Using the Yield Curve to Time the Stock Market

In the study reported here, we extended the probit model for forecasting an economic recession by using the yield-curve spread as the explanatory variable to forecast a bear stock market and tested market-timing strategies based on the model. We found that the value of the yield spread between the composite 10-year+ U.S. T-bond yield and the three-month T-bill yield holds important information about the probability of a bear stock market. We discuss our out-of-sample market-timing tests at various probability screens of a forthcoming bear market in one month. At a 50 percent probability screen, our simulations show that, for the period studied, a market timer switching out of stocks (T-bills) into T-bills (stocks) one month before a bear (bull) stock market could have realized a compound annual return of 16.46 percent versus 14.17 percent from a stock-only buy-and-hold strategy. This result is economically significant. Some researchers have suggested that the stock market returns from being able to predict turning points in the business cycle would be enormous. The unanswered question has been how investors can predict economic turning points with any precision. Recently, however, a probit model has been developed to predict, based on the shape of the yield curve, four quarters in advance whether the U.S. economy will go into a recession.Combining these two areas of research, we suggest that it may indeed be possible to time the stock market. Specifically, if the yield spread is useful in forecasting turning points in the business cycle four quarters in advance and if an investor can use that knowledge to carry out a viable stock-market-timing strategy, then conducting a stock-market-timing strategy eight months after a strong signal is received, by way of the magnitude of the yield spread, might beat a stock-only buy-and-hold investment strategy.We tested this idea by conducting an out-of-sample simulation based on an eight-month lag between the yield spread and the stock market. The results were disappointing. We were successful, however, in developing a probit model for forecasting downturns in the stock market one month ahead based on the yield spread as the explanatory variable.The basic data we used are the monthly S&P 500 Index levels and the yield spread between 10-year+ T-bonds and three-month T-bills. A bear market was defined as six or more consecutive months of a generally declining stock market.To determine the profitability of using the probit market-timing strategy versus a simple stock-only buy-and-hold strategy, we performed out-of-sample simulations. We obtained total monthly returns, including dividends, for the S&P 500 for the period January 1971 through December 1999. We also obtained corresponding 30-day T-bill returns. Using the out-of-sample time series of probabilities obtained from our probit model, we simulated a market-timing strategy beginning in January 1971 of investing 100 percent in an S&P 500 mutual fund if the probability of a bear stock market one month later was less than X percent or investing 100 percent in T-bills if the probability of a bear market was greater than X percent. In each successive month through December 1999, the investment position was reassessed and changed in accord with the probability of a bear market one month later. In the simulations, we used three probability screens (or values of X)—30 percent, 40 percent, and 50 percent.Each of the probit market-timing strategies outperformed the buy-and-hold strategy. The terminal value of investing $1 over the 29-year period in a stock-only buy-and-hold strategy would have produced $46.71 and an annual compound return of 14.17 percent. When a bear market probability screen of 50 percent was used, the terminal value was $83.02 and the annual compound return was 16.46 percent. Thus, our probit market-timing strategy yielded an extra $36.31 in terminal value or, on an annual basis, an additional 2.29 percentage points in annual return.Moreover, the strategy based on our probit model produced a significant timing coefficient when we applied the model for statistically evaluating market-timing ability. Thus, the probit market-timing strategy produced a larger mean excess return than a stock-only buy-and-hold strategy at a lower level of total and systematic risk than is inherent in the market portfolio.Finally, the parsimonious nature of our model makes it particularly attractive for use by market timers.