Professor Robert Novy-Marx Discusses Quality & Value Research

Professor Robert Novy-Marx has become well respected for his research on value and quality investing. To gain deeper insight into Novy-Marx's work he spoke to value investor Tim Melvin.

Tim: With us today is Professor Robert Novy-Marx of the University of Rochester, New York. I want to thank you for spending some time with us today. And I want to talk about two papers that you’ve written in the last few years that I think are some of the most important academic papers on investing and finance in the last decade. Your first one was “The other Side of Value”, and that came out in 2012, Correct?

I believe it was actually published in 2013.

Tim: In the paper you looked at stocks from 1973 to 2010 and found that quality stocks, defined as profitability measured by gross profits to assets, had pretty much the same power at predicting returns as book to market. As a big value investing proponent I found it remarkable, but how did you arrive at the definition of gross profits to assets as a measure of quality.

It really came about because I had a theoretical model of the value premium. It’s a little technical but it made some predictions about operating costs of firms and how that would relate to their expected returns. I knew that from some previous accounting literature that revenues to assets had some predictive power as well. I was really trying to disentangle how those two things were related and if they were related, and that’s really what got me to look at the actual measures; the difference between the two things.

Tim: I believe I saw in an earlier interview you did that you found quality stocks were those with assets less than three times total gross profits. Is that about right?

How exposed you are is going to depend on where you put that. So, if you make that threshold more severe, you’ll get fewer stocks out of the less risk averse portfolio, but you’ll get a greater exposure to this profitability factor.

There’s no magic number there, it’s really a preference of the individual investor as to how diversified they want to be; how far they want to tilt from the market.

Tim: Did you find that the lower that ratio went, the better the returns were?

Yes, but the more extreme the ratio the more exposed it will be to the profitability factor.

Tim: In your study you made a statement that I always found fascinating. Could you explain your thinking on this to our readers. You said that, “The further down the income statement, the more polluted the profitability measures become”.

For a lot of people who look at the accounting numbers a lot, I don’t think they’re that surprised by that. There’s a lot of discretion in how the accounting is used and you know there’s much less at the top of the income statement.

Tim: When you were doing the study, was there any difference between the performance of larger cap stocks and smaller cap stocks when you used the gross profitability measure?

Gross profitability is a good measure for both large and small cap stocks. It’s a little different in how you want to use it in the two places. In the small caps it really has a lot of power on its own predicting returns. In the large caps it has little less on its own, but it works particularly well with traditional price based value measures. So, if you’re going to be using it in conjunction with the price based value measure, there’s not a whole lot of difference. It works well across the size spectrum.

Tim: You also tested gross profitability with a lot of other demonstrated anomalies such as low price to earnings and metrics like that. How did that turn out?

I think it’s pretty clear that there’s a lot of so-called anomalies out there. Things that we’ve had a hard time understanding in the context of our traditional French-Fama three factor model. Taking positions on profitability and accounting for the fact that they are taking in positions in profitability really helps explain the performance of the bunch of earnings related anomalies. This is a very active area of research, not just for me, but there’s a couple of sets of research including firms who invest themselves, who are now using non-profitability in conjunction with other things.

Tim: One important finding in the study that has really struck me is that quality on its own and value on its own are not correlated at all.

No they’re quite negatively correlated, which is really what makes profitability attractive to traditional price based value investors. I like in it in some extent for a value investor that profitability’s insurance that you get paid to have. Because the strategy has good returns on its own; and it actually reduces the risk of your value exposure.

Tim: Just out of curiosity; have you ever looked at taking 50 percent of the portfolio and just doing the high quality stocks and 50 percent doing just the traditional book to market value stocks? This would be long only I think.

If you pick the half of the market that has a higher profitability, and the half in the market that has the higher book to market ratios or earnings to price ratios, you’re going to get similar information ratios on your tilts.

You’re still exposing yourself to those factors, but you will have very diversified exposures. Because they’re so diversified, you’re not going to be getting big tilts. It will still be an attractive tilt, but it will just be a very small tilt away from the market portfolio at that point.

Tim: Your second paper which you updated in 2014 was called “Quality Investing”.
Yes.

Tim: It is the single most important paper that I’ve read in the last 15 years, I’m going to say.

Thank you.

