Global factor premiums (2021)
Guido Baltussen, Laurens Swinkels, Pim van Vliet
Journal of Financial Economics 124(3), 1128-1154, URL
This week’s AGNOSTIC Paper is another out-of-sample test of the major factors (Momentum, Value, Low-Risk) and goes even further back in time than the last one. The authors examine the major factor premiums among equity indices, government bond indices, currencies, and commodities in a sample that ranges from December 31, 1799 to December 31, 2016. So this is not only an out-of-sample test with respect to the sample period, but also with respect to asset classes.
Apart from this novel dataset, the authors also examine the statistical validity of the factors and analyze the major economic explanations behind them. While all of this is important and interesting, it is beyond the scope of this post. Therefore, I will cherry-pick the (in my opinion) most important insights and refer to the paper for everything else.
Given that the Setup and Idea is basically the same as last week – testing the major factors out-of-sample – I skip this part here. To get all information, I therefore recommend to read last week’s post first.And the previous five on factor investing as well, of course.
Everything that follows is only my summary of the original paper. So unless indicated otherwise, all tables and charts belong to the authors of the paper and I am just quoting them. The authors deserve full credit for creating this material, so please always cite the original source.
Data and Methodology
To construct their sample, the authors combine data from various sources including Bloomberg, Datastream, Global Financial Data, the Chicago Board of Trade, and other research projects.For example, the “Return on Everything” dataset which I have also written about. For definitions of the major factors, they consult peer-reviewed literature and construct the factors as transparent as possible. There are few surprises in this respect. Momentum is the standard 12M-1M return, value is price/book for equities and some variation of long-term reversal for other asset classes, and low-risk is the Betting Against Beta methodology.
Note that the authors also examine three other well-researched patterns: time series momentum a.k.a. trend following, carry, and seasonal effects. Given that those are less relevant within equities, I haven’t included them in my series on factor investing (which doesn’t mean they are not important…).
There are countless details about the data collection process and the authors provide a comprehensive overview in the appendix. For the sake of this post, however, I think it is sufficient to understand that they created a long and broad sample of the following asset classes and geographies:
- Equity Indices: the authors “[…] cover the major developed markets […]”.
- Government Bond Indices: the authors “[…] consider the major developed bond markets around the globe.”
- Currencies: the sample includes the G10 countries’ currencies versus the US dollar.
- Commodities: the sample covers more than 30 of the most important commodities.
Overall, the data is (in my opinion) pretty cool. Even though not on the single-security level, the very long history allows to truly test the major factors out-of-sample, among other asset classes than stocks, and across the world. In that sense, this paper is therefore an even more robust test than last week’s long history of US equities.
Important Results and Takeaways
Momentum, Value, and Low-Risk “worked” globally, in different asset classes, and out-of-sample
The following chart summarizes the key results of the paper.For my interpretations, only the bars matter. The dashed lines are different thresholds for statistical significance. If the bar is above the line, the Sharpe ratio is statistically significant. Panel A shows Sharpe ratios of factors for the sample periods used in their original publication. Depending on the factor, that’s some period between 1969 and 2012. Panel B are Sharpe ratios for the same sample-period, but with the systematized methodology from the authors. You may have noted that the two charts look fairly similar. This is already a first success because it shows that the factor premiums are mostly robust to slightly different methodologies. However, the average Sharpe ratio somewhat decreased from 0.50 to 0.41.
Panels C and D are now the out-of-sample tests with “new history”. Panel C shows Sharpe ratios for the period between 1800 and the start of the sample used in the factor’s original publication. Once again, the picture looks very similar to the original results and the average Sharpe ratio remains around 0.40. Panel D finally reports Sharpe ratios for the entire 200+ year sample period from 1800 to 2016.
These are of course a lot of numbers, but I think the overall pattern is quite clear. Most of the factor premiums remain robust for different methodologies, regions, and sample periods. I repeat myself, but this is exactly what we would like to see to ensure that Momentum, Value, and Low-Risk are not the result of data mining.
There is little evidence for factor decay
Similar to last week, the authors also touch on the issue of factor decay, the effect that strategies become less profitable after publication. To test for this, the authors report the difference between Sharpe ratios in the out-of-sample period and the original sample period in Panel C of the following table. Again, there are a lot of numbers. But if you look at the significance stars in Panel C, you will find that there is no significant decay for almost all of the factors. Panels A and B further support this as the Sharpe ratios of most factors are comparable between the Pre- and Full-sample period.
Conclusions and Further Ideas
You know what is coming, right? This week’s paper is yet another piece of evidence that the major factors (Momentum, Value, Low-Risk) come from real economic mechanisms and are unlikely the result of data mining. Different methodologies for the same ideas, different asset classes, different markets across the globe, more than 200 years of data, and we keep finding the same overall patterns over and over again. Cheap securities with low risk and positive momentum tend to outperform over the long term. Next week, I will wrap up this series on factor investing and show you one guy who understood and applied those ideas much earlier than we did…
- AgPa #56: The Equity Risk Premium of Small Businesses
- AgPa #55: Backtests in the Age of Machine Learning
- AgPa #54: Transitory Inflation
- AgPa #53: Investing in Interesting Times
This content is for educational and informational purposes only and no substitute for professional or financial advice. The use of any information on this website is solely on your own risk and I do not take responsibility or liability for any damages that may occur. The views expressed on this website are solely my own and do not necessarily reflect the views of any organisation I am associated with. Income- or benefit-generating links are marked with a star (*). Please also read the Disclaimer.
|1||And the previous five on factor investing as well, of course.|
|2||For example, the “Return on Everything” dataset which I have also written about.|
|3||For my interpretations, only the bars matter. The dashed lines are different thresholds for statistical significance. If the bar is above the line, the Sharpe ratio is statistically significant.|