SUMMARY
This article empirically test the lower partial moments models, Sortino, Upside Potential Ratio, Omega and Kappa, comparing them with the traditional CAPM, for listed shares of Ibovespa and Dow Jones index. These two classes of models are distinguished in terms of investors' profile assumptions and risk measurement. While the CAPM considers only the first two moments of the returns distribution, assuming investor's quadratic utility function (defined in terms of mean/expected returns and variance/risk), the other measures take into account higher moments of returns distributions (assimetry and curtosis). The Hansen-Jagannathan distance, which estimates the Stochastic Discount Factor (SDF) measurement error generated by each model, showed a distinction of the models in the two markets. While the CAPM performed better for Dow Jones shares, the lower partial moments models presented better results for Ibovespa, suggesting an advantage in the use of such models in markets with lower liquidity, fat tails (greater probability of extreme events) and greater asymmetry.