6,430 Rs. lower), while marginal nationalized banks gave slightly larger loans (the average was 8,350 Rs. greater), marginal private banks gave much larger loans (35,310 Rs. more), and small private banks gave loans much larger on average (58,500 Rs. more). These results appear to confirm conventional wisdom that nationalized and public banks give smaller loans than private banks.
The most informative comparison is between what we called the “marginal” nationalized and the “marginal” private bank, which were similar in size, but the former were nationalized while the latter weren’t. Many of the differences between the marginal nationalized and the marginal private banks are large: the marginal private banks gave 5 percentage points less credit to agriculture than the marginal nationalized banks: given that the all-India share of credit to agriculture is 11 percent, this difference is substantial. The results also suggest that nationaliza- tion led to more credit to small scale industry (an increase of 2 percentage points relative to the private banks; India-wide small scale industry receives 9 percent of total credit), four percentage points more credit to rural areas (compared to a national average of 12 percent), and slightly more to government enterprises (.7 percent more; the India-wide figure is 3 percent.). These increases come at the expense of credit trade, transport, and finance (nationalized banks gave 6 percent points less, compared to the national average share of 21 percent). The final column in Table 3 gives the results of an F-test of the hypothesis γMarginal Private = γM arg inal Nationalized. The rural and government lending differences are significant at the 5 percent level, while all others are significant at the one percent level.
While this is suggestive that private and public banks behave differently, the values in the table vary not only between marginal private and marginal nationalized banks, but across other bank groups as well. Thus, from this data alone, we cannot rule out the possibility that the difference in lending behavior is attributable to bank size, rather than ownership.
To obtain an accurate measure of the impact of nationalization, we examine lending behavior at the individual bank level, adopting a full-fledged regression-discontinuity approach. We first estimate bank effects analogous to the group effects estimated in equation (2), by replacing the Bank Group dummy indicators with individual bank dummy indicators, to obtain coefficients
ˆˆ β1, ...βB. These coefficients tell us to what extent bank b behaves differently from other banks,
after controlling for the characteristics of the districts in which each bank operates. We then