May 13, 2021

Episode 149 – Professor Robert Novy-Marx: The Other Side of Value

Today’s guest is Professor Robert Novy-Marx, the Lori and Alan Zekelman Distinguished Professor of Business Administration at Simon Business School of the University of Rochester. Professor Novy-Marx is best known for his articulation of the profitability factor and has also done a ton of great work on momentum and low volatility. We kick our conversation off with Professor Novy-Marx’s thoughts on how profitability should inform portfolios. From there we hear why Professor Novy-Marx has a problem with evaluating the performance of a multi-signal strategy the same way that we would a single-signal strategy. He then talks about the trade-off between concentrated versus diversified factor exposure for capturing premiums. Next, we discuss why there is no good empirical evidence that we can time premiums. Professor Novy-Marx makes a great argument for why the regressions people use to say that the value spread works to predict the value premium can’t be taken seriously. Our conversation moves to focus on how our guest defines price momentum and what drives it, and the nuances of investing in momentum. We then hear his perspectives on the low volatility anomaly and how profitability helps to explain it. After that, we talk about whether investing in a low-vol fund is a way of accessing value and profitability, and why the five-factor model is a trustworthy factor model for regular investors. In the last part of our conversation, we talk to Professor Novy-Marx about his approach to critiquing other methods before ending off with his definition of success. Tune in for this excellent evergreen conversation.

 

Key Points From This Episode:

  • We introduce today’s guest, Professor Robert Novy-Marx, and his work. [0:00:17]
  • The significance of the relationship between profitability and stock returns for asset pricing. [0:02:45]
  • How the risk-based story around profitability is completely counterintuitive. [0:08:51]
  • The best way to go about using profitability in portfolios. [0:12:34]
  • When to target premiums individually and then combine them after the fact. [0:14:48]
  • How profitability is different from quality. [0:16:26]
  • Risks of building strategies that draw insight from different signals to identify a premium. [0:18:10]
  • The trade-off between concentrated versus diversified factor exposure for capturing premiums. [0:23:33]
  • Whether the recent decade’s run of underperformance impacts Professor Novy-Marx’s view of the value premium. [0:25:10]
  • The vagueness of the equity premium and if it is possible to time premiums. [0:27:58]
  • How Professor Novy-Marx defines momentum and what drives it. [0:32:32]
  • Whether investors should be using momentum in portfolios. [0:38:47]
  • Professor Novy-Marx’s perspectives on the low volatility anomaly. [0:44:19]
  • Whether investing in a low-vol fund is a way of accessing value and profitability. [0:32:32]
  • Why regular investors need factor models and how to choose one. [0:51:58]
  • Whether it is reasonable to pursue factor premiums in a smaller market like Canada. [0:57:45]
  • Professor Novy-Marx weighs in on writing papers that critique other methods. [0:58:44]
  • Why Professor Novy-Marx consults for Dimensional Fund Advisors. [1:01:15]
  • Insights Professor Novy-Marx has carried over from his years as a professional triathlete. [1:01:15]
  • How Professor Novy-Marx defines success. [1:01:15]
Read The Episode Transcript:

One of your landmark papers, The Other Side of Value; The Gross Profitability Premium, uncovered the relationship between profitability and stock returns. Can you talk about your findings in that paper and why they were so significant to the field of asset pricing?

The way academics been thinking about value, it was part of our standard asset pricing model, but it was all coming from sort of the traditional Benjamin Graham price-based value, where you buy cheap stocks and you short, expensive stocks and you on average earn a difference in returns between those two, because the cheaper stocks have that higher expected returns over time. My paper was sort of incorporating in what I would call Buffet value in some sense. He’s sort of repeatedly written in his newsletters about how it was back in the day that Charlie Munger taught him that it was far better to buy a wonderful stock at a fair price than to buy a fair stock at a wonderful price. I sort of viewed the profitability-based strategies as being that flip side of value, where when you buy cheap stocks, you get the Benjamin Graham style, but where you buy the high profitability stocks, you’re getting those wonderful stocks and hopefully at fair values.

So, I kind of viewed them as similar philosophically, where in both the cases you were trying to get a lot of productive capacity at a good price for what you’re getting. But I think in terms of why it kind of took off in finance, is partly because while philosophically similar, it has a very different sort of investment solution where the traditional value strategies, a lot of the stocks you buy in them, they’re good investments, but they don’t look like great firms in some way. Whereas for the profitability strategies, it’s a good investment and you’re getting good firms. So it actually has a growth tilt, which made it really attractive to value investors as a strategy to run in tandem because it tends to outperform when traditional value underperforms and vice versa. So they tend to smooth out the performance of each other and they’re just highly complimentary.

I think they also, it helps for the academics that they both come out of the value equation. There’s an economic principle on which we can hang a hat. The fact that we know that stock prices are there expected, discounted future dividends and the value strategies are trying to find… you use low prices to try and identify stocks with high discount rates, but the problem there is that the low price stocks can also be stocks and have low expected future cash flows. So, I think profitability helps you distinguish between the two reasons for low prices. It helps you find the stocks that have low prices because they have high discount rates and avoid the stocks that have low prices because they have low expected future dividends.

I also think that people were ready for profitability. It’s not like I was the only person ever to say that we should be looking to buy profitable stocks. It’s just that when people had done the empirical test of going and looking at the effect of profitability on predicting stock returns, they didn’t find much there. There the issue was really that I think people are way too hung up on bottom line income and that to find these good predictors of future returns, or even really to find good predictors of future profitability, future bottom line earnings, you have to go up the income statement. What I call profitability is a very broad measure of profitability that is pretty close to the top of the income statement.

Again, it doesn’t matter exactly where you are. These things, if they’re really robust and they’re really there, it shouldn’t matter the exact details, but there’s a lot more information near the top of the income statement than at the bottom.

From an asset pricing perspective, I think that the addition of the profitability premium also increased the explanatory power of models. Can you talk a little bit about that? Some of the anomalies that maybe existed prior to profitability?

Yeah. All these models when they’re introduced, part of the attraction is not only that there’s a reason for them and we understand the theory behind it, but also that they can explain something that we didn’t know before. So when Fama and French introduced their three-factor model I think sort of the big example there was, there was a very well-known anomaly about long run reversals, stocks that had gone up a lot over the previous five years, sort of underperformed going forward. There was a well-known behavioral explanation, but the three-factor model really nailed it. The long run reversals, the long run losers outperform the long-run winners, because the long-run losers, the stocks that had gone down in price a lot, had value tilts and were smaller. So that was a big win for that model.

