Are European stock market returns predictable? In their September 2012 paper entitled “Forecasting Returns: New European Evidence”, Steven Jordan, Andrew Vivian and Mark Wohar test the ability of fundamental, macroeconomic and technical variables to predict next-month returns in 14 developed and emerging European country stock markets both in-sample and out-of-sample. They consider four fundamental variables (using logarithms): dividend-price ratio; dividend yield; earnings-price ratio; and, dividend payout ratio (dividend-to-earnings). They consider two macroeconomic variables: the risk-free rate; and, variance of weekly stock market returns over the last 52 weeks. They consider two technical variables: monthly price pressure (ratio of number of rising stocks to number of falling stocks); and, monthly change in volume of all stocks. They test predictive power via simple linear regression, with a rolling historical window of 60 months for out-of-sample tests. They use the historical average as a benchmark forecast. To assess the economic value of forecasts, they examine whether portfolio allocations based on regression outputs beat those based on the historical average. Using monthly data for 14 European/Mediterranean stock market indexes during January 1995 (so out-of-sample tests begin in 2000) through December 2011, they find that:
- Fundamental ratios exhibit little or no out-of-sample predictive power or gross economic value.
- Macroeconomic and technical variables often beat the historical average as predictors both statistically and economically. The risk-free rate and price pressure each beat the historical average in about half the countries. In 10 of 13 countries (not enough data for Cyprus), at least one of these variables outperforms the historical average.
- Restricting macroeconomic and technical variables to ranges that “make sense” does not improve their predictive power.
- Fundamental variables consistently work better in liquid rather than illiquid markets. Macroeconomic variables work better in developed rather than emerging markets.
- The average of all simple (single-variable) models consistently (11 of 14 countries) outperforms the historical average benchmark.
In summary, evidence suggests some economic value from forecasting European stock market returns with individual macroeconomic and technical variables; the average forecast from all eight predictors most consistently outperforms the historical average.
Cautions regarding findings include:
- The sample period is very short compared to variable construction intervals.
- The study uses indexes rather than tradable assets. Costs of constructing tradable assets (trading frictions and management fees) would reduce returns, perhaps differently across markets.
- The study apparently ignores any trading frictions associated with entering/exiting indexes to estimate economic value. Incorporating these frictions would reduce performance.
- Applying different rules to the same individual markets, and the same set of rules to individual markets introduces data snooping bias, such that the best results incorporate luck. The fact that different predictors beat the benchmark in different markets undermines belief that any specific predictors link to fundamental principles of investment or human behavior.
- A monthly forecast horizon may not fit the hypothetical effectiveness of specific variables.
- Random timing may be instructive as an alternative benchmark (see “The 2000s: A Market Timer’s Decade?”).
