A Look at the Different Approaches to Factor Timing

Factor timing has always been one of the most alluring areas of investing for me. It is very clear that all factors have periods where they go in and out of favor, and if it possible to time those periods to identify the factors to invest in (and the ones to avoid), that could be a significant source of alpha.

But despite its allure, the process of timing factors is much more difficult in practice than it might seem in theory. It not only requires a system to identify the factor to over or underweight. It also requires a system to get the timing right enough to take advantage of the signals. For example, it might have been easy to identify that value was cheap in 2018, but it was much harder to time an entry point to take advantage of that given that its underperformance continued for a long time after that.

Given how difficult factor timing is, I think any discussion of the methods that can used to do it should be prefaced with two major questions:

[1] Can Factors Be Timed?  

There has been substantial research done in this area with mixed results.  AQR looked at timing using valuation in their paper 2017 “Contrarian Factor Timing is Deceptively Difficult” and concluded that it is much harder than it looks. They found better results using momentum in their 2019 paper “Factor Momentum Everywhere”.  Research Affiliates found strong results using valuation in their 2016 paper “Timing “Smart Beta” Strategies? Of Course! Buy Low, Sell High!”. And in their 2021 paper “Factor Timing: Keep It Simple”, they found that a system that combines value and momentum works best and were less optimistic about factor timing based on macroeconomic variables.  

So can factors be timed? I think the best answer I can come up with based on the research is “probably, but its complicated”.

Which brings me to the second important question.

[2] Should Factors Be Timed?

On the surface, it would seem the answer to these two questions should be the same. After all, if factor timing can increase returns, then why wouldn’t you do it? The answer to that lies in the difference between the academic world and the real one. In academic research, an investor always follows their investment strategy. They never panic when it loses money. They never alter their approach when they underperform.

But the real world doesn’t work that way. And the one thing you see in the factor timing results across the board is that sticking to these types of strategies is very hard. To understand why, think back to my value investing example from earlier in the article. Value became cheap well before it showed any significant outperformance (and its underperformance still dwarfs the bounceback). Although value investors saw strong results in the wake of the latest bear market, investors trying to time it had to sit through a lot of pain to get there. Most investors have a difficult time doing that. And this is true of the other approaches to timing as well.

The Three Major Timing Approaches

If my attempt to rain on your factor timing parade in the first half of this article didn’t deter you, I will use the second half to look at the different approaches in more detail.

Value

Value is probably the most widely used approach and the one that makes the most sense to many investors because it involves trying to buy low and well high. Generally, factor timing using value is done by looking at valuation spreads. Spreads are just the gap between the cheapest group of stocks in the market and the most expensive group. When that spread is wide, value is seen as cheap and exposure is added to it. And the same is true of other factors.

This can be used to build a portfolio in different ways, but we generally prefer to always maintain exposure to multiple factors. For the reasons I talked about earlier, going all in on one factor can be dangerous and very difficult behaviorally. We also like to use a composite of all the major value factors to limit our exposure to a single factor. In the history of our factor timing models, which goes back to 2006, value has worked as a timing factor, but the additional performance it has generated has only been marginal.

Momentum

Factor timing using momentum is exactly what it would seem. Additional exposure is added to the factors with the best intermediate-term momentum (most systems I have seen use something between 3- and 12-month momentum).  Just like with value, with our system we use multiple time frames to limit the risk that we pick the wrong one, and to smooth returns. Momentum is probably the most supported approach to factor timing in the academic research and we have found the same thing in our models, with momentum outperforming both value and macro factors by a fairly wide margin.

Macroeconomic Factors

Using macroeconomic factors can be more complicated than the other two. It requires that two different decisions be made correctly. First, you need to identify which factors work best in which environments (i.e high growth, inflation, recession etc). Second, you need to identify which environment you are in at any given time, which can be more challenging given that economic data typically has a significant delay to it. For our system, we used two primary indicators. The first is a quadrant system similar to that used by other practitioners that tries to identify where we are from both a growth and inflation perspective by looking at the rate of change in GDP growth and inflation.

The second is a series of more one-off type indicators that look for certain situations that favor a specific factor. For example, Verdad has found that high yield spreads can be used as a predictor of good times to invest in value. Or research has shown that the change in the slope of the yield curve can help to identify favorable conditions for value. The end result of all of this, however, tends to also support what the academic research has found, which is that timing factors using macroeconomic data is probably the least successful approach. The model we track that only uses these factors has underperformed the market since 2006. But we have found that when used in a composite along with value and momentum, macroeconomic factors can enhance returns.

The Challenges of Factor Timing

When I ask other factor investors I talk to about factor timing, I typically get two responses. First, they talk about the research that shows that factors are typically very difficult to time. But second, they also talk about how they actively do it in their own portfolio. And I do the same thing. So it is hard for me to sit here and discount the approach when I have used it myself throughout my investing career. But what my experience tells me is the same thing the academic research shows. And that is that the benefits of factor timing are likely small, and the challenges of implementing the strategy in the real world are likely large. For investors who can sit through long periods of underperformance, it may make sense, but for most investors, the benefit may not be worth the cost.

—

Originally Posted on October 13, 2021 – A Look at the Different Approaches to Factor Timing