A generalized CAPM model with asymmetric power distributed errors with an application to portfolio construction

We estimate the CAPM model on European stock market data, allowing for asymmetric and fat-tailed return distributions using independent and identically asymmetric power distributed (IIAPD) innovations. The results indicate that the generalized CAPM with IIAPD errors has desirable properties. It is substantially less likely to be rejected than the traditional CAPM with normally distributed errors and, moreover, backtests show that portfolios constructed using IIAPD errors outperform the portfolio constructed with normally distributed errors in terms of commonly-used performance measures.