The Journal of Financial Research. Vol. XV, No.4· Winter 1992 EFFECTS OF AGENCY AND TRANSACTION COSTS ON DIVIDEND PAYOUT RATIOS: FURTHER EVIDENCE OF THE AGENCY-TRANSACTION COST HYPOTHESIS Stephen J. Dempsey, Gene Laber University of Vermont Abstract In this study we replicate and extend an agency-transaction cost model of dividend payout previously hypothesized and supported in the literature. We find no statistical difference between the estimated regression model obtained for the original seven-year sample period, 1974-80, and that obtained for our seven-year period, 1981-87. The lat- ter period is characterized by significantly lower inflation, stronger economic growth, and lower taxes. The intertemporal stability of the model suggeststhat it is useful for pre- dicting dividend payout at the individual firm level. I. Introduction The agency-cost hypothesis is one explanation offered for solving the divi- dend "puzzle." Rozeff (1982) and Easterbrook (1984) observe that paying divi- dends forces firms to raise external capital more frequently, subjecting them to monitoring and discipline by capital markets. Rozeff (1982) combines agency and transaction costs in an equilibrium model of dividend payout and empirically tests the model on a large sample of nonregulated firms. Using five explanatory variables to proxy for agency and transaction cost effects, he explains nearly half of the cross-sectional variation in mean payout ratios for one thousand firms from 1974 to 1980. While the Rozeff-Easterbrook agency cost arguments are widely cited in the literature (e.g., Ang (1987), Copeland and Weston (1988), Smith (1990)), empiri- cal support for the model is limited to the original Rozeff (1982) findings. In this paper we extend Rozeff's analysis to the 1980s, a period characterized by lower in- flation, stronger economic growth, and lower taxes. Despite the significantly dif- ferent economic conditions between the two periods, the estimated regression model not only holds up well in terms of explanatory power, but remains statisti- cally indistinguishable from that estimated for Rozeff's original period. The model thus appears to possess structural stability over time. II. Method Following the sample selection and variable definition procedures described in Rozeff (1982), the sample is restricted to nonregulated, dividend-paying, do- mestic firms contained in the Value Line Investment Survey. After applying the We wish to thank Michael Rozeff and two anonymous reviewers for their valuable contributions. Appreciation is also extended to John Wiley for his conscientious data-gathering efforts. 317