

0.0110 
-0.00883 
0.0673 

0.024728* 
0.0126 

0.3944*** 
0.0711 

R2 0.793 R2 0.412 R2 0.625 

Transport and 
communications 

0.0313** 
0.0093 

0.2993*** 
0.0593 

0.0586 
0.0111 

0.1035 
0.09725 

0.0339** 
0.0098 

0.2641*** 
0.0608 

R2 0.640 R2 0.431 R2 0.597 

Business 
services 

0.03506*** 
0.00991 

0.4520*** 
0.0760 

0.05445 
0.00725 

-0.2164* 
0.1088 

0.03842*** 
0.0108 

0.3959*** 
0.0817 

R2 0.641 R2 0.558 R2 0.580 

Public services 
0.0429*** 

0.01063 
0.2259** 

0.0895 
0.0582*** 

0.0083 
-0.2625*** 

0.0693 
0.0501 
0.0126 

0.0907 
0.1033 

N 66 66 66 
Note. OLS estimations with  fixed effects and PCSE accounting for groupwise heteroskedasticity, cross sectional dependence and 

serial correlation. Dummy coefficients estimates are available on request. Legend: s.e. for standar deviation; * p<0.05; ** p<0.01; 

*** p<0.001. 

  



36 

 

APPENDIX B 

 

Table B. Panel data estimation of Kaldor´s Second Law 

 

Table B.1. Asia 

SECTOR (j) 
EQUATION II 

β0j / (s.e.)  β1j /(s.e.)  Ho: β1j = 1 / p-value  Ho: β1j <1 / p-value 

Manufacturing 
-.060*** 

.0082 
.695*** 

.070 Reject Ho 

(0.0000 ) 
Retain Ho 

( .99999159) 
R2 0.766 

Agriculture 
-.0307*** 

.0103 
.461*** 
.1101   Reject Ho 

(0.0000 ) 
Retain Ho 

(.9999995) 
R2 0.587 

Services  
.0252*** 

.0065 
.283*** 
.0749 Reject Ho 

( 0.0000 ) 
Retain Ho 
(1.0000) 

R2 0.333 

Trade  
.03134***    

.0083 
.1033   
.1013   

R2 0.091 

Transport and 
communications 

.0231   

.0124 
.2626**    
.0984   Reject Ho 

( 0.0000 ) 
Retain Ho 

(1.0000) 
R2 0.377 

Business services 
.0554*** 

.0108 
.2414*** 

.0491 Reject Ho 

( 0.0000 ) 
Retain Ho 

(1.0000) 
R2 0.419 

Public services 
.0319* 
.0107 

.1968 

.1168   
R2 0.453 

N 54   

 

Table B.2. Latin-America  

SECTOR (j) 
EQUATION II 

β0j / (s.e.)  β1j /(s.e.)  Ho: β1j = 1 / p-value  Ho: β1j <1 / p-value 

Manufacturing 
-0.0082   
 0.0093 

0.2992   
0.1964   

R2 0.427 

Agriculture 
-0.0073 
0.0078 

-0.1025 
0.2102   

R2   0.118 

Services  
0.0245*** 

0.0043   
0.0033 
0.0930     

R2 0.360 

Trade  
0.0223    
0.0049 

0.0185  
0.1037   

R2 0.389 

Transport and 
communications 

0.0161   
0.01348 

0.1964 
0.1467   

R2 0.075 

Business services 
0261 
.0151 

.1772* 
.0825 Reject Ho 

( 0.0000 ) 
Retain Ho 

(1.0000) 
R2 0.580 

Public services 
0.0207*** 

0.0056 
0.2691 
0.1624   

R2 0.315 
N 54   

 

 



37 

 

Table B.3. Africa 

SECTOR (j) 
EQUATION II 

β0j / (s.e.)  β1j /(s.e.)  Ho: β1j = 1 / p-value  Ho: β1j <1 / p-value 

Manufacturing 
0.0301 
0.0228 

0.718*** 
.1158 Reject Ho 

0.0152 
Retain Ho 

.99242332 
R2 0.613 

Agriculture 
0.0116    
0.0065 

-0.110** 
0.0400   

R2 0.559 

Services  
0.0281** 

0.0105 
0.30794** 

0.0949 Reject Ho 

( 0.0000 ) 
Retain Ho 

(1.0000) 
R2 0.378 

Trade  
0.0805*** 

0.0172 
0.1104 
0.109   

R2 0.305 

Transport and 
communications 

-0.0182 
0.0211 

0.666*** 
0.0907 Reject Ho 

( 0.0000 ) 
Retain Ho 
.99988027 

R2 0.556 

Business services 
0.0143  
0.0120 

0.8220***  
0.1009 Retain Ho 

(0.0779)  
R2 0.623 

Public services 
0.0031 
 0.0154 

0.3958** 
0.1180 Reject Ho 

0.0000 
Retain Ho 

.99999984 R2 0.438 
N 66 
 

Note. OLS estimations with  fixed effects and PCSE accounting for groupwise heteroskedasticity, cross sectional dependence and 

serial correlation. Dummy coefficients estimates are available on request. Legend: s.e. for standar deviation; * p<0.05; ** p<0.01; 

*** p<0.001 

 

  



38 

 

APPENDIX C 

 

Table C. Productivity growth decomposition: Per cent contribution of each effect. 

