The deep determinants of economic development in China—a provincial perspective

Abstract

There is a significant body of literature arguing that institutional quality is the key for long run economic growth and development. While the majority of these studies are based on cross-country growth regressions, we focus on the institution-economic growth nexus within a particular country, namely China. China is often regarded as an exception by having achieved miraculous growth for more than three decades despite relatively low institutional quality. Nonetheless, our findings suggest that at the provincial level, institutional quality played in fact an important role for the economic success of a province in China, even more important than geography and integration. However, when simultaneously examining the relationship between institutions, human capital, and economic development, we find that human capital ‘trumps’ everything else; however institutional quality has a highly significant indirect effect on provincial per capita income by improving human capital. We employ, among others, instrumental variable estimation techniques to address endogeneity problems.

Keywords:

• Economic growth and development
• deep determinants of economic growth
• institutional quality
• human capital
• the Chinese economy

JEL Classification:

• O11
• O43
• O53

Acknowledgements

We thank David Weil, Guanghua Wan, and Leong Liew as well as the participants of the Workshop ‘China’s Development Experience and Enlightenment’ (Chongqing, 2018), the Asian Meeting of the Econometric Society (Seoul, 2018), the China Meeting of the Econometric Society (Shanghai, 2018), the 25th Conference of the Eurasia Business and Economics Society (Berlin, 2018), and the 2nd Center for East Asia Macroeconomic Studies Workshop on ‘East Asia Macroeconomic Studies’ (Xiamen, 2018) for helpful comments.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Linda Glawe is a postdoctoral researcher at the University of Hagen (Department of Macroeconomics). She holds a M.Sc. in Economics (with honors) and doctoral degree (summa cum laude) from the University of Hagen and is a member of the Center for East Asia Macroeconomic Studies. Her research interests include economic growth and development, institutions, the European integration process, structural change, and East Asian economics.

Helmut Wagner is currently Chair and Professor of Economics at the University of Hagen (since 1995), Director at the Hagen Institute for Management Studies (since 2000) and Director of the Center for East Asia Macroeconomic Studies (since 2016). He has also held visiting positions at Massachusetts Institute of Technology (MIT), Princeton University, University of California, John Hopkins University (AICGS), Harvard University, International Monetary Fund (IMF).

Notes

1 There are some other studies that analyze institutional quality within China. Examples include Ji et al. (2014), Ang et al. (2014), and Zhou (2014). However, the latter two primarily focus on the impact of institutional quality on the firms’ R&D activity. Ji et al. (2014) analyze institutional quality within China in the context of the interplay between resource abundance, institutions, and economic growth in China with a focus on natural resources.

2 The marketization index is only available until the year 2007.

3 The government efficiency index consists of four indicators, namely “government public services”, “public infrastructure”, “government size”, and “welfare of residents”, as well as various sub-indicators and 47 indexes. However, since it also includes various educational measures that might be correlated with our human capital variable, we cannot use this index to investigate the institutions-human capital development relationship.

4 The BFE index captures how encouraging regulations are to business activities. Following Zhou (2014, 76), we calculated the BFE index as the simple average of a city’s percentile rankings on each of the four areas of business regulation and their enforcement reported in the World Bank report, namely (i) starting a business, (ii) registering property, (iii) getting credit—creating and registering collateral, and (iv) enforcing contracts (World Bank 2008, 1). The underlying data we used to construct this index is only available at the city-level (in particular for the capital cities of the 30 Chinese provinces). We assume that the data for the capital city of a province is representative for the whole province. However, one should keep in mind that the business environment is probably more encouraging in the capital city of a province than in the peripheral regions.

5 It is difficult to directly compare the institutional indicators for Chinese provinces with general, country-level institutional indicators due to methodological issues and also data availability. However, if we for instance focus on the six Worldwide Governance Indicators (WGIs) devised by Kaufmann et al. (2010) which are widely used in the academic literature, we can state that the marketization index and the BFE index are most closely related to the WGI indicators ‘regulatory quality’ and ‘rule of law’, whereas the government efficiency index resembles the WGI indicator ‘government effectiveness’.