For this model, there’s lots of things that the model does well. There have been lots of things related to profitability and cashflow that weren’t quite there, that were things that people had looked at independently. Generally, including a profitability factor really helps explain those things.

I think probably the biggest stunning win for me is if you look at portfolios sorted on market data, there’s been a lot of looking at… really because the CAPM, the capital asset pricing model, our original asset pricing model doesn’t work well, pricing stocks that were different betas, a little bit of stocks tend to have a little bit higher returns than the CAPM says they should and the high beta stocks, that’s like low returns. There’s clearly a very strong loading on high profitability and low investment if you form a strategy that is overweighted low beta stocks and underweighting high beta stocks. So for me, that’s a really remarkable result of the five factor model.

So Robert, I’ve seen you speak, I think three times now. I remember having the aha moment when you explained how the risk-based story around profitability is completely counterintuitive. I think the value story for risk is intuitive, but profitability is counter-intuitive. When you say a higher profitable stock has a higher expected return, most people will say, “Well, of course it is,” but as soon as you turn that into risk, that’s not intuitive. So could you explain that to our listeners?

The factor models, the way they’re conceptualized by… the theoretical background for them, is partly about the fact that you’re getting compensation for being exposed to some risk. Risk doesn’t just mean uncertainty, it means some systematic exposure. Assets tend to have higher returns when they pay off in good times. Assets have lower returns when they pay off in bad times, because then they provide insurance.

We tend to think about the risk that the factors in the model as maybe having higher returns, because they’re more exposed to some systematic risk in the economy. That’s been a hard thing to find. It makes sense for the value stocks in the sense that I think the way they were originally conceptualized, the value premium was something for being exposed to financial distress risk. The low price stocks were riskier, they were going to do badly in bad times. That’s been a harder story to find in the data empirically. It doesn’t really make as much sense when we’re thinking about more profitable firms, because you’d think that the more profitable firms that are in the long side of the profitability strategy, might be more stable and they might provide some insurance against bad times.

For me, it’s all about the fact… I think it’s really hard to find these things to begin with. I mean, to identify what these macro effects are. We haven’t had a lot of success in academic finance with sort of macro risk factors that underlie a lot of our theory, but what is clear is that the value equation has to hold and it doesn’t depend on risk. It’s just a statement that… it’s really just the fact that expected returns are another way to talk about prices. It’s very clear in bonds that we quote crisis with the yield and higher yields mean lower dollar prices. The same thing’s true for equities as well. Prices are just as an accounting identity, their expected future cash flows discounted at the rate of return that investors require to hold the assets. There’s nothing in that statement about why they require high returns for some assets than others.

For me, the big thing about profitability is, it makes the price signal. Prices are informative about expected returns because again, prices are discounted expected cash flows and were discounted at the expected rate of return, but including profitability makes the price signal much more informative because it tells you something about the other part of that equation, which is expected future cash flows. Once you condition on profitability, on average, when you buy a low price firm, you’re buying a firm that has a higher discount rate, but sometimes there are firms with worse cash flows.

But if you look at differences in price among stocks with similar profitabilities, the price is going to tell you more about the discount rates, because the cashflow that the profitability is telling you something about expected future dividends and so it just makes prices more informative. That’s really how I’ve always thought about it. I think the title of the paper, The Other Side of Value, is sort of pointing at the fact that I’ve always viewed it as another component of the value premium, but I tend to think about it in terms of, again, making the price signal more informative, not trying to find stocks that themselves have higher expected rates of returns independently.

Maybe this shouldn’t be a surprise, but that’s the best explanation of the risk justification for the profitability.

It helps you distinguish value stocks from value traps.

We’ve talked a little bit about how profitability should inform portfolios. I mean, you gave the Buffet reference, which I think is great, but when we think about it from a practitioner’s perspective or a retail investors perspective using profitability and portfolios, what’s the best way to go about that?

I guess I already mentioned that you have to go up the income statement. Again, you really want to trade it with value. I think there’s really interesting connections between the two in that in U.S. markets, we think about how much stronger value is in the small caps than in the large caps. It’s in the large caps that the two are negatively correlated strongly. So in the large caps in the U.S. the more profitable firms are much more expensive.

Another way of saying that is, the value firms are much less profitable. So when you form a value strategy in the large cap space in the U.S. you’ve got this big profitability headwind, so trading them together really helps you identify them better independently, and also just they work better together, but you have to use a broad measure of profitability and trade the two jointly. Beyond that it’s all about implementation, and that’s really not my side of things.

These premia that we talk about in these factor models are all fairly weak in some sense and they have to be. I don’t believe that there’s sharp ratios of three lying around for everyone to exploit. These things all have sharp ratios of 0.4, like, market like sharp ratios. Those are the kinds of things that I believe in, but to get these kinds of weak effects, you really have to be very broadly diversified to get the premium without taking… You’re already going to be taking on some risks because the sharp ratios just aren’t that high. If you’re not really broadly diversified in these things, you take even on more risk. And of course you want low turnover. That’s easier to do with these sorts of strategies, because both the broad measures of profitability and the relative pricing models are both fairly persistent, but the implementation has really been the most important thing and that’s not where I’m an expert.

Sure. One of the things that we’ve seen is, what you might call a mistake is, people combining a value ETF or fund with a profitability ETF fund where when you combine them together, like you’ve been talking about, you might end up canceling out a lot of the exposure to premiums, which is why it makes sense to combine them together and building a product. Are there any cases you can think of where it would make sense to target premiums individually and then combine them together after the fact as opposed to looking at them together?

Yeah. I think from a theoretical of view, if you ignore all the real world implementation issues, the two solutions, what I would call integrated versus siloed, are actually quite similar if we’re in a world, the academics usually think of where these are linear exposures. There are differences in terms of implementation. One is if you’re going to run them separately, and this wouldn’t happen with your UTS, you have to be careful the net trades. There is going to be trading where one portfolio is buying a stock at the same time the other one’s selling it, and you don’t want to be incurring those trading costs. So the integrated solutions are usually more efficient in terms of transactions.

