Skip to content
All library documents

Predicting Market Returns and the Weakness of Out-of-Sample Evidence

Article Quant Q&A · Author: emcor
Read the original

Summary

The document asks which models can estimate or predict broad market index returns, contrasting this task with the many factor models used to explain individual stock returns. It names cay, dividend-based measures, and average correlation as candidate predictors. The response agrees that such variables may help predict stock market returns, while emphasizing that their out-of-sample evidence is generally weak.

The note points readers toward a survey of equity premium predictors by Goyal and Welch, but does not describe the candidate variables’ construction, compare their forecasts, or report specific performance statistics. It therefore offers a caution about the gap between apparent predictive relationships and robust forecasting rather than a complete estimation procedure. Any use of these predictors requires attention to out-of-sample validation and the possibility that historical predictive patterns will not persist.

Key ideas

  • Market return prediction is distinct from estimating models of individual stock returns.
  • Cay, dividend measures, and average correlation are mentioned as candidate market return predictors.
  • The response characterizes out-of-sample evidence for such predictors as weak.
  • The document points to a survey of equity premium predictors without detailing its methods or results.

Tags

EquitiesStatisticsFactor investingUS markets
Full text
# Estimate market risk premium?


# Estimate market risk premium?












There are uncountably many factor models to estimate stock returns, such as CAPM, Fama-French, Carhart-Momentum, APT etc.

Which models can estimate the market (index) return?

I found only three models: Cay, Dividends and Average Correlation.

## Answer by phdstudent (score 3)

https://quant.stackexchange.com/a/18514

Yes, those are probably the variables that predict the better the stock market return. However, the OOS evidence is usually weak.

Goyal & Welch provide a good summary on predictors: http://rfs.oxfordjournals.org/content/21/4/1455.abstract

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.