This study empirically examines the prediction in Sikes and Verrecchia (2012) that the relation between capital gains tax rates and expected rates of return varies in the cross-section and over time with firm risk and market risk. Specifically, we test whether the general positive relation between expected returns and the capital gains tax rate becomes weaker or even reverses when (i) a firm's systematic risk is high, (ii) the aggregate market risk premium is high, or (iii) the risk-free rate is low. Using an international panel from 25 countries over the 1990 to 2004 period, we find evidence supporting these predictions. The results are particularly pronounced in countries with substantive changes in tax rates, a tradition of low tax evasion, less integrated capital markets, and less institutional ownership as well as around substantive changes in the three risk proxies. We corroborate our findings in a single country setting, using the 1978, 1997, and 2003 changes to the capital gains tax rate in the United States as events. Our results underscore the importance of macroeconomic and firm-specific factors in the determination of the effect of capital gains taxes on expected returns and show that the valuation effects can sometimes be in the opposite direction of what is generally expected.JEL Classification: G12, G15, G32, H24, K34, M41