I don’t think this is as big an issue, but you can also get bigger tilts in long-only portfolios through the integrated solution. But I think that’s usually not a binding constraint for most people. I think probably the biggest difference and I think it’s an important one is, that it’s seams in the data like these things are not all just linear factor exposures, that is, there’s interactions between the factors that are important. The best example of this from the early Fama-French work is the small growth portfolio was clearly not just the combination of small and growth. Small growth was different than falling growth.

You see it even stronger when you look at small unprofitable growth. The small unprofitable growth portfolios are just… those stocks are way less than the sum of their parts. They just do so badly in the beta. The integrated strategies allow you to account for those interactions in a way that you can’t do with the siloed strategies.

Can you talk about how profitability is different from quality?

As soon as I started presenting the profitability paper to people in industry, I heard people saying, “Oh yeah, we already run a quality strategy.” Then they’d tell me something that was different than what I was doing and different what the last person who told me they were running a quality strategy was. For me, quality is really… I think most people would call profitability a quality strategy, but the problem for me with quality is that it’s not a well-defined thing that means the same thing to everyone. It’s sort of this general thing that I kind of view as a marketing term, more than something that’s really meaningful about investment.

I actually went and looked at where the word started getting used. The term starts getting used in the early 2000s when there are big withdraws from growth funds, because growth was underperforming horribly in the NASDAQ deflation. I think there are people who just have a growth from bend. They want to invest for not only in good investments, but in stocks that look good and the quality strategies were a way to kind of capture people with a growth bent who were wanting to get in something that hadn’t had the refin serious underperformance of the actual growth space, but you would have to call profitability a quality strategy. I don’t love the term quality because I think it’s made by marketers, not economists and money managers.

I think quality is built on a bunch of different signals too, that maybe get somewhere kind of close to profitability, but in general, are there risks to building strategy like that, that try to draw insight from a whole bunch of different signals to identify some kind of premium?

Definitely. I mean, I’ve done quite a bit of work on this and it actually came out of the talking to practitioners about quality because the first quality strategy I heard was, combinations of low earnings variability, and high return on equity and low leverage. And people generally do things like they form a Z score for each of the three strategies and they roll those up and then they test that. The problem with that is, that generally we treat the back tests of strategies like that like we would treat the back test of any other strategy, and that’s really not even close to fair to do because you have a lot of degrees of freedom and how you combined… what signals you pick and how you combine them.

So, I guess I should be clear that I have no problem with using multiple signals. What I have a problem with is, evaluating the performance of a multi-signal strategy, the same way that we would a single signal strategy. The conclusions of that research is, really that if you believe in multiple signals use them, but you have to decide that you want to use each signal individually.

The problem here is, that if I was thinking about running a quality strategy, I might write down 20 things that could be candidates for quality measures. If I back test them, on average, maybe one of them will look significant at the 5% level and the other 20 won’t. But if I take three of them and combine them however I want, I can make it look really good in the back test.

It’s as if I let 20 monkeys throw darts at the Wall Street Journal, and none of them is likely to have lights-out performance, but if I pick the best three and I diversify across their recommendations, it’s going to look great because I’m going to get almost the same performances from the best one, but by diversifying across the three set of recommendations, I’m going to have much lower variability in my performance, and it’s going to make it look much more reliable.

That’s the same kind of thing that happens when you test these multi signaled strategies. By diversifying across the signals, it really reduces the tracking error of whatever strategy you’re running, but you have this overfitting on performance a little bit. Even a little bit of decent performance with the diversification across these multiple signals can make it look statistically reliable in these back test, which ignore all the degrees of freedom you had constructing the tests.

They make me really nervous. I think they’ve gotten very popular. There are hundreds, literally hundreds of ETF products that use similar methodologies. I think the reason they’re so popular is because it’s so easy to get really strong back-tested results with the methodology of using multiple signals rolled up. We just have these incredibly powerful incentives, both in academia to find things that work and statistical significance and in the money management industry to show strong past performance to get mandates.

The fact that this methodology makes the back test look so good, unreasonably good, actually makes it attractive to a lot of people because their incentives are to show positive results. So it’s something that makes me really nervous.

Geez. Yeah. I mean, you mentioned from the quality perspective when we were talking about that, is it being a bit of a marketing term. It sounds like it’s possible for other strategies based on multiple signals to be used the same way. How should people who are not you, evaluate something that’s being sold as a product with higher expected returns as based on a bunch of signals. How does someone look at that and decide whether it makes sense or not?

I mean, a lot of the paper I wrote was kind of driving different and correct statistics, but they’re very complicated. I really do think the kind of robust simple thing you can do is when someone shows you a seven signals strategy, you want them to show you the results on how each individual signal is contributing to performance. It may be that four of them are not statistically reliable contributors to the performance. Despite that fact, if they have positive performance, you can see it. It doesn’t look like anything except maybe they got a little bit lucky, but it’s going to make the overall results look much better in the way they test the seven signal strategy, but you shouldn’t actually believe the results are that much better because it’s looking better because they’re including things that you don’t individually believe in.

You have to decide on the signals you want to use on their own. You can think about how they clay with other ones. It may be a very attractive strategy because it’s negatively correlated with something else you’re doing and provides insurance, but it still has to individually be contributing to performance by providing that insurance. If it’s not individually providing a significant contribution to the performance of the strategy, and you don’t understand the economics behind it, there’s no reason to think you should be doing anything except making the back test look better in the combined strategy. It’s really just misleading you about what you should expect going forward.

You mentioned when we were talking about profitability and capturing the premiums, that the sharp ratio is not being that attractive and therefore needing to be really diversified to capture the premium. Another thing that we see from a marketing perspective is, more concentrated products that say, “We’re really going to concentrate in this factor and deliver a more extreme version of it” kind of thing. Can you talk a little bit about that trade off between concentrated factor exposure versus more diversified factor exposure in order to capture the premiums?

I mean, I guess I think about these in terms of tracking error mostly. I guess the question for me is when you think about a concentrated premium, what sort of tracking error are we talking about? I generally believe that even with the diversification, for the sort of the typical investor, you can deliver plenty of exposure to these premia with still a large degree of diversification.

It is true though, that in a long-only product, there’s certainly limits on how much exposure you can provide in a sufficiently well-diversified portfolio. At some point, you do have to start making thinner and thinner portfolios to deliver really big premia.

