2. Non-Tobin’s q in Tests for Financial Constraints to Investment

In contrast, if the firm is a price-maker, then average q is higher than marginal q by what is legitimately called the monopoly rent.

This is under the special case of no debt next period B ^′  = 0. Generally speaking B ^′ has to be solved from a system of two Euler equations, one equation for investment and another for debt.

Similar results are obtained when the sample is partitioned on the basis of bond ratings (Gilchrist and Himmelberg 1995), firm size (Gertler and Gilchrist, 1994), membership of an industrial keiretsu in Japan (Hoshi et al. 1991). A detailed review of this literature can be found in Hubbard (1998) and Bernanke et al. (1999). In contrast with these results Kaplan and Zingales (1997) find that the coefficient on cash flow does not increase monotonically across groups of firms as the degree of financial constraint increases. Actually, firms that seem less constrained according to several criteria have a higher coefficient on cash flow, as compared to more constrained firms. However, as shown by Pratap (2003), this result can be rationalized by the presence of liquidity constraints when capital adjustment costs are non-convex.

At the same time that the literature is moving toward more structural approaches one can also distinguish the trend to move in the opposite direction, toward performing “natural” experiments. This method consists of exploiting a policy change that affected the flow of credit to an identifiable subset of firms. Then the researcher computes “difference-in-differences,” that is, a twofold comparison between observed variables of “control” and “treated” firms, observed “before” and “after” the policy change. For instance, Banerjee and Duflo (2004) exploit a 1998 reform in India that increased the maximum size below which a firm is eligible to receive priority sector lending. Control firms are those that were already in the “priority” sector. The result is that bank lending and firm revenues went up for the newly targeted firms in the year of the reform, so they conclude that there are severe credit constraints. Under this approach, measuring q is optional, as it is not needed to determine the treatment effect.