Identifying Factor Risk In Smart Beta ETFs

The head of Axioma shares a look into the business of risk modeling and risk management.

Reviewed by: Cinthia Murphy
Edited by: Cinthia Murphy

Sebastian CeriaSebastian Ceria is founder and CEO of Axioma, a company that’s one of the leaders in risk modeling and risk management. He will be talking about the many facets of risk in smart-beta ETFs at the upcoming Inside Smart Beta & Active ETFs conference in New York City this week. Here Ceria offers a preview. I’d think it’d be difficult today to model risk given so many variables, so much volatility, countries so reactionary to everything. Governments typically signal things, and there's some predictability about the outcome of that signaling, but today we don’t seem to have that type of clarity. Some say it's easier to forecast risk than returns, but do investors have the right tools to forecast risk today?

Sebastian Ceria: It’s definitely easier to forecast risk than return. Nevertheless, there are periods in time when it’s easier to forecast risk than others. And that has to do with market behavior and structure relative to the past. Ultimately, to predict risk, we have to use the past to learn about the future.

When we think about predicting risk, we tend to think about at least two very different scenarios. One is when things are “normal”; in which case, history tells us a lot, and there are a lot of tools we’ve developed to help us get a pretty good grasp of what risk is going to be in those times.

Second is times of stress. There, obviously, predicting risk ends up being more of an art than a science. You have to use history to inform yourself, but you also have to be able to build some sort of scenarios you can use for stress-testing your portfolios in those times.

For example, in the context of the Italian elections or euro—the euro crisis we're seeing right now—we're helped by the fact that there are times in the past where similar scenarios have taken place. True, it was not Italy, but we do understand what happens when the euro is in distress, because there are times in the past where the euro has been in distress. We can learn from those events and try to link what a euro distress means to other factors and how they drive the risk.

That’s what risk management is about. If you're in normal times, it's about understanding from past history how things behave. If we're in an abnormal time, we use our ability to do macro analysis to figure out which macro factors will be affected, and then use history and our learnings from the past to try to see how those macro effects that we think are going to happen are going to affect the other factors that drive the risk. In these times of stress, do investors see index investing as a way to manage risk, or as providing a layer of safety—no need to try and figure out the markets; just own them?

Ceria: I don't think the move to passive investing comes from situations like this. For example, you’d never say, “Let's buy a passive European index because I'm afraid about Italy, because ultimately, European equities are going to suffer because Italy is going to suffer.” So, indexing isn’t necessarily the refuge for situations like this.

I think indexing is much more of a wider phenomenon, which is less related to the risk environment and more related to the fact that investors are not getting from active managers the net-of-fee returns they’d expect relative to a passive vehicle. The move to passive is more a return-driven phenomenon than a risk-driven phenomenon. You're speaking at a smart-beta ETF conference, so let’s talk ETFs and risk. There are two basic models for looking at factor risk: one being through regression analysis of a portfolio relative to an index; the other by looking at individual holdings. Which way is better?

Ceria: The way we look at factor risk is by building, first and foremost, factor models that are independent of the portfolio. Those factor models are informed by a combination of the literature that’s out there, and then applying those factor models to the ETF and looking at the actual holdings. That gives us a true understanding of how those individual securities are going to drive the risk of the portfolio.

Contrary to that, regression analysis has many issues. First of all, you have to figure out what you’re regressing against—what’s the “factor exposure” you're going to be regressing against and how do you define that? Second is that those betas you find for those regressions can be quite unstable. They're not very consistent over time, which means they're not going to give you a great sense of risk.

The factor model approach—which looks at holdings and then applies the factor model to get to a risk composition of the ETF—is much more accurate in terms of telling you which factor risks are driving the risk of your portfolio. With multifactor ETFs, for example, how hard is it to tell how much risk actually comes from factors as opposed to just sector tilts? Is that a difficult distinction to make?

Ceria: You're getting down to the crux of my talk [at the conference]. The problem we have in the ETF space is that whether a fund is a single-factor or multifactor ETF is not determined by the name, but really by the portfolio construction process. We see that, although some ETFs have a single-factor name, in reality, they're multifactor ETFs.

To build that single-factor ETF is a nontrivial task. You can give it the name you want, but what we find when we analyze a variety of ETFs is that most ETFs actually have multiple-factor exposures simultaneously.

