the reduced form and first stage estimates in columns (1), and (3) are presented in panel B of Table 2. Note that the coefficient in column (1) is a lower bound of the effect of working capital on sales, because the reform should have led to some substitution of bank credit for market credit. The IV coefficient is 0.75, with a standard error of 0.37. The effect of working capital on sales is very close to 1, a result which would imply that there cannot be an equilibrium without credit constraint.

The IV estimate of the impact of bank credit on profit is 1.79, though again the sample is limited to firms with positive profits. This is substantially greater than 1, which suggests that the technology has a strong fixed cost component. However, these coefficients also allow us to estimate the effect of credit expansion on profits.

We can use this estimate to get a sense of the average increase in profit caused by every rupee in loan. The average loan is Rs 8,6800. Therefore, and increase of Rs. 1,000 in the loan corresponds to a 1.15% increase in loans. Taking 1.79 as the estimate of the effect of the log increase in loan on log increase in profit, an increase of Rs. 1,000 in lending causes a 2% increase in profit. At the mean profit (which is Rs. 36,700, this would correspond to an increase in profit of Rs. 756.^{13 }

A last piece of important evidence is whether big firms become more likely to default than small firms after the reform: the increase in profits (and sales) may otherwise reflect more risky strategies pursued by the large firms. In order to answer this question, we collected additional data on the firms based in the Mumbai region (138 firms, a bit over half the sample). In partic- ular, we collected data on whether any of these firm’s loan had become non-performing assets (NPA) in 1999, 2000 or 2001, or were NPA before 1999. The number of NPAs is disturbingly large (consistent with the high rate of NPAs in Indian banks), but large and small firms are equally likely to have a non-performing loan: 7.7% of the big firms and 7.29% of the small firms (who were not already NPA), default on their loan in 2000 or 2001. Among the firms in Mumbai, 2.5% of the large firms, and 5.96% of the small firms had defaulted between 1996 and 1998. The fraction of firms that had defaulted thus increased a little bit more for large firms,

^{13 }This estimate may be affected by the fact that the firms with negative profits are dropped from the sample. We have also computed the estimate of the marginal product of capital using data on sales and cost instead of using profits directly. We found that an increase in Rs 1,000 in the loans leads to an increase in Rs 730 in profits.

11