When facing uncertainty surrounding the risk loading, or beta uncertainty, arbi- trageurs allocate less arbitrage capital to anomalies. In this paper, we introduce a Beysian stochastic CAPM that explicitly accommodates separate random processes in beta and idiosyncratic volatility to estimate beta uncertainty in anomaly portfolios. We provide both theoretical and empirical evidence that beta uncertainty serves as arbitrage barriers to slow down arbitrage activities and thus is positively associated with future anomaly returns. We extend the analysis to firm level and show that beta uncertainty amplifies the effect of mispricing score on stock returns and reduces short selling activities. We also discuss the high correlation between beta uncertainty and idiosyncratic risk and the pitfalls of conventional estimation.