What moves the stock market?

Campbell, John Y.

Stock prices tend to move with the state of the economy. This fact is both familiar--it describes the United States, the United Kingdom, and many other national markets in the postwar period, the interwar period, and the 19th Century(1)--and surprisingly hard to explain. My most recent work, with John H. Cochrane, proposes a simple model that accounts for this and other aspects of aggregate stock market behavior.

Financial theorists describe asset prices as expected present values of future payments, discounted to adjust for time and risk. This can be seen as a matter of accounting: if an asset's price is high today, then either its price must be even higher tomorrow, or its dividend must be high tomorrow, or its rate of return between today and tomorrow (the ratio of price plus dividend tomorrow to price today) must be low. If prices cannot grow explosively forever, then an asset with a high price today must have some combination of high dividends and low returns over the indefinite future. Furthermore, investors must recognize this fact in forming their expectations, so when an asset price is high, investors must expect some combination of high future dividends and low future returns.

In a series of papers, some written jointly with Robert J. Shiller, I have developed this simple insight into a quantitative framework for analyzing the variation of asset prices. As a matter of accounting, price movements must be driven by some combination of changing expectations ("news") about future dividends and changing expectations about future returns; the latter in turn can be broken into news about future riskless real interest rates and news about future excess returns on risky assets over riskless ones.(2)

Until the early 1980s, most financial economists believed that there was very little predictable variation in stock returns and that dividend news was by far the most important factor driving stock market fluctuations. The first challenge to this orthodoxy came from Stephen F. LeRoy and Richard Porter, and from Shiller.(3) These authors pointed out that dividends, and plausible measures of expectations about future dividends, are far less volatile than stock prices. A few years later, Eugene F. Fama and Kenneth R. French studied the predictability of stock returns from past returns and dividend-price ratios. They pointed out that the predictable variation in stock returns becomes increasingly impressive--in the sense that it accounts for an increasing share of the overall variation of returns--as one lengthens the horizon over which returns are measured.(4)

In several papers, some written with Shiller and one with John M. Ammer, I have shown that these observations are intimately related. When the expected return on an asset changes slowly over time, any change in today's expected return has a large effect on the asset price. This increases the volatility of prices; it also may increase the volatility of short-horizon returns relative to the volatility of long-horizon returns, so that the predictable variation of returns is a more important component of the variation of total returns at long horizons. The dividend-price ratio reflects the underlying variation in the expected stock return, and therefore is a good forecasting variable for long-horizon stock returns. The message of this research is that changing expectations of future stock returns are a powerful force driving movements in stock prices.(5)

None of this work addresses the important question of what economic factors cause the expected return on the aggregate stock market to change over time. As a matter of logic, the expected stock return can change only if the short term riskless real interest rate changes, or if the expected excess return on stock over short-term debt changes. The expected excess return may change if the riskiness of stock changes, or if the "price of risk" (the extra return demanded by investors for bearing risk) changes.

There are many theoretical asset pricing models in which the riskless interest rate changes over time. One standard model suggests that assets are priced as if there is a representative investor who consumes aggregate consumption. When consumption is low and predicted to rise, the representative investor would like to borrow from the future; the riskless real interest rate must rise to deter the investor from doing so, and this drives down asset prices in recessions.(6) Unfortunately, there is little empirical evidence for the strong countercyclical variation in the real interest rate required by this story; and, it cannot explain why excess stock returns over Treasury bills are just as predictable as raw stock returns.

Many researchers have explored the idea that risk varies over time. Econometricians have found that the conditional variance of stock returns is highly variable. If conditional variance is an adequate measure of risk, then perhaps this can explain the predictability of excess stock returns. But this approach has several problems. First, changes in conditional variance are most dramatic in daily or monthly data, and are much weaker at lower frequencies. There is some business-cycle variation in volatility, but it does not seem strong enough to explain large movements in aggregate stock prices.(7) Second, forecasts of excess stock returns do not move proportionally with estimates of conditional variance.(8) Finally, one would like to explain the cyclical variation of stock market volatility, rather than treating it as an unexplained phenomenon.

The remaining possibility is that the price of risk moves cyclically. In a recent paper with Cochrane, I have proposed a simple asset pricing model with this property.(9) Cochrane and I suggest that assets are priced as if there is a representative investor who consumes aggregate consumption; but in a departure from the standard model, the investor's utility is a function of the difference between consumption and "habit," where habit is a slow-moving nonlinear average of past consumption. This utility function makes the investor more risk averse during recessions, when consumption is low relative to its past history, than in periods when consumption is high relative to its past history.