6 We use the most current version, namely NBS (2017).

7 We use the average years of schooling in 2009 instead of 2010 since the data for 2010 is inconsistent.

8 The great-circle distance (in km) ($D$) between two provinces can be obtained via the following formula: $D=\mathit{arccos }\left(\hspace{0.17em}\text{sin}\hspace{0.17em}\left({\varphi }_1\right)*\hspace{0.17em}\text{sin}\hspace{0.17em}\left({\varphi }_2\right)+\hspace{0.17em}\text{cos}\hspace{0.17em}\left({\varphi }_1\right)*\hspace{0.17em}\text{cos}\hspace{0.17em}\left({\varphi }_2\right)*\hspace{0.17em}\text{cos}\hspace{0.17em}\left({\lambda }_1-{\lambda }_2\right)\right)*{\mathrm{R}}_{\oplus },$ where ${\varphi }_1$(${\varphi }_2$) is the latitude of province 1 (province 2) and ${\lambda }_1$ (${\lambda }_2$) is the longitude of province 1 (province 2). ${\mathrm{R}}_{\oplus }$ denotes the earth radius (approximately 6.371 km). Using an alternative formula developed by Thaddeus Vincenty which is not based on a sphere but on an ellipsoid (Vincenty 1975) does not affect our results; the significance levels stay the same and even the coefficients are almost identical compared to those obtained via the great-circle distance formula.

9 Data on the student-teacher ratio for China’s provinces is not available prior to 1994.

10 Examples include the studies of Easterly (2007), Becker and Woessmann (2009), Becker et al. (2010), and Naritomi et al. (2012). Moreover, several of the studies listed above use the distance to a city or a country as instruments (For example, Becker and Woessmann 2009 use the distance to Wittenberg as an instrument for Protestantism.).

11 We empirically tested instrument validity by performing the Hansen J test using Lewbel’s (2012) constructed instruments. Moreover, as suggested by Baum (2008), we tested the exclusive restriction by including the instruments as regressors; however, they are statistically insignificant. As another robustness check, we also included the distance to the coast as an additional regressor in our core specification; however, our main results remain unchanged. Regarding the first-stage relationships, the distance to the coast is not significantly correlated with trade or institutions.

12 The standardized variable $x^{*}$ is obtained by using the following formula: $x^{*}=\frac{x-\mu }{\mathit{sd}},$ where $x$ denotes the original variable and $\mu$ ($\mathit{sd}$) is the mean (standard deviation) of $x.$

13 Scatter plots of the bivariate correlation between our alternative institutional measures and provincial per capita income is provided in the Appendix B, Figure B1.

14 The corresponding scatter plots for our two alternative institutional measures are provided in Appendix B, Figure B2.

15 Using the average of the (log) trade share in GDP and average of the marketization index for the years 2001 to 2007 does not change our results regarding the primacy of institutions.

16 We derive very similar results when using our alternative measures of institutional quality. The only difference is that when using the government efficiency index, latitude has no direct effect on per capita income but only an indirect effect via institutional quality. When using the BFE index, latitude has no significant effect on the provincial per capita income government (neither direct nor indirect).The results are not presented here to save space but are available upon request.

17 It has to be noted that in the strict sense, this “rule of thumb” only applies to the single endogenous regressor case (as in Columns 1-4 of Table 3). However, as argued by Rodrik et al. (2004), F-statistics far exceeding the threshold (as in our case) are nonetheless a good sign that our results do not suffer from weak instruments.