Until you get to fairly large tilts, you shouldn’t expect a sharp ratio or the information ratio on the tilt they’re providing to change much, but when you do get to the really concentrated portfolios, you do start getting additional, non-factor risk. It’s sort of idiosyncratic risk for being under-diversified. To get sufficiently large tilts in long-only portfolios, you really do need to… it does negatively impact the information ratio you’re able to achieve. So, there are limits on how big you can sort of cleanly access these premia in well-diversified portfolios. Of course, the more tilts you’re taking, the more signals you’re trading to, it reduces the magnitude of the tilts you can get to other things as well.

So, here’s a question I know many listeners are asking in their mind right now about the value factor. So does the recent decade, plus or minus, run of underperformance impact your view of the value premium?

For me, not really very much. At my heart, I’m a profitability guy. It’s kind of my baby, but if you were to press me on which long-term premia, I believe most strongly in, other than the equity premium, it’s the value premium. It’s really just because prices are directly observable and the fact that low prices are signals of high discount rates, it’s an accounting identity. It’s very hard to come up with a world where low prices aren’t at least somewhat associated with higher expected rates of return. So as soon as I’m in a world where I kind of admit, and I do strongly believe that there’s differences in expected rates of return across stocks, there are kind of a value premium comes about.

Now, I’ve already said I don’t think these premia can be too large. It’s just there’s what they call good deal bounds in academic finance. There’s just, I don’t believe in these super high sharp ratios that some people claim, and it’s just the nature of arbitrage. I’m willing to believe there are lots of nickels lying around and if you are efficient at picking up the nickels, it can be worth looking for them. But I don’t think there are a lot of hundred dollars bills lying around waiting to get picked up.

What does that mean? Well, it means that if the sharp ratios aren’t that big, it means there have to be long run periods of underperformance. If not, if there were no long-run periods of under-performance of any of these risk factors, they’re not risky, then they’d be obviously good deals.

It’s the same thing with the market. We saw the whole 2000s were a lost decade for equity investors, but at the end of that run where the market was down over a 10-year period, no one was saying, “Well, the equity premium, is it gone?” I mean, we thought it was coming back. We just had a decade of under-performance and it’s the kind of thing you expect.

I always expected to live through some decade where value had a negative premium, realize the negative premium, but that doesn’t mean that I think the expected premium is gone. I wake up every morning expecting value stocks to outperform growth stocks by a couple of basis points. I expect to be right 50.5% of the time over my lifetime, but that means there are long periods where it’s not true.

Really just for economic reasons, I kind of think there has to be an expected value premium, but again, I’m not surprised that there was a long period of under-performance. It’s unfortunate, since I believe in value and invest in it, but it doesn’t make me think value has gone away.

One of the side effects of the poor performance of value over the last decade or so has been this widening value spread, which is a term that AQR uses. I don’t know if it’s a common term in academia, but the valuation difference between value and growth stocks and that is now extremely wide relative to the past. Is there any indication in your research that economic variables maybe like the value spread, maybe sunspots, I don’t know, that’s a joke from your paper, is there any indication that we can time premiums?

I would say that there’s no good empirical evidence that we can. There may be economic reasons to think that you might be able to, but you’re not going to be able to see these in the data. It’s really just a limitation we have statistically. It’s not because we’re not smart enough to figure out how to do it. It’s just that the data doesn’t lend itself to.

When I talk to investment advisors, I love to ask them, “What is the equity premium?” I get a range of answers, but almost everyone thinks it’s somewhere between 4 and 7% and 6% is what I hear most. Then when you ask people where they come to that number, they say, “Well, you look at the last 100 years and it’s been 6%.”

That’s true. If you look at 100 years of data, you see a 6% equity premia, but if you get that mean equity premia by running the regression, it also gives you a confidence bound on your estimate. It turns out that the estimated equity premium that you get by looking at the past data is 6% plus or minus the standard deviation of like two and a quarter percent.

So, if you just are a frequentist probablist, you do the frequentist statistics, you’re basically saying you’re 95% sure that the true equity premium is between one and a half percent and 10 and a half percent. You just have no idea. With 100 years of data, we can’t come close to agreeing on what the equity premium is.

Timing any of these premiums requires that you use a short sample to precisely estimate whether the premium is particularly high or negative and you want to get out. It’s something that we can’t do. The problem is, expected returns probably do move around over time, but we don’t observe expected return. We view realized returns and they’re incredibly noisy estimates of expected returns. If expected returns are moving around a little bit, then our incredibly noisy data we have on what the expected returns, we can’t get a precise estimate of what those things are. It’s not something that can be done using the data.

Now, there might be economic reasons to think that something predicts something and the value spread. There are academic papers about the value spread. Some of them claim that they have statistical evidence for it as well, but I don’t believe any of that. I think if you’re going to try and time these spreads based on these things, it has to be for purely economic reasons. You can’t rely on the statistics because the statistics will lie to you. That’s kind of what my paper was about.

We had this sort of this long literature from the 70s and 80s on using things to predict market performance, like short-term interest rates and the slope of the yield curve, stuff about market volatility. At some point, we kind of realized that none of these predictors that we found actually worked going forward. They only worked looking backwards. It’s because of some statistical problems that I’ve been talking about.

Then I saw people doing this again with other anomalies and so I sort of wrote this paper that I guess is my finance version of Jonathan Swift’s Modest Proposal, which is you just treat it seriously, but you come up with a ridiculous conclusion. I kind of show that you can predict the performance, to highly statistically reliably predict the performance of all sorts of equity market phenomena using the weather or global warming or sunspot activity, the aspects of the planets, using the same regressions people use to say that the value spread works to predict the value premium and show that sunspots are an incredibly powerful predictor of momentum and post-earnings announcement drift. I just can’t take it seriously. So it calls into question the whole methodology for me.

On the statistics side, I get it. I’ve been through the paper, which is excellent and funny in its own way. Is there an economic reason to believe that the value spread would predict anything because it is related to price?

The value spread is not just about… it’s not like when expensive stocks get more expensive and value stocks do poorly. The value spread doesn’t always widen because there’s rotation between the portfolios. Part of this depends on exactly, is the value spread being driven because maybe the value stocks have an even higher expected rate of return? But it may be something about, what sort of stocks are in the value portfolio now? And I just try and ground all my empirical work in theory, but at heart I’m an empiricist and this is not a question that I can answer with the data. It’s hard for me to take too strong a stance on this, because to me, it’s a question that’s not answerable with the data we have.

