Cliff Asness: Fight The Fed Model
Fall 2003
The relationship between future returns and stock and bond market yields.
Tune to CNBC or the like for more than about 15 minutes, and you will hear a strategist, portfolio manager, or market pundit of some stripe explaining that the high market multiples of recent times are justified by low interest rates and/ or inflation. “Well, Maria, you have to understand-stocks might look expensive, but that is fine because interest rates and inflation are low.” Or so the refrain goes. In fact, to many on Wall Street and in the financial media this assertion has been elevated to the status of conventional wisdom.
The most Widespread version of this comparison of stocks to bonds is often deemed the Fed model. This model, allegedly found in the annals of a Fed report, not named because of any official Fed endorsement, comes in various forms, but generally asserts that the stock market’s earnings yield should be compared to current nominal interest rates (the earnings yield, or E/ P, is the inverse of the well-known price-to-earnings ratio or P/E).
Letting Y represent the yield on ten-year Treasuries, the model says we should look at E/ P versus Y. In its simplest form, it asserts stocks are cheap when E/P exceeds Y, expensive when Y exceeds 13/ P, and fairly valued when Y and E/ P are equal.
Even pundits who are united in their belief in the Fed model do not always agree on what it is telling them. Of course, as recent times make clear, the E in E/P is not a simple observable number. In addition, some adjust the basic comparison of E/ P and Y for a growth assumption or a required equity risk premium, or change the functional form of the relationship.
The basic widespread core belief implied by the Fed model, though, is that the stock market’s E/P must be compared to Y, and that low interest rates permit a low E/P or, equivalently, a high market P/E (and vice versa). It is this core belief (whether or not it is labeled the Fed model) that I study here.
The evidence strongly suggests that the Fed model is fallacious as a tool for long-term investors. Essentially, the comparison of E/ P to Y is erroneous as it compares a real number (P/E) to a nominal one (Y). The important point is that the stock market’s P/ E does not have to move with inflation since nominal corporate earnings already do so. Empirical evidence supports this theory. Investors forecasting future long-term stock returns would do much better relying on simple P/E, or the like, rather than the Fed model?
While the Fed model fails as a predictive tool for future long-term stock returns, it does work as a descriptive tool for how investors choose to set current stock market P/Es. Even here, however, the simple Fed model needs help. Applying a relationship studied in Bernstein [1997b] and Asness [2000], it is clear that the Fed model relationship must be conditioned on the perceived volatility of stocks and bonds. Without conditioning on perceived volatility, the simple Fed model is a failure over 1926-2001, even to describe how investors set P/Es. Conditioned on perceived volatility, however, the Fed model explains the puzzle of why the relative yield on stocks and bonds has varied so greatly over the last century.
Note that this finding that the Fed model has descriptive power for how investors set P/ Es in no way contradicts the finding that the Fed model fails as a predictive tool for stock returns. If investors consistently err and follow a poor model, it is not surprising that this same model fails those investors for making long-term forecasts.
See full PDF below.
