Academics have researched factor premiums for many decades, using datasets that sometimes span very long time periods. Can you explain what triggered this new research effort?
“One of the concerns frequently raised by skeptics of factor investing is that while factors may have been extensively documented by academics in the more recent period2, we can’t really be sure they will persist in the future. The main reason is p-hacking, which implies that several academic findings are the result of pure chance. If this is indeed the case, then targeting explicit exposure to factors may not necessarily lead to higher risk-adjusted returns in the long-run.”
“So, for our clients, it is very important to know which factors truly exist and offer attractive expected returns. One powerful way to determine whether factors are just a statistical blip or a result of the data mining of relatively recent data sets is to study their existence over a very long period. This is in essence what we did, going back to 1800 on the major global markets and studying factor portfolios.”
OK, but data mining concerns relate not only to the relatively short length of the time series that are typically analyzed by researchers in finance. How did you address this issue?
“To avoid objections on this front, our approach was to limit the degrees of freedom you have as researcher and consider the six most extensively documented factors in key peer-reviewed scientific journals. We replicated these anomalies and studied data spanning two centuries to find evidence of their existence in equity indices, government bonds, currencies and commodities.”
Ontdek de nieuwste inzichten via onze maandelijkse e-mailupdates
Ontvang onze Robeco-nieuwsbrief en ontvang als eerste de nieuwste inzichten of bouw het groenste portfolio op.
How did you compile such an impressive data set?
“We actually used a combination of different data sets. Historical data on financial markets dating back several centuries is very difficult to come by. Compiling this kind of information requires intensive manual work – scanning, checking, transcribing and cross-referencing with data from archival sources.”
“For this study, Robeco sponsored Erasmus University Rotterdam to obtain historical data from various sources to conduct academic research, and made it available to the Erasmus University community. We then combined this data with information from more conventional providers such as Bloomberg and Reuters.”
“
We found convincing evidence of most factor premiums across almost every decade in the period we analyzed for all asset classes.
What would you say has been your most prominent finding?
“Without any hesitation, I would say the very high level of persistence of factor premiums. Based on earlier research on recent data sets, we were expecting these premiums to be persistent over time. But our results clearly exceeded our initial expectation. In particular, we found significant, persistent and robust momentum, value, seasonal and carry premiums within the four asset classes considered.”
“The evidence supporting the existence of the factor premiums we considered in this study is not limited to any sub-period. We found convincing evidence of most factor premiums across almost every decade in the period we analyzed for all asset classes. We also identified the existence of significant factor premiums regardless of the market and economic contexts, in bull and bear markets inflationary vs. non-inflationary periods, as well as during periods of economic boom and recessions.”
“Our results also show that a diversified multi-factor multi-asset portfolio can deliver very stable returns over very long periods of time. This is important as it reinforces our belief that factor premiums are a permanent feature of financial markets rather than just another passing market phenomenon.”
Did your results provide any insights into the underlying causes of factor premiums?
“Well, one of the main conclusions of our work is that the factor premiums we considered are definitely not compensation for risk. This reinforces the behavioral hypothesis explanation for these anomalies. It is also consistent with previous research by Robeco and academics which finds that factor premiums are not necessarily compensation for taking higher risk.”3
From this perspective, multi-asset factor investing seems like a fairly straightforward style that could be implemented relatively easily, through smart beta equity and fixed income products, for example. Is it really that simple?
“Not quite. For one, harvesting factor premiums within and across multiple asset classes requires a much more sophisticated approach than the one generic factor solutions typically apply. To achieve consistently strong returns, you need to remain well diversified across asset classes, markets and factors – at all times. And to ensure good diversification, you really need a holistic approach. A combination of several generic factor solutions will not provide that.”
Footnotes
1 Baltussen, G., Swinkels, L. and Van Vliet, P., 2019. “Global Factor Premiums”, working paper available on SSRN
2 See for example: Dimson, E., Marsh, P. and Staunton, M., 2017. “Factor based investing: The long term evidence”, The Journal of Portfolio Management, Vol. 43, No. 5, pp. 15-37.
3 See for example: De Groot, W. and Huij, J., 2018. "Are the Fama-French factors really compensation for distress risk?", Journal of International Money and Finance, Elsevier, vol. 86(C), pp. 50-69.