Oh that’s a great answer.

So, you mentioned price momentum, which is another item that gets a ton of attention in both academic and practitioner literature. I know we see it all the time in our literature that comes across our desk. How do you define price momentum and what drives it?

Yeah, so the most common way to measure price momentum in the academic literature is, you look at the stock performance over the proceeding year. We usually skip the most recent month because empirically we observe not momentum, but reversals at a one month horizon, which is partly due for microstructure effects, liquidity and trading effects. You don’t really see that same thing in momentum and asset classes that you see in the cross-section of individual stocks.

In terms of what drives it, this fact that stocks that have gone up a lot over the last year on average have out-performed the stocks have gone down a lot over the last year, over the coming month. That’s a $64,000 question for academics. So if anyone could give a simple, compelling answer to that question, they’d be instantly famous, at least in my narrow little nerdy circles.

But what I do think I can say, I’ve thought a lot about momentum, and what I do think I can say is, that a lot of the high average returns to the winner minus loser strategies are actually driven by, it’s called fundamental momentum, which is a much older anomaly, 20 years older than price momentum that came out of the accounting literature, which is stocks that have seen their year over year quarterly earnings increase a lot, tend to outperform… that there’s a big bump-up when they announce their earnings improvement, but they tend to continue to outperform going forward for a while.

The earnings winners have generated high average returns as have the price winners, but I think what you see in the data is, that if you look at the stocks that have had big price increases without having seen big improvements in earnings, you don’t see them continuing to do well. Whereas if you look at stocks that had big improvements in their earnings, but didn’t have a big price bump-up, you can’t expect them to perform well going over time. On average, when you just buy price winners, you’re buying stock. On average, the stock that have seen big increases in their price have also, they’ve gone up in price, partly because a lot of those stocks had seen improvements in their earnings. The earnings is a more robust cause of this momentum going forward than the price momentum. So in academia, people have paid a lot more attention to price momentum.

So, if fundamental momentum is driving price momentum, what are you getting if you invest in a momentum fund?

Oh, I would say that by buying the price winners, you’re getting some earnings momentum. That’s clearly in what you’re getting. You’re also getting a bunch of industry concentrations because industries do tend to do well or poorly together. It doesn’t help with your high expected returns, but it does contribute to the volatility of the strategy. The price momentum strategies are more volatile and they have occasional periods of terrible underperformance that I think are driven partly by these concentrations.

You’re also making bets on the broad market that are time-varying. So it turns out that in up markets, the stocks that are doing the best tend to be high beta stocks. In up markets, when you trade momentum, you tend to overweight high beta stocks, which hurts you when the market takes a big drop.

In down markets, the relative winners tend to be the low beta stocks. So you tend to have… the momentum strategies tend to be short the market, which also hurts you when the market turns around and goes up. So it’s not clear what that portfolio of things is, that you’re getting, but I don’t know if those are the bets you want to be making. But I mean, I think it is probably useful to momentum investors to understand that when they’re betting on price winners, they’re exposing themselves to these different things. Part of it is earnings momentum, part of it is industries that got lucky recently, and part of it is, a time-varying, business cycle-dependent bet on the market.

Hmm. Say, that’s driving a lot of momentum, does that get captured by any other known factors like profitability or anything like that?

So, very low frequency things that we think about like value and profitability are not ever going to do a good job explaining the performance of the anomalies that take place at higher frequencies, not high frequency [inaudible 00:37:06], but in the academic literature turnover, we balance the value and profitability factors once a year, whereas momentum or earnings momentum are things that get rebalanced monthly and no, they’re not really going to help explain those things.

I would say though, that the academic literature has way overstated the performance of momentum and earnings momentum portfolios. Because for a practical, applied, academic field, we spend a shocking little amount of time paying attention to real world issues like transaction costs and implementation. So in academia, when we construct price momentum or earnings momentum portfolios, we just try and get the biggest spread we can and we don’t care about turnover at all, and we completely ignore implementation issues. So, the strategies that academics look at have 700% annual turnover. If you construct them a little more efficient in terms of transaction costs, they still in the back test do have positive performance, but it’s half as large maybe as what the academics typically claim. I really do think we’d be better off as a academic profession and we’d have a lot more impact on the real world if we took implementation issues more seriously. I just think it would help us be more relevant.

Interesting. You mentioned some downsides of investing in momentum like concentration and maybe some industry bets. Should investors be using it at all in portfolios? If it’s not in a momentum fund because of the downsides that you mentioned, what are some ways that it can be used?

The big thing about momentum was partly that it was negatively correlated with value. It wasn’t a huge negative correlation, but it made it look even more impressive with value. That’s because the winners were stocks that had gone up a lot. They’d gotten more expensive. They tended to look a little more like growth stocks. The momentum strategies were overweighting in growth and so consequently, negatively correlated with value.

That incentive for trading momentum, I think is much weaker when you’re also including profitability, because profitability has a bigger negative correlation. When you’re trading profitability and value together or jointly, it’s just much less related to the momentum. Momentum is still a diversifying strategy, but it’s not a hedge anymore. The extremely high turnover makes it much more difficult to deal with. It’s just a much more active strategy. It’s also makes me a little nervous because we don’t understand it at all so that also just makes it less reliable for me.

I think that there are ways that people try and trade it at lower transaction costs. One of the ways people try and trade things at lower transaction costs, which I think is completely sensible to do is, if you’re running other strategies, you have some natural turnover in your other strategies and some of that, you can adjust that natural turnover you would have done in the way to help you capture a little bit of say momentum. It gets you very small exposures to momentum, but I guess because it’s such a high turnover strategy with such high costs, it also has this extreme negative skew where occasionally it blows up.

So spring of ’09, the standard momentum factor lost 50% of its value in three months. That’s not a unique episode within the momentum data. I’ve thought about momentum a lot from an academic point of view. I’m not as big a fan of momentum personally, though I understand that many people love it, but I think that you do want to… if you’re going to try and trade it… Again, implementation is always really important. The first order thing here for momentum, as well as trying to understand what you’re trading is, to do it at the lowest cost possible and again, implementation issues.