Cochrane and I develop a version of this model in which aggregate consumption is a random walk. For simplicity, we model stocks as assets that pay aggregate consumption as their dividends. Hence, by construction, there is no predictable variation in dividend growth. If expected stock returns were constant, the dividend-price ratio would be constant, and stock returns always would equal aggregate consumption growth.

We also can parameterize the model so that the riskless real interest rate is constant. Thus we can shut down all the standard channels by which macroeconomic volatility is communicated to the stock market. Yet we can still explain the observed volatility and long-horizon forecastability of stock returns through time-varying risk aversion. When consumption falls, the aggregate stock market falls because dividends are lower, but more importantly because investors are more risk averse. This drives the dividend-price ratio up and is followed, on average, by high returns as risk aversion returns to more normal levels. The model predicts that stock returns are volatile, and that they are more volatile in recessions, when the habit-formation effect is more powerful.

Our model has important implications for both finance and macroeconomics. Financial economists have been puzzled by the fact that the traditional Capital Asset Pricing Model (CAPM), which prices assets by their covariance with the aggregate stock market, seems to work better than the consumption CAPM, which prices assets by their covariance with aggregate consumption. Our model accounts for this fact. Although aggregate consumption is the driving force in the model, the effect of a consumption shock depends in a complicated way on the past history of consumption The movement in aggregate stock prices is a better proxy for the underlying shock to investors' welfare than is the consumption shock itself, and this explains the comparative success of the CAPM.

Macroeconomists have tended recently to downplay the importance of economic fluctuations in favor of an emphasis on long-term economic growth.(10) But in our model, fluctuations have important negative effects on welfare because they move consumption in the short term, when consumers' habits have little time to adjust. The evidence from asset markets is that consumers take fluctuations extremely seriously, and macroeconomists should not neglect this fact.

1 E. F. Fama and K. R. French, "Business Conditions and Expected Returns on Stocks and Bonds, "Journal of Financial Economics (November 1989), pp. 23-50; and J. J. Siegel, Stocks for the Long Run, 1994.

2 J. Y. Campbell and R. J. Shiller, "The Dividend-Price Ratio and Expectations of Future Dividends and Discount Factors," Review of Financial Studies (Fall 1988), pp. 195-228, and "Stock Prices, Earnings, and Expected Dividends," Journal of Finance (July 1988), pp. 661-676; and J. Y. Campbell, "A Variance Decomposition for Stock Returns," Economic Journal (March 1991), pp. 157-179.

3 S. F. LeRoy and R. Porter, "The Present Value Relation: Tests Based on Variance Bounds," Econometrica (May 1981), pp. 555-577; and R. J. Shiller, "Do Stock Prices Move Too Much to Be Justified by Subsequent Changes in Dividends?" American Economic Review (June 1981), pp. 421-436.

4 E. F. Fama and K. R. French, "Permanent and Temporary Components of Stock Prices, "Journal of Political Economy (April 1988), pp. 246-273, and "Dividend Yields and Expected Stock Returns, "Journal of Financial Economics (October 1988), pp. 3-25. See also J. M. Poterba and L. H. Summers, "Mean Reversion in Stock Returns: Evidence and Implications, "Journal of Financial Economics (October 1988), pp. 27-60.

5 J. Y. Campbell and R. J. Shiller, "The Dividend-Price Ratio . . .," op. cit., and "Stock Prices, Earnings . . .," op. cit.; J. Y. Campbell, "A Variance Decomposition . . .," op. cit.; and J. Y. Campbell and J. M. Ammer, "What Moves the Stock and Bond Markets? A Variance Decomposition for Long-Term Asset Returns," Journal of Finance (March 1993), pp. 3-37.

6 A simple version of this model is given by L. P. Hansen and K. J. Singleton, "Stochastic Consumption, Risk Aversion, and the Temporal Behavior of Asset Returns, "Journal of Political Economy (April 1983), pp. 249-268.

7 J. Y. Campbell and L. Hentschel, "No News Is Good News: An Asymmetric Model of Changing Volatility in Stock Returns, "Journal of Financial Economics (June 1992), pp. 281-318, argue that predictable variation in returns arising from changing volatility cannot account for much of the overall variation in returns.

8 C. R. Harvey, "Time-Varying Conditional Covariances in Tests of Asset Pricing Models, "Journal of Financial Economics (October 1989), pp. 289-317, and "The World Price of Covariance Risk, "Journal of Finance (March 1991), pp. 111-157.

9 J. Y. Campbell and J. H. Cochrane, "By Force of Habit: A Consumption-Based Explanation of Aggregate Stock Market Behavior," unpublished paper, Harvard University and University of Chicago, September 1994.

10 R. J. Barro, for example, in the Summer 1994 NBER Reporter, writes that "Economic growth is the part of macro-economics that really matters."