Comments and Discussion

Robin Brooks: Do changes in the population age distribution account for long-run cycles in the U.S. stock market? If so, what will happen to stock prices when the baby-boomers retire? Before this paper, the literature on demography and the stock market offered conflicting answers. On the one hand, Bakshi and Chen reported that, since 1945, the real S&P price index had moved in step—to a striking degree—with the average age of the U.S. population older than twenty.¹ They explained this finding by saying that an older population means more people saving for retirement, which pushes up the real price of stocks. This reasoning suggests that real stock prices will fall when the boomers run down their savings in retirement. On the other hand, Poterba found little association between demographics and real returns on stocks and other financial assets.² He argued that forward-looking, rational investors anticipate demographic change, so that there should be little contemporaneous association. According to this line of thinking, the coming retirement of the baby-boomers is already priced into stocks.

Among the many important contributions of this paper is that these seemingly conflicting results may be perfectly consistent. The authors construct an overlapping-generations endowment economy, which they use to simulate the effects of changes in the U.S. age distribution on stock prices and the equity premium. The model yields several predictions: First, real stock prices are positively related to the ratio of the middle-aged population to the population of young adults (the MY ratio), even when investors are forward-looking and rational, but that real stock returns are positively related to the change in this ratio. Second, demographic fluctuations alone account for about half the peak-to-trough variation in stock prices, but, after business cycle shocks are taken into account, these fluctuations account for virtually all of the peak-to-trough variation over the postwar period. Third, the equity premium is negatively related to the MY ratio, so that the premium is high when the working population is tilted toward the young and stock prices are low. Fourth, these effects reduce the lifetime consumption of large generations like the baby-boomers but make smaller generations, such as the boomers' parents and children, better off. This is what the paper calls the favored cohort effect.

The authors test these predictions empirically. The predictions for stock prices and returns fit the data relatively well, but the model performs less well when it comes to interest rates and the equity premium. This empirical exercise suffers from some of the standard problems in this literature. First, there are, in effect, few observations because demographics is such a slow-moving process. Second, at long horizons certain common shocks, notably the Great Depression and World War II, drive both the age distribution and the stock market. Third, because these shocks are global, data from other countries do not represent independent observations. As a result, the empirical part of the paper—through no fault of the authors—is somewhat less compelling. I will therefore focus my remarks on the predictions of the model, hoping to put some of them into perspective and to qualify others.

My comments are structured as follows. I begin by noting that the authors' choice of benchmark model—an exchange economy—has implications for the magnitude of demographic effects on stock prices. The authors are up front about this, but I believe it is worth reiterating. I then comment on saving and portfolio behavior, which the authors implicitly hold constant over time, even though their simulations span the past 100 years. Next, I focus on the favored cohort effect. From a policy perspective, this is perhaps the most important result, because a better understanding of which generations benefit, and which generations lose, from the asset market effects of the baby boom is critical for allocating the costs of Social Security reform across generations. I will show that the favored cohort effect is rather sensitive to some underlying assumptions and should thus be interpreted carefully. Finally, I provide some perspective from my own...