There is a lot of research that dice and dice stock returns. However, most analyze the same data spanning the 1960s to the present day. This is why recent research on the history of the market before 1926 is enlightening.
A new study breaks new ground by examining an older period (1866-1926) than most researchers estimate. It still provides six decades of relatively new data on almost a total of 1,500 stocks, with around 200 individual stocks in each of the individual decades.
This helps with out-of-sample testing. This is useful because with current computing power levels, data can be easily manipulated. With today’s technology, it can be easy to find relationships that look statistically impressive, but lack predictive meaning.
Looking outside the original period can be a good way to validate or contradict a theory. The research is conducted by researchers at Robeco Quantitative Investments and titled Cross-section of stock returns before 1926 (and beyond).
So looking at this new data set, what do we find? Many of the more robust relationships we’ve seen in the typical more modern data set historically hold, although the strengths of the factors change.
A value-driven approach has historically contributed to investment returns over time, on average. This also holds in this earlier period. Here, this analysis is done on the basis of dividend yields, showing that stocks with higher dividends perform better over time than the wider market. In fact, this effect is stronger than in more recent periods. Other factors also hold.
Momentum, which is the concept that stocks that have risen in price over the past few months continue to perform above average in the following months. However, the dynamic effect seems weaker in this earlier period than with more recent data.
The inversion also holds. Related to momentum is the idea that stocks with extreme upward or downward reactions in price tend to rebound over short periods of time. Again, the inversion has a smaller impact on the older data period than on the more recent data.
The size effect is that smaller stocks tend to outperform larger ones. Again, this is evident in the earlier data period, although, as with more recent history, this has a weaker effect on market returns than other well-known factors.
This is subject to the usual problems with data-driven analysis in that returns can always be wrong and markets change over time, especially as trading costs decrease and the flow of information. is improving. That said, these data provide a useful additional perspective on factors that have held up well in the United States and abroad.
Even as the market changes over time, it appears that many of these factors, perhaps related to human nature, may persist over the long term.