Banking Structure, Labor Intensity, and Industrial Growth : Evidence from China

While China has seen rapid economic growth over the last three decades, many studies document a negative correlation between traditional measurements of banking development and economic growth in China. Among many efforts to explain the low efficiency of Chinese banking sector, two arguments focus on the dominance of the four big state-owned banks but they emphasize different facets of the “Big Four.” The ownership-structure view argues that the state ownership of the largest banks and corresponding government interventions in capital allocation should be responsible for the bad performance of the banking sector. The size-structure view states that it is the improper size structure of the banking sector that leads to its low efficiency. This paper aims to empirically disentangle these two different explanations. With data on the banking sector and 28 manufacturing sectors in 30 provinces of mainland China over the period 1999-2007, this paper investigates the differential effects of banking structure on the growth rates of different industries. In order to identify the channel through which banking structure affects industrial growth, the paper constructs two interaction variables: interaction between labor intensity of each industry and banking structure at each province, interaction between the share of non-state-owned firms and banking structure. To capture the purely technological feature of each industry, the paper uses the labor-capital ratio of each U.S. manufacturing sector as a proxy for labor intensity of the corresponding Chinese manufacturing industry. We find that more labor-intensive industries grow faster than more capital-intensive industries in provinces with more active small banks, compared to provinces with more dominant Big Four branches. These results are robust to alternative measures of labor intensity. But the interaction variable used to capture the ownership-structure view is not significant in all the regressions. Thus the results are consistent with the size-structure view while not support the ownership-structure argument