 

Table C.1. Asia 

SECTOR 
Labour 
productivity 
growth 

Structural 
Effect 
(SCE = 
SSE + 
DSE) 

Static 
Structural 
Effect (SSE) 

Dynamic 
Structural 
Effect (DSE) 

Intra-
sectoral 
Effect (ISE) 

TOTAL +1.655 
0.289 

(17.5%) 
+0.479 

(28.9%) 
-0.190 

(-11.5%) 
+1.366 

(82.5%) 

      
Agriculture (1.338) -0.223 -0.084 -0.139 +0.225 
Manufacturing (2.917) 0.044 +0.035 +0.009 +0.414 

Other industry (3.991) -0.149 +0.085 -0.234 +0.290 
Services (1.027) 0.617 +0.443 +0.174 +0.437 

      
Trade (0.992) 0.194 +0.107 +0.087 +0.162 
Transport & 
communications 

(2.039) 
0.073 

+0.038 +0.035 +0.106 

Business services (0.758) 0.239 +0.211 +0.028 +0.038 
Public services (0.849) 0.111 +0.087 +0.024 +0.131 

Sectoral contributions to each effect 
 (adding the TOTAL by columns) 

Agriculture  -77.2 -17.5 73.2 16.5 

Manufacturing  15.2 7.3 -4.7 30.3 
Other industry  -51.6 17.7 123.2 21.2 
Services  213.5 92.5 -91.6 32.0 

Subsectoral contributions to each effect 
(adding the SERVICES by columns) 

Trade  31.4 24.2 50.0 37.1 

Transport & 
communications 

 
11.8 8.6 20.1 24.3 

Business services  38.7 47.6 16.1 8.7 
Public services  18.0 19.6 13.8 30.0 

 

 

 

 

 

 

 

 

 

 



39 

 

 

Table C.2. Latin-America 

SECTOR 
Labour 
productivity 
growth 

Structural 
Effect 
(SCE = 
SSE + 
DSE) 

Static 
Structural 
Effect (SSE) 

Dynamic 
Structural 
Effect (DSE) 

Intra-
sectoral 
Effect (ISE) 

TOTAL +0.008 
0.006 
(75%) 

+0.338 
(4225%) 

-0.332 
(-4150%) 

+0.002 
(25%) 

      
Agriculture (1.180) -0.085 -0.040 -0.045 +0.085 
Manufacturing (0.226) -0.057 -0.035 -0.022 +0.046 

Other industry (0.585) -0.035 +0.025 -0.060 +0.027 
Services (-0.310) 0.183 +0.388 -0.205 -0.156 
      

Trade (-0.473) 0.072 +0.167 -0.095 -0.089 
Transport & 
communications 

(0.414) 
0.022 

+0.022 +0.000 +0.019 

Business services (-0.385) 0.065 +0.163 -0.098 -0.033 
Public services (-0.219) 0.024 +0.036 -0.012 -0.053 

Sectoral contributions to each effect 
 (adding the TOTAL by columns) 

Agriculture  -1416.7 -11.8 13.6 4250.0 
Manufacturing  -950.0 -10.4 6.6 2300.0 

Other industry  -583.3 7.4 18.1 1350.0 
Services  3050.0 114.8 61.7 -7800.0 

Subsectoral contributions to each effect 
(adding the SERVICES by columns) 

Trade  39.3 43.0 46.3 57.1 
Transport & 
communications 

 
12.0 5.7 0.0 -12.2 

Business services  35.5 42.0 47.8 21.2 
Public services  13.1 9.3 5.9 34.0 

 

 

 

 

 

 

 

 

 

 

 

 

 

 



40 

 

 

Table C.3. Africa 

SECTOR 
Labour 
productivity 
growth 

Structural 
Effect 
(SCE = 
SSE + 
DSE) 

Static 
Structural 
Effect (SSE) 

Dynamic 
Structural 
Effect (DSE) 

Intra-
sectoral 
Effect (ISE) 

TOTAL 
+0.407 

0.184 
(45.2%) 

+0.483 
(118.7%) 

-0.299 
(-73.5%) 

+0.222 
(54.5%) 

      
Agriculture (0.299) -0.071 -0.061 -0.010 +0.042 
Manufacturing (0.119) 0.027 +0.057 -0.030 +0.005 

Other industry (0.987) -0.059 +0.056 -0.115 +0.166 
Services (0.122) 0.287 +0.431 -0.144 +0.009 
      

Trade (-0.315) 0.112 +0.239 -0.119 -0.060 
Transport & 
communications 

(0.667) 
0.023 

+0.021 +0.002 +0.029 

Business services (0.200) 0.080 +0.074 +0.006 +0.005 
Public services (0.492) 0.064 +0.097 -0.033 +0.035 

Sectoral contributions to each effect 
 (adding the TOTAL by columns) 

Agriculture  -38.6 -12.6 3.3 18.9 
Manufacturing  14.7 11.8 10.0 2.3 

Other industry  -32.1 11.6 38.5 74.8 
Services  156.0 89.2 48.2 4.1 

Subsectoral contributions to each effect 
(adding the SERVICES by columns) 

Trade  39.0 53.6 82.6 -666.7 
Transport & 
communications 

 
8.0 4.9 -1.4 322.2 

Business services  27.9 17.2 -4.2 55.6 
Public services  22.3 22.5 22.9 388.9