So, how many stocks might fall in that momentum bucket? Like the there capacity constraints compared to say the value factor?

Yeah. So in the academic implementation for the factor models, it’s about a 30% of the market is what would be called a winner. In terms of the capacity constraints, that’s something that people have written about. I’ve done a lot of transaction costs work including on momentum, but not thinking about capacity constraints, just thinking about for a small investor on the margin, how much less profitable it is that even if you’re only going to trade a dollar? There is literature thinking about capacity, how much of it can you trade, and they came up with wildly different estimates. So there are estimates of almost nothing before momentum profits go away to people saying that you can trade hundreds of billions of dollars before it falls to an unattractive sharp ratio.

Again, this is not my research, so I can’t speak authoritatively on it, but I do think this literature has a problem, which is, the way people estimate capacity is by looking at the price impact of trading. In the academic estimates of the pricing cap [inaudible 00:42:09], the trading actually industry as well, what you do is, you look at how big is the price impact on days with different order flow? So you’re looking at days when there are big imbalances versus days when there are small imbalances of trade and how much more do prices move when there are big imbalances and when there are smaller imbalances. They basically just trace out that curve and use that as an estimate of if you double your trade, well, that makes a bigger imbalance and you would look to see how much that impact prices.

The problem with me for that, and the reason I haven’t done any work in this space is, that to me, this is the same exercise as going out and watching people drink Coke in public and noticing that people drink a lot more soda on hot days and concluding that if you can get people to drink more and soft drinks, it’ll make the weather warmer. So the problem is, when people go and they look at the order imbalance and how big the price impact to trading is, they’re ignoring the fact that the choice to trade is an indogenous decision and if you need to work a big trade, you’re going to choose a day when the price impacts are small and the market’s particularly liquid to trade a lot.

So, if you just take the lessons and say, “Okay, well, I see that on this day when there was double the amount of trading, there was only double the impact on prices,” that doesn’t mean that if you now double the trading, you’re only going to double the price impacts, because maybe the trading was twice as big because of the… So there’s this endogeneity issue that makes it an incredibly difficult exercise and the bias is all in the way of underestimating how big. And if you exogenously decide you’re going to trade a lot, the estimates you get from the way people do the estimates probably underestimates how big your pricing back to trading is. So, because of those issues, I tend to think about the high capacity estimates as being overstated, but I don’t really have a strong opinion on what the true capacity of momentum is.

That was fascinating, really, really interesting to think about. You alluded briefly when we were talking about profitability to low volatility, which is another anomaly that profitability, I think helped to explain. Can you talk a little bit about the low volatility anomaly?

Sure. And while I’m at it, I’ve written a lot about defensive in general so I’ll talk about low vol. I guess also bring in low beta here a little bit, because they’re sort of related though distinct. So the low volatility anomaly, and I love that you call it the low volatility anomaly because I don’t think about it that way, but everyone talks about it that way and it is what we call it. I call it that to you, but what you see in the data, what you really see is that only after 1968, you see that the most volatile stocks in the market have significantly underperformed the rest of the market. So we call it low volatility, but really it’s a high volatility anomaly. It’s about the fact that the highest volatility corner of the market has in the second half of our good data sample, underperformed, and it’s underperformed by a lot.

So, I guess the question I always have, because I try and understand what’s causing things is, what tilts do they take? To get to that question, you have to ask what sort of stocks have the highest volatility? So if you’re going to try and pick stocks with the highest volatility without looking at volatility directly, what sort of stocks would you choose? Well, I think the one that’s kind of most obvious to most people is that they’re probably small. So that’s the really high volatility stocks tend to be tiny. When we’re talking about the high volatility stocks, even if we’re talking about the 20% of the stocks with the highest volatility, it’s just a sliver of the market in terms of market cap, a percent or maybe slightly 1.2% of the market.

It turns out that statistically, there’s an even stronger predictor than market cap, which is operating profitability. Stocks with higher operating profitability, perhaps not surprisingly, tend to have lower volatilities. It also turns out that if you pick an average growth stock in an average… just a random growth stock and random value stock, they have basically the same volatility on average, but on average, the growth stock, the more expensive stock is more profitable. We know that high profitability is associated with lower volatility so the fact that the average gross stock has the same volatility as the average value stock, despite being more profitable, means that high relative price is also associated with higher volatility, at least after controlling for profitability.

So, what does that all say? Well, it says if you want to find the most volatile stock in the market, you should be looking at small unprofitable, expensive stocks. We know that that small unprofitable growth corner is the corner of the market that has just dramatically underperformed the market. So the high volatility stocks unquestionably have performed really poorly in the post ’68 sample, but I think they do so primarily because small unprofitable growth stocks, which to be high volatility have been by far the worst underperforming segment of the market. So, that’s kind of my take on most of what’s going on in the low volatility anomaly.

The other half of defensive equity, the other big piece of defensive equity is actually much older. The low volatility stuff sort of comes out of the mid 2000s, whereas low beta or beta arbitrage strategies were actually suggested 50 years ago by Fischer Black. So they come out of this again, this failure of the CAPM where the Fischer Black was in one of the two earliest groups of researchers to do good empirical test of the capital asset pricing model and that he saw that low beta stocks had lower returns in the market, but not as much lower as the CAPM said they should. It suggested buying low beta stocks and levering up on them a little bit if you wanted a beta of one.

Now, here again, the issue is the other tilts. I already talked about them a little in the success of the five factor model. Well before academics realized it, industry people were already noticing that when you kind of run a low beta strategy, you were taking a value tilt. It explained part of the returns, but the value tilt wasn’t big enough to explain all of the returns, but it turns out that that value tilt is really there because not because of the value tilt so much, but because there’s a very big investment tilt on the low beta strategies. The stocks that have low beta are firms that are not doing a lot of investment. They’re not increasing their asset basis, whereas the, I’m sorry, the high beta stocks tend to be ones that are doing a lot of acquisitions or a lot of internal investment. The low beta stocks also tend to be more profitable. So, the beta arbitrage strategies, the low beta strategies, can take big tilts, quite big tilts to profitability and especially investment that really explain their performance.

