Scale effects in growth, positive effects of the population size on per capita output growth, have been rejected by cross-country regressions. This paper, however, finds that long-run time-series data supports the effects. Moreover, although scale effects in growth seem to be inconsistent with the fact that a substantial increase in the R&D labor in the postwar United States did not raise its growth rate, the theoretical part of this paper proposes costly international knowledge diffusion as its possible reason, suggesting that growth did not improve most likely because additional R&D labor was devoted to knowledge diffusion, rather than innovation. Calibration analysis shows that the key variables predicted by the model are not very different from their actual values in the postwar United States.