The economic literature has long attributed non-zero expected excess returns in currency markets to time-varying risk premiums demanded by risk-averse investors. This paper, building on Bacchetta and van Wincoop's (2021) portfolio balance framework, shows that such returns can also arise when investors are risk-neutral but face portfolio adjustment costs. Models with adjustment costs but no risk aversion predict a negative correlation between exchange rate levels and expected excess returns, while models with risk aversion but no adjustment costs predict a positive one. Using data from nine inflation targeting economies with floating exchange rates (2000–2024), we find strong empirical support for the adjustment costs framework. The negative correlation persists even during periods of low market stress, further evidence that portfolio adjustment costs, not risk premium shocks, drive the link between exchange rates and excess returns. Our model predicts that one-year expected excess returns should have predictive power for multi-year returns, with longer-term expected returns as increasing multiples of short-term expectations, and the predictive power strengthening with the horizon. We confirm these findings empirically. We also examine scenarios combining risk aversion and adjustment costs, showing that sufficiently high adjustment costs are essential to generate the observed negative relationship.These findings provide a simpler, testable alternative to literature relying on assumptions about unobservable factors like time-varying risk premiums, intermediary constraints, or noise trader activity.