In this study, we develop a dynamic factormodel that incorporates features of price dynamics across assets as well as through time. With the dynamic factors extracted via the Kalman filter, we formulate two testable asset-pricing models: the risk-adjusted pricing model (RAPM) and the bias-adjusted pricing model (BAPM). We then conduct asset-pricing tests in the in-sample context. In addition, we perform out-of-sample tests for competing models, presenting pair-wise comparisons of the accuracy in one-step-ahead forecasts. We provide evidence that the ex post dynamic factors alone do a better job than the Fama-French (FF, 1993) three factors both in-sample and out-ofsample. Our analyses also demonstrate that the ex ante factors are a key component in asset pricing and forecasting. By employing the ex ante factors together with ex post ones, the BAPM further improves upon the explanatory and predictive power achieved by the naive benchmark, the FF 3-factor model, and the RAPM. In particular, the BAPM can even explain and better forecast the momentum portfolio returns, which are mostly missed by the FF 3-factor model.