The aggregate dividend payout ratio forecasts excess returns on both stocks and corporate bonds in postwar U.S. data. High dividends forecast high returns. High earnings forecast low returns. The correlation of earnings with business conditions gives them predictive power for returns; they contain information about future returns that is not captured by other variables. Dividends and earnings contribute substantial explanatory power at short horizons. For forecasting long‐horizon returns, however, only (scaled) stock prices matter. Forecasts of low long‐horizon stock returns in the mid‐1990s are caused not by earnings or dividends, but by high stock prices.

Using data from the 1986 oil price decrease, I examine the capital expenditures of nonoil subsidiaries of oil companies. I test the joint hypothesis that 1) a decrease in cash/collateral decreases investment, holding fixed the profitability of investment, and 2) the finance costs of different parts of the same corporation are interdependent. The results support this joint hypothesis: oil companies significantly reduced their nonoil investment compared to the median industry investment. The 1986 decline in investment was concentrated in nonoil units that were subsidized by the rest of the company in 1985.

When the discount rate falls, investment should rise. Thus with time‐varying discount rates and instantly changing investment, investment should positively covary with current stock returns and negatively covary with future stock returns. Aggregate nonresidential U.S. investment contradicts both these implications, probably because of investment lags. Investment plans, however, satisfy both implications. These investment plans, from a U.S. government survey of firms, are highly informative measures of expected investment and explain more than three‐quarters of the variation in real annual aggregate investment growth. Plans have substantial forecasting power for excess stock returns, showing that time‐varying risk premia affect investment.