Now, you can make them look better on paper to where you can’t explain them with the standard models. There’s a very influential paper called Betting Against Beta that came out maybe eight years ago that shows really lights out performance of these anti beta strategies. But if you really dig down on what’s going on there, the strategy is totally driven by nano cap stocks. For every dollar invested in the strategy constructed the way it is in the paper, you ended up taking a dollar five and positions in stocks that make up the bottom 1% of the market is it quite a bit of turnover in the strategy the way it’s constructed and the lifetime performance is really driven by sort of exploiting the fact that there’s lots of stocks among the stocks that institutional investors can easily trade that looks anomalous.

Even value, if you trade value kind of constructed the way this betting against beta strategies is constructed, there’s no way that our standard model with the value factor can explain the performance because value looks just outrageously strong among the inner cap stocks. But I don’t take it very seriously because again, it’s not something that’s designed to be anything close to the way you might implement the strategy in the real world.

If low volatility or high volatility is subsumed by other known factors for someone who has, for whatever reason like in Canada, we have a smaller investment product landscape. If someone has trouble finding a factor product, but they want exposure to value and profitability, is investing in a low vol fund a way to get access to that?

Yeah. I mean, I guess I think about this sort of the same way I think about profitability and quality. I mean, I think there were quality strategies out there that took little tilts towards profitability. If you didn’t know about profitability, they would have been a good thing to have in your portfolio because they get you a little bit of profitability, which you want. But once you know about profitability trading, these, some of these other measures of quality are a pretty inefficient way to get those tilts.

I kind of view the low volatility or low beta in the same way in that again, I do think that what they’re doing is, getting you these other tilts. I mean, certainly again, if you didn’t know about these other factors, having a high volatility exclusion from your equity portfolio would have given you some benefits. I mean, it’s not outrageously big because again, the high volatility stocks tend to be so small, but it’s a relatively small exclusion in terms of market cap, but it would get you a significant information ratio by helping you avoid the small unprofitable growth stocks.

Again, at least using an exclusion, I think it’s an inefficient way to get a good exclusion. I’m not as sold on true low volatility products, because again, I think this is really concentrated in a narrow sliver of the market cap. When you start talking about a true low volatility product, I’m not sure what part of the market you’re talking about so I’m just not as clear on it. But yeah, avoiding a few percent of the market that was the most volatile is probably better than nothing, but yeah, it’s probably an inefficient way to get a small unprofitable growth exclusion.

So, I want to take this chance in having such a great expert, join us to ask a relatively layman-type question about factor models. There’s been a proliferation in factor models, certainly over the past 10 years. Why does it matter to have a factor model if you’re a regular investor and how does someone choose which one to follow?

I guess I’m not sure why it matters to the regular investor, except that we do use it for performance evaluation, to some extent. There’s a lot of richness and detail in how I could answer this so I’ll try for some of it. We have this issue in that we often try and evaluate factor models by seeing how they do on anomalies. The problem here that the definition of an anomaly is that it doesn’t work for one particular model, the Fama-French model. So you’ve stacked the deck against the Fama-French model whenever you’re doing these horse races, because we’re testing the Fama-French model against the other models, using the strategies that thousands of people have spent decades trying to find to challenge the Fama-French model.

We have some other criteria, more sophisticated criteria that are sort of taking off for trying to evaluate performance models relative to each other. I actually have been working in this space, but again, back to my issues of frustration with the profession, ignoring implementation issues, it turns out that a lot of the models that under these criteria do better than the Fama-French model in some sense, are models that have factors that are much more expensive to trade. We ignore those transaction costs when we’re trying to do this model evaluation and it makes a significant difference in the conclusions that you get from these tests.

For a typical investor, I’m not sure that it matters tons because for them, I guess the main thing is, sort of what sorts of big important signals are they using in these factor models? Those are basically different ways of measuring the same basic few things. They include relative price and size and profitability, and maybe an investment, maybe momentum, that’s sort of the big one you have to decide whether you want there or not.

I guess there’s another issue, which is, I think a lot of the time that we use the factor models, one of the big uses we have for them as academics is, to do performance attribution backward-looking. And there, what you want out of the model is not necessarily what helps you win the horse race the way that the models are evaluated against each other in practice.

In practice, sort of what wins the horse race when you’re evaluating these models is, that some combination of the factors that you come up with after the fact would have performed super well in the back tests. So it’s about what we would call ex-cost deficiency. So there are things like the size factor, the size factor has not had a huge contribution to a high sharp ratio ex-cost but the size factor is probably the second most important factor next to the market factor in actually evaluating how portfolios performed in the past. Lots of strategies have size tilts and including that size factor in your model helps you explain more of the realized performance. It can help explain why a strategy did particularly well at some point in the past. That’s important when using a factor model, but it’s not going to help you win this horse race that a lot of academics seem overly interested in.

The premise of the paper that sparked that question, as I understood it anyway was, basically that with these model evaluation metrics, the Fama-French model doesn’t look so great. And there’s a lot of other models that look better, but if you account for transaction costs, all of a sudden the Fama-French model looks as good or better than the other models.

Yup, yup, and part of the problem is, in general, it’s like new products. If you have a tiny marginal, incremental improvement on a product, we know the Fama-French model and we trusted him, I don’t even see the marginal improvements, but if there was one, it’s not good enough. You need a really big improvement to try and convince anyone that they really should be using something else and I don’t see it.

That paper is also one where I illustrate the problems with the methodology, with an example that’s ridiculous. I can come up with a one-factor model that’s complete nonsense that’s very expensive to trade. And if you ignore the transaction costs under these metrics people are using to compare factor models, it looks far better than any model anyone’s using. No one in 100 years would consider it as a better factor model. So I just think to some extent that the paper is a critique about the methodology that people are using as much as trying to do the evaluation, because the models are all very similar. They’re based on the same idea, it’s they’re tweaks on the same basic factors.

Oh, it was a great paper because we occasionally hear that all you guys are using these products that follow the Fama-French five-factor model. But what about this other model? Why aren’t you looking at that? So your paper was, I guess reassuring.

Yeah, I guess one of the reasons we use it is, because we’re all comfortable with it and we can talk about results in a way that are meaningful to each other. That has great benefits.

Comfortable with it, which is great, but seeing your paper showing analytically why it’s actually as good as any other model, I thought that was, like I said, reassuring.

