Market development and bank profit efficiency in China: application of the generalized Malmquist productivity index

Abstract

We study the total factor productivity of Chinese banks (the generalized Malmquist productivity index) and the impact of market development on bank profit efficiency using a unique sample of 150 Chinese commercial banks for the 1999–2008 period. Employing an output-oriented stochastic distance function approach, our analysis shows that the productivity growth of Chinese banks over time can be attributed mainly to improvements in technical efficiency and technical change. In addition, the efficiency of Chinese banks is heavily influenced by market development variables, including the proportion of non-state business, level of government intervention in the market, competition in the financial industry and competition in credit allocation. The effects of these factors on bank profit efficiency differ depending on the type of banks.

Keywords:

• generalized Malmquist productivity index
• market development
• bank profit efficiency
• China

JEL Codes:

• D24
• G21

Acknowledgements

We appreciate the helpful comments from participants at the 2010 APJAE Symposium on International Trade and the China Economy in Hong Kong. The research that led to this paper was conducted when Dr Baozhi Qu was at the City University of Hong Kong. Jianhua Zhang acknowledges the financial support from the National Natural Science Foundation of China (Project No. 71073021: The measurement of international seigniorage).

Notes

aThe value indicates the number of bank-years in the sample (or the sub-group of the sample) that take the value one for this variable and the number in the parentheses shows the number of different banks.

1. Given that the banking sector in China has been under heavy regulation of the government and most of the reforms related to this sector followed a “top-down” procedure, we expect that the causality goes from market development to banks’ profit efficiency and not the other way around.

2. A major advantage of the distance function approach is that it can be applied in the case of multiple inputs, multiple outputs, or in the absence of price information when the traditional dual approach is inapplicable (Jiang et al. 2009).

3. Simultaneous equation bias may exist when both inputs and outputs are included in the distance function as regressors. After the normalization procedure, the output ratios can be treated as exogenous (Coelli and Perelman 1996).

4. The output distance function requires that , but may not be equal to 1.

5. In our analysis, we take the logarithm of both the input and the output factors. To avoid taking the logarithm of zero or a negative number, it is a common practice in the literature to find the minimum value for each factor (usually a negative number), calculate its absolute value plus one (∣y∣ + 1), and then add this number to the initial variable value before taking the logarithm. However, if the quantitative level of ∣y∣ is exceptionally large, then the input–output relationship may change significantly for some banks. Here, we adopt the following procedure to tackle this issue. We use negative one million as the benchmark and add 1.01 million to the initial value of the variable. If the result is less than one, then we treat the logarithmic value of this observation as zero. For the rest, we directly use the logarithm. Although this approach effectively imposes a “penalty” on negative values and sacrifices some information by smoothing out the variation in some variables with negative and large absolute values, it allows us to better control the potential distortion of the estimation results resulting from exceptionally large negative values. Our main results are not sensitive to the choice of the benchmark value or to using the common practice approach to obtain the logarithmic values.

6. Jiang et al. (2009) have only 14 CCBs in their sample, compared with 133 CCBs in our sample.