On the Persistence of Capital Structure - Reinterpreting What We Know

Abstract:     
Despite the known statistical problems with pooled regression specifications, we argue that current literature has identified a good list of factors that affect firms' capital structure decisions. In particular, to address the potential biases in both pooled and fixed effects specifications, we offer an econometric model that (a) separates cross-sectional and time-series effects of each factor, (b) uses a Tobit mixed normal specification, and (c) includes a lagged leverage term to capture leverage persistence. The flexibility of our model allows us to nearly match the explanatory power of the fixed effects specification with only a few parameters, while the linearity of the model allows for a straightforward interpretation of the parameter estimates. We find that, while leverage is persistent, it is not stable (time-series variation is only 30% smaller than cross-sectional variation); the known capital structure factors have differential explanatory power for cross-sectional and time-series variations in leverage, and lagged leverage has the largest explanatory power for time-series changes in leverage. We find that cross-sectional differences in leverage persistence are significantly related to a number of firm characteristics as well as persistence of these characteristics.

Number of Pages in PDF File: 50

Keywords: Capital Structure, Fixed Effect, Leverage, Mixture Distribution, Persistence

JEL Classification: G32

working papers series