Financing constraints and heterogeneity in investment behavior

Abstract

During the 1980s, a substantial body of new theory has been built on the premise that informational problems between borrowing and lending institutions may limit the capital spending of several classes of firms in a rational economy. Inspired by these theories, several authors have tested the importance of the impact of rationing on investment across different classes of firms and for different countries [Fazzari et al., "Financing Constraints and Corporate Investment," Brookings Papers on Economic Activity, I, 1988; Devereux et al., "Investment, Financial Factors, and Cash Flows: Evidence from U.K. Panel Data," in Hubbard (ed) Asymmetric Information, Corporate Finance, and Investment, 1990; Hoshi et al., "Evidence on q and Investment for Japanese Firms," Journal of the Japanese and International Economies, 4, 1992]. Unfortunately, since constrained firms are not directly observable in available data sets, empirical tests of financing constraints have generally relied on a priori sorting of firms into potentially constrained and unconstrained groups. Statistical tests are then used to see if internal liquidity actually plays a more important role in shaping the investment patterns for the constrained groups. Clearly, to the extent the a priori sample sorting is incorrect, this procedure will produce biased test results for both groups. This paper proposes a new strategy for investigating the impact of credit rationing on investment in data sets where constrained firms are not directly observable. The strategy is designed to circumvent the potential estimation biases due to incorrect sorting. This is achieved by building the parametric impact of competing sample sorting criteria directly into model specification. Each sorting criterion implies a different sample stratification and a different non-nested specification. The competing specifications are then evaluated against each other and the best stratification (if any) is chosen by applying one or more of the existing family of non-nested hypothesis tests [Goodman et al., "Sample Stratification with Non-Nested Alternatives: Theory and Hedonic Examples," The Review of Economics and Statistics, 1, 1990]. The results from the best stratification are then analyzed to see if they are consistent with the predictions of the rationing models. An application of the procedure to Canadian panel data for 299 companies over the period 198689 produced instructive results. First, among firm size, the retention ratio, and Tobin's q, the best sorting criterion is Tobin's q. Second, the results from conditioning on the best sorting criterion reveals that, although financial factors exert an important general influence on investment for all types of firms, the data provide no evidence of a systematically larger impact of liquidity for investment for constrained firms over that for unconstrained groups. The procedure proposed here has important implications for future research in the area. Future research must account for the impact of sorting before drawing any inference about the existence of credit rationing.