We talked earlier about the importance of diversification in capturing premiums. We’re in Canada. We use factor-tilted products of the Canadian equity market, which is, I mean, compared to the U.S. market it’s tiny. Do you think it’s still reasonable to pursue factor premiums in a smaller market like Canada, where we cannot be as diversified by the number of names that you can in the U.S.?

If you’re explaining the factors purely within the domestic market, it does make it a little harder. I mean, it makes it harder to get the same level of diversification. To some extent what it means is, to be sufficiently diversified, you can’t take these larger tilts. It’s remarkably consistent the performance of these tilts across different equity markets and even across different asset classes. So I do think you can get a lot of reassurance about the existence of the facts looking elsewhere. Again, as long as you’re not trying to take big tilts, I think it’s completely reasonable to pursue them in Canada.

That’s good. Good to hear for us. Your papers are, in a lot of cases, what you might colloquially referred to as a mic drop, where you’re, like you said, you’re kind of going after the methodologies that are being used by other academics. I’m just curious as a point of interest because we’re not academics, we’re not in that world. What’s the dynamic like when you drop a paper, that’s saying basically all these other people were wrong?

It really is not a very good way to make or keep friends or even to publish papers. It’s hard to publish papers that have negative results. I’ve read the some of these with some junior coauthors and I guess that they worry a lot more than I do at this point in my career. At least on the academic side, I don’t want to lose friends, but at least on the academic side, I’m pretty confident about it.

At the same time, I kind of view it as not my problem. I think as a profession, we have an obligation to present robust results. If you don’t, if you present results that are fragile and due to kind of your misleading use of methodology, it’s your fault if you get called on it. So, I don’t really carry too much weight from doing it. I think it’s the author themselves’ fault if they’ve written a paper that can be called out like that.

I kind of understand why people do it at the same time. I think there’s always this tension between playing a short game and a long game. So certainly in academia, it’s easier to get a paper published and to get a lot of early sites, if you make a big outrageous claim and so it helps you get published. The problem is that if the paper’s important and people pay attention to it, they will try and replicate it and you might end up paying the price later. So I personally have always been more comfortable playing the long game. So I try and really, if anything, be conservative on how I present the data and the kind of claims I make and the methodologies I use and then I’m confident that people will be able to replicate the results and see that they’re just as strong as I claim.

I think this actually, this long game versus short game thing is something that you and your listenership see in a very different context, which is in industry, people are doing exactly the same thing. You have huge incentives to overclaim what you can do to win mandates, but the bigger your claims, the harder it is to deliver in the longterm. I guess I would say that sort of my equivalent is, I want to deliver in the longterm. That’s what I’m trying to do.

Do you have friends in academia that you’ve torn the paper apart of, and then had a friendly conversation about it afterwards?

Torn the paper apart? Is that what I do? I mean, I point out. Yes, I have very good friends who I’ve written papers about their papers pointing out issues with their papers. Again, I’m not trying to tear anything apart. I’m trying to shine light on what’s actually going on.

Okay. Maybe those were the ones that you tore apart.

Or maybe the ones who stayed my friends were my real friends to begin with.

So our listeners know that we use Dimensional Fund Advisors products to build portfolios and have for a long time. We have many of our clients that are regular listeners. Now you have consulted for Dimensional Fund Advisors, I believe for seven or eight years. Can you talk about why you decided to consult for them and what makes that work with them attractive to you?

I guess it’s for the same reasons I’ve just been talking about. I mean, I think Dimensional is also playing the long game and are fairly conservative in what they claim and have done a pretty good job delivering it.

I guess at some level, it really comes down to skepticism. I mean, I guess you could characterize the mic drop papers I’ve written and my whole approach to research is partly about trying to be skeptical and being robust in the work I do. I think Dimensional’s been a very comfortable place for me because of that.

Part of it was also just interest in happenstance. The way the relation actually came about was, I was in Austin, Texas to give a paper on public pensions. I’ve done a lot of non-finance work on what I would call public finance around public pension systems in the U.S. and I was there so I invited David out to dinner.

Unbeknownst to me, they were getting ready to implement profitability in their first traded product within the next few months. So this was something that… I was on the faculty at the University of Chicago with Gene at the time, so he had seen early versions of my profitability paper. Profitability had been something they had pursued earlier because Gene and Ken had some work on it, some of the work that said, “Oh, we think it should be there, but we don’t see it in earnings.”

So, when Gene saw me presenting the early versions of my paper, it was something that they went back to looking at at Dimensional and were getting ready to launch. So it made perfect sense for me on that issue. I mean, I really was just taking David out to dinner, but it became a job offer because it worked perfectly in the timing for my interests and theirs.

I just have to say, it’s been a wonderful relation. I talk to people there all the time and I love the people I work with there. The research is great. Love the PMs. It’s just been, it’s been a really good experience for me.

That’s very cool. That’s a cool story to hear. You spent some years as a professional triathlete before going into academia. You’ve mentioned earlier, when you were talking about your profitability paper, what you won on, what the wins were with the paper. What insights have you carried over from professional sports into your research and investing?

There’s certainly character transference. I mean, there certainly has been reward for me in both pursuits for hard work and persistence. I mean, I can’t understate how much persistent matters. I will grind against the problem sometimes because I have a sense that I can get to the answer. Even before I know how to get there, I will sometimes feel like I can and it often is just hard work.

But at the same time, I have to say, I think I would’ve gotten a lot farther faster in academia if I hadn’t had the detour into triathlon. Going back to graduate school, my first year in graduate school was the hardest thing I ever did. Just going back to doing measure theoretic based probability, not having thought about math in seven years was brutal, but I’m not unhappy I did it. Life isn’t a straight line race and I think both things have been good for me.

It’s a perfect segue to our last question, Robert. How do you success in your life?

I guess I try and appreciate what I have and trust it and take care of my family and the people I love, and really important for me is staying open to new possibilities and opportunities, trying to be a lifelong learner.

About The Author
Cameron Passmore
Cameron Passmore

Cameron Passmore has been a leading advocate for evidence-based, systemic investing for over 20 years in the Ottawa area. Today, Cameron and his team serve a broad range of affluent clients across Canada.

Benjamin Felix
Benjamin Felix

Benjamin is co-host of the Rational Reminder Podcast and the host of a popular YouTube series.

Meet With Us

Meet With Us Flyout Form
What are you looking for?