This article uses variation in access to a targeted lending program to estimate whether firms are credit constrained. While both constrained and unconstrained firms may be willing to absorb all the directed credit that they can get (because it may be cheaper than other sources of credit), constrained firms will use it to expand production, while unconstrained firms will primarily use it as a substitute for other borrowing. We apply these observations to firms in India that became eligible for directed credit as a result of a policy change in 1998, and lost eligibility as a result of the reversal of this reform in 2000, and to smaller firms that were already eligible for the preferential credit before 1998 and remained eligible in 2000. Comparing the trends in the sales and the profits of these two groups of firms, we show that there is no evidence that directed credit is being used as a substitute for other forms of credit. Instead, the credit was used to finance more production–there was a large acceleration in the rate of growth of sales and profits for these firms in 1998, and a corresponding decline in 2000. There was no change in trends around either date for the small firms. We conclude that many of the firms must have been severely credit constrained, and that the marginal rate of return to capital was very high for these firms.

Most papers that employ Differences-in-Differences estimation (DD) use many years of data and focus on serially correlated outcomes but ignore that the resulting standard errors are inconsistent. To illustrate the severity of this issue, we randomly generate placebo laws in state-level data on female wages from the Current Population Survey. For each law, we use OLS to compute the DD estimate of its “effect” as well as the standard error of this estimate. These conventional DD standard errors severely understate the standard deviation of the estimators: we find an “effect” significant at the 5 percent level for up to 45 percent of the placebo interventions. We use Monte Carlo simulations to investigate how well existing methods help solve this problem. Econometric corrections that place a specific parametric form on the time-series process do not perform well. Bootstrap (taking into account the autocorrelation of the data) works well when the number of states is large enough. Two corrections based on asymptotic approximation of the variance-covariance matrix work well for moderate numbers of states and one correction that collapses the time series information into a “pre”- and “post”-period and explicitly takes into account the effective sample size works well even for small numbers of states.