It is well known that the two-country real business cycle model (as initially proposed by Backus, Kehoe and Kydland (1992)) has difficulty producing positive cross-country correlations of hours worked and investment. Recently, Baxter and Farr (2005) found that adding endogenous capital utilization rates to a two-country model raises these correlations. However, their results show that for realistic parameter values, the model still cannot match both correlations simultaneously as the cross-country correlation of investment is consistently too low. A first contribution of this paper is to show that learning-by-doing is another mechanism that raises the cross-country correlations of investment and hours. A second contribution is the finding that a two-country model with learning-by-doing and endogenous capital utilization produces positive cross-country correlations of hours and investment that are in the range observed empirically. A third contribution is to show that the model with learning-by-doing and capital utilization can produce positive cross-country correlations of output, consumption, investment and hours when technology shocks are uncorrelated across countries and there are international spillovers in the accumulation of organizational capital.