Empirical Estimation of the Cost of Equity: An Application to Selected Brazilian Utilities Companies.

We provide an extensive set of alternative models for the estimation of the real cost of equity in a sample of utitlites firms in Brazil with monthly data from March 2006 to June 2011.The main results are that the traditional CAPM is regected and, together with the Fama-French factors, give a poor fit, low beta and low predicted real return on equity. Additional factors improve the fit of the modesl and the estimated betas and real cost of equity increase relative to the traditional CAPM and Fama-French. Accounting for conditional heteroskedasticity shows that autocorrelation of variances is more important than news effects. The inclusion of higher order terms shows that the third order term is mostly significant and positive indicating preference for skewness in this sample period. Our estimates of betas and the implied predicted real cost of equity show that, across the best models, betas are significantly below unity in the range 0.26-0.73. The predicted real cost of equity, across the best models, for Brazil in this sector and smaple period averages 11% per year and is in the range of 8.7% to 13.2% per year.