Changes in real stock-market prices have a lot of explanatory value of the growth rate of U.S. aggregate business investment, especially for long samples that begin in 1891 or 1921. Moreover, for the period since 1921 where data on a q-type variable are available, the stock market dramatically outperforms q. The change in real stock prices also retains its predictive value in the presence of a cash-flow variable, such as after-tax corporate profits. Basically similar results apply to Canadian investment, except that the U.S. stock market turns out to have move preditive power than the Canadian market. I discuss some possible explanations for this puzzling finding, but none of the explanations seem all that convincing.