To test whether variations in earnings-price correlation have any predictive value for stock returns, we ran regressions of correlation levels versus trailing annualized returns.
The R² between the S&P Composite’s earnings and the worth from 1871 to 2024 may be very high at 0.95. Given the strength of this long-term relationship – and the relative rarity of periods of low correlation – it is cheap to ask whether these periods could act as buy or sell signals. In other words, do variations in earnings-price correlation help predict future returns?
I evaluated this query over several rolling time horizons. The resulting R² values ​​– which link correlation levels to subsequent annual returns – were far lower than the R² between earnings and price itself. For the rolling 10-year and five-year windows, the R² fell near zero, suggesting that there may be virtually no predictive relationship.
The rolling 50-year period showed the strongest relationship with an R2 of 0.53.
