Forecasting the equity risk premium

(MoneyWatch) There are many academic studies on the ability of macroeconomic variables (such as dividend yield, earnings-to-price ratio, volatility of stock prices, book-to-market ratio, and inflation) to predict the equity risk premium. However, very little attention has been paid to technical indicators, such as moving averages and momentum.

The likely explanation is that there's no economic reason for technical indicators to have predictive value. Thus, academics are suspicious of any research -- it could be the result of data mining, random outcomes that have no out-of-sample predictive value.

To remedy this situation, the authors of one paper analyzed the forecasting ability of technical indicators and compared it to the forecasting ability of popular macroeconomic variables. They sought to answer the following two questions:

• Do technical indicators provide useful information for forecasting the equity risk premium?
• Can technical indicators and macroeconomic variables be used in conjunction to improve equity risk premium forecasts?

They then compared the value of such forecasts to forecasts based on the use of the historical average equity risk premium. The following is a summary of their findings:

• Monthly equity risk premium forecasts based on technical indicators produced economically significant statistics and frequently outperform forecasts based on macroeconomic variables -- 12 of the 14 individual technical forecasts outperformed the historical average forecast, and eight of the 12 produced statistics that are significant at conventional levels (5 percent).
• Out-of-sample forecasting gains were highly concentrated in recessions for both technical indicators and macroeconomic variables -- they did superior job of forecasting the equity risk premium than using the historical premium.
• Technical indicators detected the typical fall in the equity risk premium near business-cycle peaks.
• Nine of the 14 macroeconomic variables examined outperformed the historical average benchmark forecast, but only three of the nine were significant at the 10 percent level -- there was still a 10 percent chance the outcome was random. Dividend related metrics produced the best result.
• Both approaches seem useful for predicting returns, and they appeared to complement each other. Combining them outperformed any of the forecasts based on individual technical indicators or macroeconomic variables.

In his paper, "Equity Risk Premiums (ERP): Determinants, Estimation and Implications," which included a survey of the research on the ERP, New York University finance professor Aswath Damodaran analyzed the predictive power of four methods:

• The current implied ERP
• The five-year average of the implied ERP
• The default premium
• The historical premium

He found that the most commonly used metric -- the historical premium -- was negatively correlated with both the implied next years' premium and the actual risk premium over the next 10 years. As a forecasting tool, historical premiums fail, and using macroeconomic models has provided superior results.

The results of the study on the role of technical indicators provide evidence that perhaps using technical indicators can improve results, especially at peaks and troughs. That said, the evidence makes clear that estimating the ERP isn't a simple task. If it were easy, we would be able to do what no one has yet been able to do persistently well -- forecast stock market returns. Thus, when estimating the ERP we should have a healthy skepticism as to the accuracy of forecasts and be sure to treat the estimated only as the mean of a wide potential dispersion of potential outcomes.

Image courtesy of Flickr user yomanimous