18 There are various (empirical and theoretical) articles that find a positive effect of institutional quality on the trading activity. Examples include the studies of Anderson and Marcouiller (2002), Levchenko (2007), and Francois and Manchin (2013). For example, Levchenko (2007) constructs a model of international trade in which institutional differences are modelled within the framework of incomplete contracts. He shows that institutional differences as a source of comparative advantage imply, among others, that the less developed country may not gain from trade, and factor prices may actually diverge as a result of trade. In the empirical part of the paper, Levchenko (2007) finds that institutional differences are an important determinant of trade flows. In addition, Anderson and Marcouiller (2002) find that corruption and imperfect contract enforcement reduce international trade. There are also studies arguing that there exist rather indirect positive effects of institutions on trade, for example via the investment channel (e.g. Knack and Keefer 1995) or through productivity (e.g. Hall and Jones 1999). Apart from that, there is also a branch of the literature examining the impact of trade on institutions. One of the arguments is that trade openness resulting from a more efficient resource allocation increases per capita income, the communication of ideas and, thus, the demand for institutional quality (see Lipset 1959). Empirical evidence is provided, among others, by Ades and Di Tella (1999), Wei (2000), Rodrik et al. (2004), and Rigobon and Rodrik (2005).

19 Note that we follow Rodrik et al. (2004) by omitting the feedback effect from per capita income to institutions and integration.

20 In particular, we obtain these values by solving the system of equations implied by Column (6) of Table 3, Panel A, and by Columns (1) and (2) of Table 4, Panel B.

21 Again, using the average of the marketization index for the years 2001 to 2007 does not change our results.

22 The results obtained using the BFE index are mostly consistent with the corresponding estimation results for the marketization index (however, the F-statistic of the institutional instrumental variable drops below the critical threshold of ten). As already mentioned above, we cannot perform robustness checks using the government efficiency index as alternative measure of institutional quality since it contains various indicators that might be correlated with our human capital variable.

23 Tebaldi and Elmslie (2008) and Dias and Tebaldi (2012) provide a theoretical framework for the interconnection between institutional quality and human capital. Cross-country empirical studies are provided among others, by Glaeser et al. (2004) and Coe et al. (2009).

24 There are also various studies arguing that human capital (accumulation) and trade have a positive effect on each other. However, this discussion is beyond the scope of this paper. Empirical evidence is provided by the studies of Gould and Ruffin (1995), Stokey (1991, 1996), and Hanson and Harrison (1999). In addition, Owen (1999) develops a theoretical model that shows that trade changes the incentives to accumulate human capital and also the ability of individuals to purchase education by altering the distribution of income.

25 We obtain these values by solving the system of equations implied by Column (4) of Table 6, Panel A, and by Columns (1) and (2) of Table 7, Panel B.

26 This may be partly due to the fact that within our cross-sectional framework, we treat physical capital as exogenous (as most other cross-country analyses do) (and also due to the absence of an adequate instrument); in a panel data setting, difference GMM estimation allows us to control for endogeneity of all explanatory variables. However, even the OLS panel estimates where we do not control for endogeneity (just as in the cross-sectional framework), show that physical capital is also insignificant (see Table 10, Panel A).

27 In the panel data framework, we compute the average years of schooling of the year 2010 by taking the mean value of the average school years of 2009 and of 2011 since, as already mentioned in Section 2, the data for 2010 is inconsistent. However, within the cross-sectional framework, we prefer to not take a computed mean but original data for the year 2009.

28 Regarding the educational quality, the PISA rankings indicate that China performs very well, emerging as the top performer in all categories (math, science, and writing) in 2012 (OECD 2014, 5). However, these results should be treated with caution as only Shanghai-China is covered in the PISA 2012 study. Shanghai’s share in the country’s total population is rather small (ca. 1.78%) and it is one of the most developed provinces in China (NBS 2017, own calculations). Thus, the performance of Shanghai’s student needs not to be representative for the whole country. This concern is further confirmed by the newest PISA results: In 2015, when the PISA test participation was extended to additionally include Beijing, Jiangsu, and Guandong (“B-S-J-G China”), accounting for roughly 17.07% of the country’s total population, China was barely able to retain its top 10 ranking in the average PISA score of reading, science, and math (OECD 2018). It would be very interesting to see how including more (less developed) provinces would affect the PISA test outcome.