This paper analyzes the important time variation in U.S. aggregate portfolio allocations. To do so, we first use flexible descriptions of preferences and investment opportunities to derive optimal decision rules that nest tactical, myopic, and strategic portfolio allocations. We then compare these rules to the data through formal statistical analysis. Our main results reveal that i) purely tactical and myopic investment behaviors are unambiguously rejected, ii) strategic portfolio allocations are strongly supported, and iii) the Fama-French factors best explain empirical portfolio shares.