Swedroe: Consumption’s Impact On Returns

February 09, 2018

There is a large body of literature on stock return predictability, with most predictive variables being financial variables such as the earnings yield (E/P), dividend yield (D/P) and the Shiller CAPE 10 ratio. However, there is not much evidence in favor of returns being predictable from aggregate consumption data. Stig Vinther Moller contributes to the literature on expected equity returns with his November 2017 study, “Cyclical Consumption and Time-Variation in Expected Stock Returns.”

Cyclical Consumption & Excess Returns

The objective of the study was to determine whether movements in expected stock returns have any direct relation to cyclical consumption (consumption varies with the state of economy) and, if so, whether cyclical consumption contains information about expected stock returns that isn’t already captured by existing predictive variables. Moller found that predictive regressions of excess stock returns on cyclical consumption (CC) show there is a strong negative relation between it and future excess stock returns.

This is consistent with economic theory, in which expected returns are driven by the cyclical time-variation in the price of consumption risk—during bad economic times (such as during the great financial crisis), investors demand a larger risk premium.

The implication is that expected excess stock returns are high when CC is low during cyclical downturns. On the other hand, expected excess stock returns are low when CC is high during cyclical upswings. Following is a summary of Moller’s other findings:

  • When using real-time data, CC strongly outperforms the historical mean return as a predictor of future returns.
  • CC contains considerable information about future stock returns not already explained by labor capital and financial wealth. Together, these variables explain as much as about 20% of the variation in one-year-ahead excess stock returns.
  • Financial predictive variables such as D/P, the term structure of interest rates (the slope of the yield curve) and the spread between investment-grade and high-yield corporate bonds have relatively modest in-sample predictive power.
  • CC contains substantially more information about expected stock returns than price/dividend ratio and other popular predictive variables.
  • In out-of-sample tests, CC strongly predicts stock market returns in several developed countries. In addition, the CC slope coefficient is negative across all countries.
  • The global component of cyclical consumption generally contains more predictive information than the local counterpart, consistent with evidence of increasing financial market integration.

Moller’s results were both statistically and economically significant. He found that average excess returns following negative and positive CC values over the sample period October 1953 to October 2016 were, respectively, 10.5% versus 0.9%.

His results imply that, on average, low values of CC are followed by high returns, and vice versa. Thus, expected returns are high when CC is low in economic contractions and expected returns are low when CC is high in economic expansions.

Summary

Economic theory posits that investors require high expected returns when cyclical consumption is low in economic contractions and low expected returns when cyclical consumption is high in economic expansions. Moller’s findings are consistent with that theory.

This has important implications for investors who are prone to selling during economic contractions when consumption risk is high. If you are in that group, you need to understand that you would be selling when risk premiums are high, and, therefore, so are expected future returns. Perhaps this understanding will help investors remain disciplined, avoiding the panicked selling that dooms so many financial plans to failure.

Larry Swedroe is the director of research for The BAM Alliance, a community of more than 140 independent registered investment advisors throughout the country.

 

Find your next ETF

CLEAR FILTER