ling for bank size in 1980.19 The idea behind this comparison is that the relationship between size and behavior should not change dramatically around the cutoff, unless nationalization itself causes changes in bank behavior. This will allow for credible causal inference on the role of bank ownership on bank behavior.
In order to get a sense of the magnitude of lending differences among bank types, we first divide the banks into five groups, based on their size in 1980: State Bank of India and its affiliates, large nationalized banks (nationalized in 1969), “marginal” nationalized banks (nationalized in 1980), “marginal” private banks (relatively large, but just too small to be nationalized in 1980), and small private banks. Because the geographic districts in which banks are located vary (soil quality, rural population, etc.), and face different economic shocks, we focus here on comparing differential bank behavior within each district. Our outcomes of interest include average loan size, residual interest rate,20 and share of bank lending to the following areas: agriculture, rural credit, small scale industry, government credit, and “trade, transport and finance.” The unconditional, India-wide means of these variables are given in column 1 of Table 3. To estimate bank-group effects, we regress credit outcome variables for each bank group g in district d on D district dummy variables, and BG1, ..., BGG bank group dummy variables. Specifically, we estimate:
i γiBG +
δiDistricti + εb,d,t
[TABLE 3 ABOUT HERE] The estimated bank group effects, ˆγ1, ..., ˆγG give the average share each bank gives to each sector, after controlling for differences across districts. These coefficients are presented in Table 3. (We use data from 1992, 1993, 1999, and 2000.) For example, the estimated ˆγ for the average loan size from banks in the State Bank of India group is Rs. 56,190. Compared to the average loan size of the State Bank of India, nationalized banks gave slightly smaller loans (an average of
19 Cole (2004). 20 The residual interest rate is obtained by regressing the interest rate on a wide range of control variables, such
as: small scale industry indicators, borrower occupation dummies (at a three-digit level), district fixed effects, size of loan, an indicator for whether the borrower is from the public or private sector, and dummies indicating whether the loan is given in a rural, urban, semi-urban, or urban area.