In this section, we discuss our main results of inter-sectoral linkages.
The primary empirical results of long-run Granger causality by using value- added growth and labor productivity growth are presented in Tables 5.3 and 5.4, respectively. As we noted before, we estimate the dynamic model of equation (3) by using oLS with annual data. We include the lags of each variable up to the sixth order into the oLS regression, as well as the country- specific intercepts. Empirical results of other model specifications and estimation techniques are discussed in the robustness check section.
Conclusions based on panel Granger causality analysis are threefold.
First, asymmetric Granger-causality relations among different sectors in the middle-income stage do exist. In Table 5.3, where we investigate the inter- sectoral linkages by using value-added growth data, the significantly positive signs of manufacturing industries in relation to other sectors indicate that there are substantial effects running from manufacturing growth to all other three sectors’ development: if a country’s manufacturing sector continues to grow fast, its services sector growth rate, as well as the agriculture and non-manufacturing sector growth rate, will be higher. This finding shows that the manufacturing sector is of great importance to a country’s development.
Table 5.3: Long-Run Granger Causality Test: Value-Added Growth with Relative Criterion
Agriculture Non-manufacturing
Industry Manufacturing Services Short-run effect
Agriculture / 0.071 0.149** 0.178**
Non-manufacturing industry 0.007* / 0.551*** 0.616***
Manufacturing 0.006 0.040 / –0.036
Services 0.069 0.064** 0.188*** /
Long-run effect
Agriculture / 0.043 0.090** 0.109*
Non-manufacturing industry 0.148 / 0.413*** 0.457***
Manufacturing 0.007 0.045 / –0.041**
Services 0.089 0.079*** 0.218*** /
Note: ***, **, * denote statistical significance at 1%, 5%, and 10%, respectively.
Source: Authors. See section 5.2.1.
Table 5.4: Long-Run Granger Causality Test:
Labor Productivity Growth with Relative Criterion
Agriculture Non-manufacturing
Industry Manufacturing Services Short-run effect
Agriculture / –0.290 0.127 0.017
Non-manufacturing industry –0.211 / 0.658** –0.657
Manufacturing 0.137 –0.095 / 0.174
Services –0.00004 –0.250 0.670*** /
Long-run effect
Agriculture / 0.043 0.090** 0.011***
Non-manufacturing industry 0.148 / 0.413*** –0.419
Manufacturing 0.007 0.045 / 0.677
Services 0.089 0.079*** 0.218*** /
Note: ***, **, * denote statistical significance at 1%, 5%, and 10%, respectively.
Source: Authors. See section 5.2.1.
Moreover, due to small persistence in industry value-added growth rate, the long-term effect is slightly larger than the short-run effect. Also, when we switch to labor productivity data, we reach similar conclusions except that in the short run, we cannot observe significant effects of manufacturing growth in the agriculture sector. However, there exist major and positive effects after considering the persistence of agricultural labor productivity growth, which means that the development of manufacturing industry labor productivity is likely to benefit the agriculture sector in the long run. All these findings indicate that manufacturing sector development is central to the whole economy during the middle-income stage.
Second, such pulling effects are not only statistically significant but also economically significant. In the short run, a 1% level increase in the value- added growth rate of the manufacturing sector is likely to lead to a 0.149%
increase in the growth rate of the agriculture sector, a 0.551% increase in the non-manufacturing industry, and a 0.188% increase in the services sector.
Also, in the long run, a 1% difference in value-added growth rate in the manufacturing sector will generate a gap of 0.090% in the agriculture sector, a 0.413% gap in non-manufacturing industry, and a 0.188% gap in services. Such magnitudes are quite substantial. When we use labor productivity growth data, the empirical results in Table 5.4 also prove that the manufacturing sector has strong spillovers to and externalities for other sectors during the middle-income stage.
Third, the characteristics of all the other industries, including services, are different from those of manufacturing. Agriculture and non-manufacturing industry fail to contribute to the development of other sectors, which is in fact not a very surprising outcome. As for services, although it is likely to Granger cause the development of agriculture and non-manufacturing industry, it cannot provide significantly positive contributions to the development of the manufacturing sector. In the short run, there is no significant effect from services to manufacturing. However, in the long run, a 1% increase in the value-added growth rate of the services sector is likely to Granger cause a 0.041% decrease in the manufacturing growth rate. We obtain a similar
conclusion when using the labor productivity growth rate as the dependent variable, but we are on the fence about the generality of this finding because we find that when we change our model specifications, the result changes.
It turns out that the services sector can make a significantly positive, a significantly negative, or an insignificant contribution to the growth of the manufacturing sector. Therefore, we cannot draw any general conclusion from this baseline regression. However, the positive externalities of manufacturing to all other sectors are quite robust across different modifications in
econometric methods.
All these findings indicate that the links between sectors may depend on the development stage. For developed economies, the services sector may contribute a substantial proportion to the growth of the total economy.
But for middle-income economies manufacturing remains the key engine of growth. on the one hand, for middle-income economies, the tradable manufacturing sector is viewed as the primary channel through which a developing economy absorbs the best practices from advanced economies (Hausmann, Hwang, and Rodrik 2007; Jones and olken 2005). Thus, it plays a significant role in adopting state-of-the-art technology from developed economies and diffusing the technology and knowledge across sectors.
on the other hand, as industrialization progresses, the manufacturing sector increasingly stimulates demand for service inputs (Park and Chan 1989), thus further promoting the development of the services industry.
our results are also consistent with many other papers that focus on the inter-sectoral linkages between the manufacturing sector and the service sector. According to Kaldor (1957) and many references therein (Naudé and Szirmai 2012; Szirmai 2012), manufacturing picks up services because the manufacturing sector has several important qualities that are not shared by other sectors. Therefore, they conclude that spillover effects are stronger in the manufacturing sector than in any other sectors. Guerrieri and Meliciani (2005) argued that the emergence of modern service activities depends on the improvement in the manufacturing structure. Park and Chan (1989) found that manufacturing development is important to both
nonmarket services and market services because it provides an increased demand for those intermediate inputs. Based on this, they conclude that the development of the services sector depends more on that of manufacturing than vice versa. Szirmai and Verspagen (2015) proposed the same idea regarding the relation between these two important modern sectors.