This thesis examines whether the level of Swedish economic policy uncertainty (EPU) can predict excess returns on the Swedish stock market. We run out-of-sample forecasting using an EPU-based predictive model constructed with the official Swedish EPU index developed by Armelius et al. (2017). Forecasting errors for one-, two-, three-, six-, and twelve-month holding periods and four measures of central tendency are analysed and compared against a random walk benchmark. The findings suggest that EPU has limited forecasting ability for excess stock returns in Sweden, and the EPU-based model demonstrates superior forecasting accuracy only in two out of twenty instances, both for the one-month holding period. However, the forecast errors remain relatively large, casting doubt on the model's ability to outperform the market. Furthermore, the EPU-based model consistently underestimates excess returns, questioning its usefulness as a predictor. Notably, the random walk benchmark's forecast error improves with longer holding periods, raising doubts about the predictability of market movements in the long term.