Tim: And you’ve combined several different definitions of quality here. You looked at Ben Graham’s definition, Joel Greenblatt, Jeremy Grantham, Joseph Piotroski and a few others that I don’t have written down here for some reason.

I look at Richard Sloan’s earnings quality measure and also at some defensive measures.

Tim: You combine these with stock trading at low price to book values that also adopt the particular high quality measures. Your findings were remarkable. Can you tell us a little bit about what you discovered as you tested these against each other?

I see that there are some commonalities to the strategies. So, the things that people sort of broadly market as quality, they do all seem to be getting at some of the same stuff. Perhaps it’s not surprising to you that since I write the paper, but I do basically find that if you were to use a single quality measure, that profitability seems to be clearly that the strongest one. It’s particularly so for long-only investors and a lot of the quality measures that people use really don’t have high returns themselves but that they do have high what we’d call ‘risk adjusted returns’, because they had very low or negative exposures to the market and to small caps and to value factors. You know they’re going to again diversify the risk for value investors.

Tim: So using the quality measure, you’re kind of doing what Piotroski and even Edward Altman have tried to do over the years and that’s get rid of the ones that aren’t going to work.

In particular, you’re really trying to get rid of the non-value stocks. These measures all work better when you use them in conjunction with value measures. It is not a term that I’m big fans of I guess, but I think that there are people who would say that really what you’re trying to do is avoid the value trap: use these measures to avoid the stocks that are cheap with really good reasons and try and find those stocks that are cheap because people are just counting their expected future cash flows at a really high rate.

Tim: What kind of out-performance did you find with the various strategies?

It depends on how extreme you make your strategy.

How good the risk reward trade-off on these tilts is often better than the risk reward trade-off on the market itself. So, if that’s the case, I guess the person trying to find the ultimate portfolio…a long-only investor should really be trying to tilt fairly far from the market and in the directions of these profitability and value factors because the risk reward trade-off there is quite high.

Tim: I’ve done two back tests using the limited data sets that I’ve had. They were August 2012 through August 2013, and January of 2014 through 2015. I used below 90% of book and total assets that were less than three times gross profitability.

Yes.

Tim: So much that I went back through and checked the data by hand. So it was it was that good.

Yes, it’s a very short example.

Tim: It is.

It’s a very short example but yes, gross profitability has performed remarkably well in the last couple years.

Tim: Did you find any difference between financials and non-financials? It would seem given the tremendous amount of assets that financial companies have to use, this almost isn’t going to work.

The problem there is not that the measure itself does not work in financials. It actually works reasonably well among financial firms, quite well. The problem is it makes it very difficult to compare financials and non-financials.

Because, what we mean by assets, the denominator of the ratio is just a very different thing for financials than it is for a firm that’s using its assets as operating assets. So, there are things you can do. In the second paper that we’re talking about, the quality paper; I actually do include financials. I just use the measures separately for financials and non-financials when I rank firms. The other thing you can do is go to a slightly nearer measure of profitability. One that takes out interest expenses; cash flow that goes to the bond holders and uses the book to equity rather than book to assets for the denominator. So , that would be an equity levered measure of profitability.

Tim: One thing that really surprised me is that the maximum drawdowns in both large and small stocks using the strategy were a fraction of the index drawdowns.

It’s not really just that the profitability strategies have such low drawdowns. This is really back to this fact that profitability and value are so highly negatively correlated. And so I think the drawdowns that you are talking about are when your trading strategies that combined the profitability measure with the value measure.

The reductions in the drawdowns there and especially among the large cap stocks is really in some ways the most remarkable things about the strategies.

Tim: I think you were talking about 18 percent maximum drawdown in large caps stocks, over a very long period of time.

It was less than 19 percent over the sample I think that started in ’63 from the large cap strategies that combined value and profitability. The traditional value portfolios had drawdowns. The biggest drawdown was well over 40 percent and the market itself had an absolute drawdown of more than 50 percent three different times over that period. So, a maximum drawdown of 19 percent.
It was the most surprising thing for me in the whole research.

Tim: Given taht the periods; you’re talking about include the crash of the 70’s, 87, there are some bad market periods in there where many stocks were down much, much more than that. The other thing I noticed is that you had much fewer periods of out-performance than a traditional value take.

