Does P/E10, current real (inflation-adjusted) level of a stock market index divided by associated average real earnings over the last ten years, usefully predict stock market returns for developed stock markets other than the U.S.? In their March 2012 paper entitled “Value Matters: Predictability of Stock Index Returns”, Natascia Angelini, Giacomo Bormetti, Stefano Marmi and Franco Nardini test the ability of P/E10 to predict future returns for 12 developed country stock market indexes: Australia, Belgium, Canada, France, Germany, Japan, the Netherlands, Norway, Sweden, Switzerland, the UK and the U.S. Using monthly stock market index levels, aggregate price-earnings ratios and consumer price indexes during January 1871 through March 2011 for the U.S. and during December 1969 through December 2010 for other markets, they find that:
- Over available sample periods, P/E10 predicts future long-run gross returns in some developed markets other than the U.S. Predictive power generally increases with forecast horizon. Specifically, P/E10 exhibits:
- Impressive predictive power with R-squared statistics exceeding 80% at long forecast horizons for Belgium, France, Germany, Japan, the Netherlands and the UK.
- Overall good predictive power for Norway, Sweden and Switzerland, but with inconsistent results at long forecast horizons.
- Very poor predictive power, with modest or negligible R-squared statistics for Australia and Canada.
- Predictive power is generally robust for both monthly average and cumulative returns.
- Incorporating the long term interest rate (10-year Treasury note yield) for the U.S., as suggested by the Fed model, reduces prediction accuracy.
In summary, evidence indicates that P/E10 predicts future stock market returns in most developed markets.
Cautions regarding findings include:
- It is not obvious that P/E10 has a stable, predictable average. Consequently, an investor operating in real time may well set different thresholds for high and low P/E10 at different times, confounding use of P/E10 for market timing. See “P/E10 and Future Stock Market Returns”, which includes investigations of out-of-sample exploitability of the in-sample predictive power of P/E10 with respect to nominal U.S. stock market returns.
- As noted in the paper, sample periods are short for markets other than the U.S. in terms of number of economic cycles and, especially, in terms of independent ten-year measurement intervals in each country. Evidence across countries is not multiplicative because of correlation of economic and market behaviors during 1970 through 2010.
- The study employs indexes rather than tradable assets. Findings may be different for real assets with trading frictions (which vary considerably over time) and management fees.
