Term limits and rotation of Chinese governors: do they matter to economic growth?

Abstract

Since the 1980s, the economic system in China has evolved into one that combines fiscal decentralization with political centralization. Local officials are successfully motivated to foster economic growth. One of the motivation and control mechanisms established in China's political and bureaucratic hierarchy is an introduction of term limits and a rotation system for higher level government officials. Using a panel data covering detailed information for all the provincial governors between 1978 and 2004, this paper finds a positive impact of both term limits and rotation of governors across provinces on local economic growth. It also finds that term limits and economic growth exhibit a weakly inverted U relationship. Although the rotation of governors matters to local economic growth, regional variation is observed, and the positive impact of rotated governors on local growth turned out to be more obvious in the eastern provinces than in the other provinces.

Keywords:

• term limits
• rotation
• economic growth
• China

JEL Classification:

• H83
• P26

Acknowledgements

The authors would like to thank the financial supports both from a national grant named ‘985 project’, which contains a sub-project on financial deepening, decentralization and regional economic growth in China, and from Chinese Ministry of Education Grant for national key research projects (grant #05JJD790076). We are also grateful to two referees for insightful comments. Any errors remaining are of course the authors'.

Notes

∗Significantly different from 0 at the 10% level

∗∗Significantly different from 0 at the 5% level

∗∗∗Significantly different from 0 at the 1% level.

∗Significantly different from 0 at the 10% level

∗∗Significantly different from 0 at the 5% level

∗∗∗Significantly different from 0 at the 1% level.

∗Significantly different from 0 at the 10% level

∗∗Significantly different from 0 at the 5% level

∗∗∗Significantly different from 0 at the 1% level

∗Significantly different from 0 at the 10% level

∗∗Significantly different from 0 at the 5% level

∗∗∗Significantly different from 0 at the 1% level

∗Significantly different from 0 at the 10% level

∗∗Significantly different from 0 at the 5% level

∗∗∗Significantly different from 0 at the 1% level

∗∗Significantly different from 0 at the 5% level

∗∗∗Significantly different from 0 at the 1% level

1. Huang (2002) provides an institutional rationale for the control mechanisms of both appointments and rotation of officials in China.

2. We use résumés of governors in our sample to identify and record whether he or she was rotated governor.

3. The vertical axis measures the difference between the growth rate of GDP in each province and national growth rate of GDP. By such a method, the annual macro effect of the business cycle is controlled for.

4. Note that Hainan and Chongqing are new provinces established in 1988 and 1997 respectively, therefore observations for both provinces are fewer than other provinces. We also exclude some observations because some party secretaries and governors began to serve after 2003. In short, our whole sample contains 30 provinces and covers 26 years, and every province has one party secretary and one governor, so the number of observations by year is 30 × 26 × 2 = 1560. After deleting some years for Hainan and Chongqing, and excluding some governors who served after 2003, the total number of observations is actually 1413.

5. It is because most datasets contain only hundreds of observations, if we control for fixed effect for every province, the degree of freedom in regression will decrease sharply. If the observations are not so many, it will influence the significance of its results to some extent.

6. The eastern region includes Beijing, Tianjin, Hebei, Liaoning, Shanghai, Shandong, Jiangsu, Zhejiang, Fujian, Guangdong, Hainan; the Central region includes Henan, Shanxi, Anhui, Jiangxi, Heilongjiang, Jilin, Hubei, Hunan; the Western region includes Inner Mongolia, Guangxi, Sichuan, Chongqing, Guizhou, Yunnan, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang.

7. As discussed in the second section, it is mainly related to Chinese politics and is rather complex. If the Party has a preference for higher ability governors to be rotated, it means the rotated governors' abilities will be higher than those who remain. Consequently, in our regression, the rotation variable should measure the governors' abilities. Logically, governors who are rotated from the eastern to the central-western area should have a significant influence on economic growth of both central and western areas. However, there is no evidence in our regressions that could support this argument.

8. Li and Zhou (2005) provide evidence that governors with better performance are more likely to be promoted, therefore it may actually reduce the number of observations for outstanding governors in our sample.

9. Acemoglu et al. (2001) run robustness tests by adding variables that probably influence the explained variable, and then testing whether coefficients of the core explanatory variables will vary significantly. If these coefficients do not vary so much, those variables newly added do not have a direct effect on the explained variable.

10. We notice that Jones and Olken (2005) explore the relationship between governors' turnover and economic growth without controlling abilities, and they also treat governors' age as an exogenous variable unrelated to the error term.

11. Generally speaking, the effect (coefficients (β₂)) of governors' earlier economic growth on economic growth later on is positive. In western democratic countries, voters hope to reselect governors with better performance and better abilities, see Banks and Sundaram (1998) for details. In China, however, governors with better performance are more likely to be promoted to other positions, and governors whose performance is below the average are most likely to remain in the same positions. In our sample there are 134 cases where governors were promoted, including 62 cases in which promotion was from governor to party secretary of the same province. Therefore, the link between earlier economic performance and tenure is uncertain. However, our empirical results show that the relationship seems to be negative with δ < 0. According to econometrics, E(β₁) = β₁ + β₂ δ if β₂ δ < 0 then E(β₁) < β₁, and downward bias exists.

12. Please note that in Table 5 we only reported IV results. In the first stage we run the regression by using instrumental variable and other exogenous variables as explanatory variables and treating tenure as an explained variable. When we calculate the real effect of rotation on economic growth, we should combine the effects of these two stages. For example, in column 1 of Table 5, during the first stage, the coefficient of rotation is −1.167, and the coefficient of tenure on GDP growth is 0.0702; therefore, the real effect of rotation on growth is 1.167 × 0.0702 + 1.223 = 1.1141. Ceteris paribus, in column 2, the real effect is calculated to be 0.902.

13. This increases our observations, and the relationship between age and tenure is more significant. Moreover, earlier on we treat tenure as exogenous an variable when tenure < 4(5), but now we treat it as an endogenous variable to run regressions. It is clear that this can enhance robustness of results.