Yes. For the traditional value strategies, it depends on if we’re talking large caps or small caps. Traditional value is very successful in small caps. Among the small caps, it beats the market about almost two out of every three years. For large cap value you’re only up eleven out of twenty years value relative to the market. But, when you combine profitability in the large cap you’re up seven out of ten years.

Tim: In September you published a paper on defensive stocks. I’m only part way through that because it takes me a lot of crunching the numbers. What you find in that paper has been widely discussed as of late.

I was interested because there are claims of other people that the defensive strategies, particularly the ones that are low market data strategies are really a way to get value in disguise. I see that in the data as well, at least if you look at the last 50 years of data. If you go back more than 50 years the low data strategies actually look like growth strategies, but they also didn’t have any out-performance at that point.

I was a little more interested in the low volatility strategies which have been harder for people to understand and have seemed a little more anomalous to the academics. But, I guess I had a hunch that there was a profitability issue there as well. I was to ask you, “What are the two single things about a stock that you think will predict volatility?” I think everyone would say, “Being small”. Small stocks are more volatile. The one that I don’t know if many people would say right off the bat, but I think it makes sense looking back is actually profitability. I mention profitability there a little differently than I do in my other papers just because it correlates with volatility more strongly. It appears that the stocks that are less profitable, I guess not surprisingly if you’re running at razor thin margins, and there’s a little bad news, then it’s going to push you into the red. Whereas if you’re running at giant margins, a little bad news means you’re running at still pretty big margins and your stock isn’t quite as sensitive to that. The stocks that are running at much lower levels of profitability tend to be both smaller and much less profitable. And so, the defensive strategies that trade on volatility end up being strategies that buy large cap profitable stock that also tilt a little to value and they really avoid the stocks that are small and unprofitable growth stocks. And so I guess I would say that avoiding small unprofitable growth stocks is generally a really good idea. They’ve been the worst performing segment of the market for as far back as you want to go.

Tim: Unfortunately, that’s where the general public likes to speculate.

I would say that defensive stuff; the low volatility strategies are avoiding the really high volatility stocks. Relative to not doing it is a good thing, because it does help you to avoid that segment of the market which are just terrible performers. I would also suggest though that it’s probably better to just avoid those stocks directly and it’s certainly going to be more transactionally efficient. Because those portfolios turn over a little more so I think if you want to avoid the small unprofitable growth stocks I would say to do it directly.

Tim: Now just going back to the quality investing paper for a second, because something very important just occurred to me. What would your criteria be for eliminating a stock for the portfolio? That is, when do you sell?

The portfolios there are just constructed to try and take a segment of the market, generally 30% of the market that looks the best in terms of a very simple combination of valuations and profitability, or valuations and quality more in general. So in that paper I rebalance it annually and I sell when they fall out of the top 30%. You can do things to even reduce turnover a little more. If you were setting these up in practice I would have a buy/hold spread on them, where you might buy if they get into the top 25% and you don’t sell until they get out of the top 35%. It just avoids you buying and selling stocks that are essentially equivalent because they’re right on the opposite sides on the razor thin boundaries. So it would help you avoid a little bit of transaction cost for no real gain. But to be honest again these portfolios that we’re talking about here, turn over once every four years, so transaction costs are really already quite small and it’s not going to make nearly the kinds of improvements you’d see from my buy/hold spread in a strategy that turned over more frequently.

Tim: You’ve been quite prolific with papers over the last few years. You’ve done, in addition to these three, some work on public pensions and real estate investing that I found interesting. What are we going to see next from you? Anything fascinating you’re working on?

I have two things that I’m actively working on and I’m quite excited about. One of them is something on momentum and I’m not going to say too much about it because it will be available in a couple of weeks. The other one is a little farther form being done, but it’s really me trying to investigate and quantify the effects of these strategies that use multiple signals. But I mentioned this earlier that it’s hard for me to really evaluate what really good back-tested performance means when we’re back-testing something based on a bunch of things that worked pretty well in the past. So this is a little more, I guess, academic and a kind of metric. But I think it’s going to have really important implications for how investors evaluate claims on strategies that people say, “Well you have to select a stock based on these five characteristics”. The basic conclusion of that paper is that, “Yes, if there’s a few different things you believe in, you should probably trade them all together”. But when you want to evaluate each of the signals, you have to evaluate each signal individually. It’s cheating to put five things together and say, “Well these five things work well together”. You can’t expect it to work nearly as well going forward.