Theoretically, risk-averse investors who expose to higher level of risks should expect to be compensated with more returns. However, prior studies empirically found inconclusive and often counterintuitive results. Motivated by advanced econometric techniques and choices of risk and return recently introduced in literature, we justify and choose the appropriate models and proxies to study the risk and return trade-off pattern in The Stock Exchange of Thailand. Instead of using historical returns mentioned in prior studies, this study chooses the earning ratio to proxy the expected return. For the risk and return trade-off testing model, we employ the ICAPM model with Glosten, Jagannathan and Runkle-GARCH (GJR-GARCH) and GJR with Continuous and Jump Variations (GJR-CJ) as volatility function. Consistent with the theory, results show that the baseline models can explain the positive relationship between risks and returns. By adding the effect of a rare event which can be estimated from the jump process of the data, the forecast accuracy is more improved. In addition, the results find that it is possible that this proxy can be applied in another model which is based on the mean and variance models. Finally, the results from the forecasting model and other models indicate that using an Earning to Price (E/P) ratio as a proxy for expected returns is more appropriate than using the historical return.