Now, for those already marketed as multifactor, we should check whether those multiple factor exposures are consistent with at least what the manufacturer of the ETFs says those factor exposures really are. There again, we frequently find that the factor exposures the ETF holds don’t always agree with what's in the marketing brochure. That’s because the portfolio construction techniques that these providers use are frequently very naïve, and they get unintended factor exposures they don't know about in the portfolio construction process. We see obviously a risk in that.

What we need much more in ETFs is to put the ETFs through the factor analysis filter to really find out what the real exposures are. If you’re an advisor, is there an easy way to do this type of due diligence?

Ceria: In this day and age, it's within everybody's reach to do this kind of precise factor analysis. How hard is it to get a blood test? You just have to go to a doctor and you have to ask for it. If you try to do it at home, you're not going to be able to get a very accurate one.

I always give the analogy of health, because many people try to tell you whether you're healthy or not just by the way you look, whether you're tanned, whether you’re overweight, etc. The reality is, to know if you're healthy or not, you have to do more work than that. Yes, you can get an initial sense, but you have to put this portfolio through more sophisticated filters to really find out what the factor risks are, what the exposures really are and whether those are consistent or not with what was advertised. Do multifactor ETFs in general do a good job at mitigating factor cyclicality?

Ceria: That all depends on the portfolio construction. There’s a general idea that one very hard thing to do is timing what you call “cyclicality.” It’s very difficult to know ahead of time when a certain factor is going to work or not. The idea of multifactor ETFs is that if you diversify your factor exposures, you're going to get certain factors that work certain times and others that work others. It’s the best you can do, because timing of factors is very difficult, right? You're diversified and you're invested in multiple factors at the same time. There are three big competitors in the risk model space—Axioma, MSCI and Northfield. Is there a real difference between them for a sophisticated investor?

Ceria: There are many differences, but let me just try to cover some of the fundamental ones. I think that Northfield, MSCI and Axioma have this concept of cross-sectional models. The basic premise of cross-sectional models is you identify factors that drive risk; you use cross-sectional regressions to estimate those models; you put a portfolio through the models and then essentially the system is going to give you the factor risk of your portfolio and a risk decomposition that tells you where risk is coming from.

We all agree with this notion that cross-sectional models work best versus other types of models, so there's not a big difference there. But the different providers differ in other things [e.g., using different parameters]. Most of the differences come in how you pick the factors. The big difference between Axioma and MSCI is that MSCI believes one model is good for everybody, whereas Axioma believes the model should be customized to your investment strategy so that the factors you have in your model are aligned with the factors you use as part of your investment strategy.

For example: If you have a momentum ETF, there are many ways to calculate momentum. When MSCI decides it's going to put together a model, it's going to pick one way of doing momentum. Sometimes some of the troubles occur because there’s a misalignment between the way the risk provider defines momentum and the way the investment provider defines it. Axioma believes in customization. That's the most important philosophical difference between us and the rest of the world. At the conference, your talk is, “Will the Future of Smart Beta Investing Resemble the Past?” What's your key forecast?

Ceria: The point here is that there’s a technique used in risk management that’s very useful, which is the notion of stress testing. You take a smart-beta portfolio, you come up with a bunch of scenarios, and then you see how the portfolio’s going to work in those scenarios. Some of those scenarios could be historical. It could also be something that you make up—say, what would happen to the portfolio if rates would go up by this much?

To do stress testing, you have to link that factor to all the other factors through historical analysis. One of the key questions is, “What do we learn when we apply that kind of historical analysis to the way those funds would behave?” Most importantly, do we see “similar funds” that behave similarly in the context of those scenarios?

This is a very helpful tool for an advisor to compare portfolios. Say you’re interested in a dividend yield fund, or in a low-volatility fund. Take two or three “similar” ETFs and see how they behave in the context of rising rates. By doing stress testing, you’ll see those behaviors among similar ETFs are quite different, and that helps you pick the right fund for your client based on what scenarios they’re most worried about. The name of an ETF is not enough.

Contact Cinthia Murphy at [email protected]

Cinthia Murphy is head of digital experience, advocating for the user in all that does. She previously served as managing editor and writer for, specializing in ETF content and multimedia. Cinthia’s experience includes time at Dow Jones and former BridgeNews, covering commodity futures markets in Chicago and Brazil equities in Sao Paulo. She has a bachelor’s degree in journalism from the University of Missouri-Columbia.