4.3 Exploring endogeneity
Our hypotheses posit that the cost of debt is influenced by dividend payouts. Nevertheless, it is conceivable that the direction of causality may be reverse. That is, managers may set dividend payouts based on the cost of debt. To explore possible endogeneity, we utilize the 2003 dividend tax cut as an exogenous shock. The jobs and growth tax relief reconciliation (JGTRRA) of may 2003 reduced the top marginal tax rate on dividend by 20 percent. This tax change was pro proposed and signed into law with unusual speed, and thus represents a largely unanticipated and exogenous increase in the after –tax value of dividend to individual invertors. As a result, due to higher after-tax returns, dividend-paying stocks should become more attractive to invertors. In response managers should respond by increasing payouts. In sum one should expect corporate dividend payouts to increase after the passage of JGTRRA in 2003. We use this exogenous increase as instrument in two-stage estimation model.
We execute a two-stage regression. We construct a dummy variable equal to one for the period after 2003 and zero for the period up to 2003. In the first- stage regression, we regress dividend payouts the tax cut dummy and control variables. Then, in the second stage, we replace dividend payouts by the “predicted” payouts from the first stage. This way we capture exogenous dividend increases – not determined by previous credit spreads – to explain variance in corporate bond yields. The results are shown in 4. Model 1 represents the first- stage regression. As expected, the coefficient of the tax cut dummy is positive and significant, suggesting that dividend payouts increase significantly after the tax cut. Model 2 is the second – stage regression. The coefficient of the predicted dividend payouts is negative and significant, whereas the quadratic term carries a positive and significant coefficient. Taken together, the evidence suggests that the direction of causality likely runs from dividend payouts to the cost of debt than vice versa. The 2003 dividend tax cut provides an exogenous shock not related the cost of debt and allows to have a relatively clean identification strategy. Control variable coefficient are consistent with the results previously reported in table 3.