This study investigate the ability of various risk aversion measures to predict U.S. future real economic activity (REA). Considering that existing risk aversion measures widely used by researchers and financial institutions have different tendencies, we re-examine their empirical validity under diverse criteria (e.g., leading indicator, counter-cyclicality, high degree of persistence, and low volatility). We test whether risk aversion predicts future REA using in-sample and out-of-sample tests, along with sub period analysis. We find that most risk aversion measures perform well in-sample but exhibit mixed out-of-sample results depending on economic conditions. Our findings highlight the state-dependent nature of risk aversion and its usefulness as an economic indicator.