David Blitz, Robeco’s Chief Researcher, provides insights into factor returns across various market conditions in his latest study.
- Research on 150+ equity factors reveals long-term positive premiums.
- Positive-beta factors exhibit no alpha; others perform best in bear markets.
- Study confirms seasonal, momentum effects, enabling robust investment strategies.
Focusing on over 150 equity factors, Blitz uncovers steady alphas post-beta adjustment and reaffirms seasonal and momentum effects. These insights into factor behavior aid the development of more robust factor-based investment strategies. We sat down with him recently to discuss the paper: ‘The cross-section of factor returns’.
Natural curiosity
“For many years, as researchers we’ve been using the Fama-French factors which they provide on their website with a long history. These factors have been examined to death basically. They include the size factor, the value factor, the momentum factor and a couple more they added recently. There's so much research done on them that by now we understand these factors very well.”
"It's like being in a candy store. Before, you could eat five flavors, and now there are 150 boxes with all this candy, and they all have their own special taste.
“But in the meantime, the academic literature has suggested all kinds of factors. Now there's a whole zoo with hundreds of proposed factors. Several academics have made these factors publicly available. Instead of the classic three, four, or five factors, we now have libraries containing over 150 factors. It's like being in a candy store. For years you could only choose between five flavors, while now there are suddenly over 150 with so much variety in taste.”
“So, there’s a natural curiosity on our end. Are they really 150 different factors? Turns out, not really. For instance, there are like 20 value factors. It’s just value in different shapes and forms. At the end of the day, there’s probably one overarching value effect behind this.”
Basic condition
“Also, some factors in these libraries don’t seem to belong there. The number one condition for a factor is that it gives you a better return in the long run. Some don’t meet that basic condition. Other factors appear to be misclassified, so there’s also still a lot of room for improvement there.”
“Now, if you have 150 factors, you can also look at things like which factors have high alphas, which ones have low alphas, is there anything they have in common? Which periods do they deliver? With five factors, you can have the problem that in some periods they're just very weak. But if you have 150, surely there should be some which do well in every conceivable environment.”
“Having 150 factors opens up a whole new dimension. It's at a higher abstraction level. If you just look at individual factors, there's already some thinking behind it. But when you build strategies on top of factors, you’re operating at another layer.”
Fama-French: mature or obsolete?
“Also, the classic Fama-French data library that has been used for decades might be becoming a bit obsolete or even mature. It's still a benchmark, but there's so much more at our disposal now. Sticking with just this handful of factors is way too restrictive.”
“In my research, the abstract and conclusion contain one observation that’s led to a lot of questions. Most of these factors have an implicit low risk tilt, which also explains why they tend to do better in bear markets. If you look at the raw factor returns, it seems like they get most of their returns in bear markets, which is only 20% of the time. The rest of the time, their return is much weaker.”
“This is quite shocking. So, does this mean that you should only do factor investing if you think we're heading towards a bear market? Usually, investors have a bullish outlook. If they're really pessimistic, they wouldn't be investing in equities in the first place. The findings are strong and statistically solid, but we have to be careful with the implications. We’re not saying here that you should only invest in factors if you expect a bear market.”
Not an argument against factor investing
“The paper shows that if you see a factor with a low beta, you know it's naturally inclined towards low beta stocks. But you can correct this in advance. There's also an analysis that indicates if you adjust for beta, you can virtually eliminate the sensitivity to bull and bear markets. There are two graphs, exhibit 8 and 9, that demonstrate this. One shows a substantial difference in raw return during bull and bear markets, but when you look at the beta-adjusted return, it's much more stable.”
“What this means is that you shouldn't just blindly implement factors. You need to be mindful of the sensitivity to market conditions, which can be significant, and perhaps make adjustments before you implement a factor. We manage these risks carefully by having various checks and balances in place. So, this is not an argument against factor investing, but rather against a naive implementation of it, as that could lead to unintended risks in your portfolio. We neutralize those risks and carefully manage them. It's important to have an understanding of this before going ahead.”
“Something else I’d like to mention is that while the factor library contains 150 factors, effectively there might be only about 15 or 20. This is because there are many variations of the same theme. Moreover, all the factors in this library are based on historical data like prices or financial statements, and it doesn’t even include analyst-based data or alternative data sources. Even though the library sounds extensive, it's still a subset of what’s available. I would rather have a more diverse set of 100 factors than 150 that are quite similar.”
“Having said all that, I do want to commend the academics who have provided these datasets. They've done the industry a great service by taking the initiative to create these resources. However, it's important to recognize that this is just a first step and there’s still much more that can be explored. The key takeaway is to be cautious and not to implement factors blindly.”
Read the Full Paper Here
