Vlog: A Trillion-Dollar Question

After a lengthy internal replication that lasted two and a half years, we are finally ready to circulate the next installment of our ongoing empirical work: “Investment-based Costs of Equity.”

Title: Investment-based Costs of Equity
with Yicheng Liu (Ohio State) and Chen Xue (U. of Cincinnati)

This paper develops the q5-characteristics model that estimates costs of equity as out-of-sample forecasts from cross-sectional regressions. The q5-cost of equity outperforms the accounting implied cost of equity in predicting cross-sectional returns. Our cost of equity is precise at the industry level and aligned with average factor premiums. Its firm-level distribution is weakly left-skewed, whereas the accounting implied cost of equity is right-skewed. However, the accounting cost of equity outperforms ours in the time series. Factor models perform poorly in all out-of-sample tests. Gradient-boosted trees improve on cross-sectional regressions in the q5-characteristics model, but not reliably.