Factor anomalies 1900 to 1925

Long-term readers will know that one of my key tenets about investment returns is to try to falsify findings out of sample. This can involve testing an anomaly observed in the US in other regions or countries, or testing it out of sample with data that has not been used to identify the anomaly. Another approach is to utilise ancient historical data that wasn’t previously available and predates the data used to identify an anomaly.

Fabio Braggion, Joost Driessen, and Lyndon Moore combined two of these three approaches to test whether the momentum, size and low volatility factors are real. They also tested the short-term and long-term reversal factors and betting against beta. This was done by examining the performance of these six factors in nine international stock markets for the period 1900-1925, i.e., before the famous Ibbotson data begins in 1926, which forms the core of US factor analysis.

I am not particularly interested in betting against beta or the long- and short-term reversal effects, so I will disregard them here (in any case, the study finds that they do not exist before 1926). I am more interested in the factors that investors more commonly implement: momentum, size, and low volatility.

The chart below shows the results for these three factors for seven out of the nine markets. The results for Germany and Austria are excluded due to the hyperinflation that affected both countries following World War I. However, the paper discusses these two countries in detail and finds that before World War I, both countries showed similar results to the other countries. Over the whole period (1900 to 1925), results were qualitatively the same despite the massive disruption caused by hyperinflation.

Momentum, size and low volatility factor returns 1900 to 1925

Source: Braggion et al. (2025)

To me, the key takeaways from this study are that momentum worked extremely well even before 1926, which reinforces the findings of this study. But the evidence for small-cap outperformance vs. large caps is more mixed. Small caps outperformed large caps by a wide margin in Amsterdam and London but underperformed in New York. In general, it seems to have worked with five out of seven markets showing statistically significant outperformance of small caps. Finally, when it comes to the low volatility factor, four out of seven markets show a significant outperformance of low volatility stocks over high volatility stocks. Still, three markets do not, providing only a marginal confirmation of the low volatility effect in the early years of the 20^th century.