We argue that the cointegrating relation between dividends and consumption, a measure of long-run consumption risks, is a key determinant of risk premia at all investment horizons. As the investment horizon increases, transitory risks disappear, and the asset’s beta is dominated by long-run consumption risks. We show that the return betas, derived from the cointegration-based VAR (EC-VAR) model, successfully account for the cross-sectional variation in equity returns at both short and long horizons; however, this is not the case when the cointegrating restriction is ignored. Our evidence highlights the importance of cointegration-based long-run consumption risks for financial markets. (JEL G1, G12) How does the riskiness of equity returns change with investment horizon? We show that at long horizons, risks are dominated by long-run consumption risks in dividends, while at short horizons, additional price risks may also matter. The cointegrating relation between dividends and consumption (i.e., the longrun consumption beta of dividends) provides a measure of long-run risks in dividends. This cointegrating relation has important conceptual and empirical implications for measuring risks in asset returns at all horizons. Empirically, we document that consumption betas determined by the cointegrationbased vector-autoregression (EC-VAR) can account for the cross section of mean equity returns at both short and long horizons. This, however, is not the case when the cointegration restriction between consumption and dividends is ignored. Hence, our evidence highlights the economic importance of long-run