Abstract The evidence from the repo market is more supportive to the expectations hypothesis, but term structure anomalies still remain. Using the Bekaert–Hodrick–Marshall (2001) method, we investigate whether term structure anomalies can be explained by peso problems by estimating a regime-switching model for the overnight repo rate. We find that term structure anomalies can largely be accounted for by peso problems, probably along with a small time-varying risk premium for the full sample. However, peso problem explanations cannot resolve term structure anomalies for the postcrisis sample. In addition, we find that three regimes are related to calendar effects in the repo market.