Theoretically, two-country real business cycle models with time-separable preferences and complete markets predict that cross-country investment correlations will be negative. The opposite is true in the data. This phenomenon has been described by Backus et al. [in Cooley (ed.), Frontiers of Business Cycle Research, pp. 331–356 (Princeton, NJ: Princeton University Press, 1995)] as a quantity anomaly. This paper proposes to address this discrepancy by allowing the nonseparability of preferences over time. Here, we incorporate internal habit formation into consumption. Our model predicts the empirically plausible value of cross-country investment correlation without sacrificing other business cycle statistics. The results are robust to the degree of spillovers and persistence in the specification of the productivity shocks.