Regional determinants of FDI in China: a factor-based approach

Abstract

We empirically investigate the factors that drive the uneven regional distribution of foreign direct investment (FDI) across Chinese provinces from 1995 to 2006. We first perform a factor analysis to summarize information embodied in around 40 variables and derive four FDI determinants: ‘institutional quality’, ‘labour costs’, ‘market size’, and ‘geography’. Applying these estimated factors, we then employ instrumental variable (IV) estimation to account for endogeneity. In line with theoretical predictions, we find that foreign firms invest in provinces with good institutions, low labour costs, and large market size. The Arellano-Bond dynamic panel generalised method of moments (GMM) results show strong agglomeration effects that multinationals tend to invest in provinces which attract other foreign firms, consistent with the economic geography literature. Several robustness tests indicate that low labour costs combined with improvements in institutions are the key for attracting FDI in China.

Keywords:

• China
• FDI
• regional distribution
• factor analysis

JEL Classifications::

• F21
• F23
• O18
• O53
• R11

Acknowledgements

The authors gratefully acknowledge Harry Garretsen (University of Groningen), Salil K. Sarkar (University of Texas at Arlington), Nick Horsewood (University of Birmingham), Michiel Gerritse (VU University Amsterdam), and Xiaohui Liu (Loughborough University) for their suggestions on earlier drafts. The authors would like to thank the two anonymous reviewers and Xiaming Liu for their constructive comments.

Notes

Notes

1. With firm-level data it is important to note that often they restrict the analysis to cross-section only, since there is no investment pattern at the firm level recorded over time. However, clearly reverse causality is a problem when using firm-level data.

2. This interpretation is supported by La Porta et al. (1999), who show that infrastructural quality is highly correlated to government performance.

3. The standard fixed effects transformation can only be applied if the instruments are strongly exogenous because the time-demeaned lagged values are correlated with the errors in the case of weakly exogeneity (Wooldridge 2002). By contrast, the forward orthogonal deviation transformation produces the transformed error based only on current and future values of the original error (, the superscript F denotes for future values). Therefore, weakly exogeneity implies E[z_is ()] = 0 for s ≤ t, where z_is are possible instruments. We use the forward orthogonal deviations IV estimator to keep efficiency. Nevertheless, our main findings are robust to using the first differencing IV estimator, as indicated by the standard Arellano-Bond GMM estimation results.

4. One has to keep in mind that the factors (clusters of variables) change over time, although some of the variables are rather static.

5. As evidence for the prevalence of vertical FDI in China, according to the National Bureau of Statistics of China, exports by foreign funded enterprises (FFE) take around 60% in total exports from China in 2006, with an increasing trend over the past 15 years.

6. Since there are no data for Chongqing before 1998 and no FDI inflows data for Tibet throughout the sample period, our sample excludes Chongqing and Tibet. The Eastern provinces are Beijing, Fujian, Guangdong, Hainan, Hebei, Heilongjiang, Jiangsu, Jilin, Liaoning, Shandong, Shanghai, Tianjin, and Zhejiang.

7. Removing the suggested variables results in significantly lower values of alpha. For example, Cronbach's alpha of the adjusted factor 4 (without imports) is 0.466; the new alphas of factor 2 (without government expenditure, exports, and junior high school, respectively) are 0.450, 0.450, and 0.297. Note that, overall, the internal consistency of the other factors remain unchanged, or reach alpha values above 0.600.