The effect of tax law changes on corporate investment and financing behavior: Empirical evidence from changes brought about by the Economic Recovery Tax Act of 1981.

This dissertation examines the relationship between debt and investment-related tax shields using changes in these classes of tax shields scaled by expected operating earnings following the passage of the Economic Recovery Tax Act(ERTA) in 1981. The substitution effect predicts that a negative relationship between changes in the two classes of tax shields will be observed in response to the increased investment-related tax shields offered by ERTA. Debt tax shields should decrease following ERTA since the probability of losing the tax benefit of tax shields would rise as investment-related tax shields increased following ERTA. Firms' probability of losing the deductibility of tax shields is used to segregate the sample into two groups. For the group of firms with a low probability of losing the deductibility of tax shields, the substitution effect is inapplicable and the relation between changes in the two classes of tax shields simply represents the debt securability effect. Since fixed assets can be used as collateral for debt, the debt securability hypothesis predicts a positive relationship between changes in debt and investment-related tax shields after the passage of ERTA. The model developed to segregate debt securability from the substitution effect reveals that, as predicted, the debt securability effect is positive for all firms and that the substitution effect is negative for those firms with a large probability of losing the benefits of tax shields. This reverses the findings of prior research. Controls for pecking order theory effects are introduced into the model to assure that the substitution effect observed is not due to debt ratio as predicted by Myers (1984). The findings described above remain intact except that the debt securability effect does not exist and the substitution effect is weaker for high-debt firms. Furthermore, support is offered for the pecking order theory. These results are robust to alternate specifications of time periods tested, variable definitions, data screening criteria and model specifications