with major frauds. This list consists of eighty-seven officials in public sector banks between the years 1992 and 2001. While nature of the fraud with which they are charged is not known, we do know that approximately 72 percent of frauds relate to illegal extension of credit, while the balance is classified as kite-flying or “other.” 35 Since our hypothesis is that vigilance results in a decrease in lending activity, the inclusion of spurious non-credit related vigilance activity should bias coefficients towards zero. Summary statistics for credit data, and the CVC fraud data are given in Table 8.
Empirical analysis: The first approach is to use bank level monthly lending data to estimate the effect of vigilance activity on lending, using the following equation,
y i t = α i + β t +
where yit is log credit extended by bank i in month t, αi is a bank fixed-effect, βt is a month fixed effect, and Di,t−k is an indicator variable for whether vigilance activity was reported by the CVC for bank i in month t-k. Standard errors reported are adjusted for serial correlation and heteroscedasticity. The basic idea is to compare the bank that was affected by the vigilance activity with other public sector banks, before and after the vigilance event. Which event window to use is not immediately clear: the appropriate start date would most likely be the month in which it became known that vigilance proceedings were under way, or perhaps the date bankers learned of the judgment. The data published by the CVC give only the date at which the CVC provided advice on the case, and the date on which action was taken. Nor is it clear how long it should take before an effect appears, or for how long one would accept this effect to last. We therefore let the data decide, by estimating models which allow effects ranging from one month to four years.
[TABLE 9 ABOUT HERE] Table 9 presents estimation results from several similar specifications. Columns (1), (2), and (3) provide estimates for windows of one, twelve, and 48 months. There appears to be a clear effect of vigilance activity on lending decisions. Vigilance activity in a specific bank results in a reduction of credit supplied by all the branches of that bank by about 3-5 percent.
This effect is estimated precisely (and is significantly different from zero at the 5 percent level 35 Government of India, (2000).