Risk-adjusted momentum returns are usually estimated by sorting stocks into a regularly rebalanced long-short portfolio based on their prior return and then running a full-sample regression of the portfolio returns on a set of factors (portfolio-level risk adjustment). This approach implicitly assumes constant factor exposure of the momentum portfolio. However, momentum portfolios are characterized by high turnover and time-varying factor exposure. We propose to estimate the risk exposure at the stock-level. The risk-adjusted return of the momentum portfolio in month t then is the actual return minus the weighted average of the expected returns of the component stocks (stock-level risk adjustment). Based on evidence from the universe of CRSP stocks, from sub-periods and size-based sub-samples, from volatility-scaled momentum strategies (Barroso and Santa-Clara 2015) and from an international sample covering 20 developed countries, we conclude that the momentum effect may be much weaker than previously thought.