During his recent interview with Tobias, Partha Mohanram, a Professor of Accounting and John H. Watson Chair in Value Investing at The University of Toronto discussed Picking Winners From Losers Using The G-Score. Here’s an excerpt from the interview:
Tobias: Well, your name was brought to my attention by the Practical Quant, Jack Forehand, he’s a partner at Validea, in a podcast that we did. And he pointed out that the best-performed strategy, I think last year or the year before was your G-score. This is because they track a variety of strategies. And I thought it was absolutely fascinating. And I think I said it was funny. I didn’t mean to offend you, but I thought it was funny in the sense that it’s explicitly looking in the most expensive stocks.
Partha: Exactly. So, if I can just sort of give you a little bit of history on that. By the way. I think I have met this gentleman or maybe one of his partners in like the mid-2000s because if I’m not mistaken, they are based in Connecticut somewhere north of New York City.
Tobias: That’s right.
Partha: And they came down to Columbia to talk with me. The thing about Validea is they don’t really have formal relationships with these professors who publish papers. They just take it and interpret it as they want to. So, for example, my G-score paper is actually a longshot idea, but they are just focusing on the long side, and the idea has done phenomenally. To be fair, the reason why the ideas done really well is not just because it’s a great idea. It is, but it’s also because I think, in general, value has not done well while growth has done well in the last decade. That’s kind helped, so that rising tide has lifted this boat as well.
So, the basic idea of that paper is, everybody knows Piotroski, at least not just academics. And so I didn’t start off as a valuation guy. So, I got my PhD from Harvard. My thesis was the area of disclosure, I got a more of how do firms communicate and how do they improve their information environment, that was my area of research. But I ended up teaching FSA, financial statement analysis, and ratio analysis. And also I got interested in valuation from a very practical perspective, and I came across Piotroski’s paper. And I just love that paper. By the way, I know Joe really well. So I’m not just saying this, but I did because I know him. But I like that paper because it was a very simple practical idea of applying stuff that people do.
And he basically said, “Let’s test fundamental analysis. But let’s test it in a setting where we know it should work.” These are value stocks. Nobody looks at them. So, it’s quite likely that there’s information in the public domain, financial statements that people just haven’t bothered to look at and therefore, it’s not been impounded into prices. If you just look at the F-score, it’s basically DuPont Analysis. It’s like, are you profitable? Is your profitability growing? Is your asset turnover improving? Is your profit margin improving? And then something’s pertinent to value stock. Like, is your liquidity getting better? Is your solvency getting better? Are you not doing stuff like issuing equity, which is a sign of weakness and so those are all the signals. And then, he shows that the stuff really works. His is like a simple test of fundamental analysis in a setting where you think it ought to work.
So, when I saw his paper, I had this thought experiment. I said, “This is awesome. But what if we think about the opposite quadrant? What if we look at a setting where we think fundamental analysis should not work?” If it works in that setting, it basically shows that there is value in fundamental analysis. Showing it works in a setting where it ought to work is great, but that’s like setting a low bar. So, I was trying to set a really high bar and see if it can work for that.
So, the other thing I also noticed was that the F-score doesn’t work that well in growth stocks. So, I said, “Let’s try to see if I can tailor fundamental analysis for the purposes of growth stocks.” Obviously, this notion that these firms are ignored and nobody’s looking at them can no longer be true because these are firms that are in the public domain and people are looking at them. They have a fair amount of analysts following, following in the business press, institutional investors, and so on.
Tobias: And high prices.
Partha: And high prices too. Probably, those things are related. But just because everybody is looking, doesn’t mean everybody is looking the right way. So, maybe it’s a case of fools rush in, everybody has fallen for the hype. So, can we still apply the basics of fundamental analysis to separate the solid growth firms to the hype firms? If you think about how Piotroski or how most people use to sought firms into value and growth, the most common ratio is the market to book or book to market ratio. Let’s use the word book to market.
Piotroski uses high book to market firms, so firms which have low market values related to the book values and calls them value. So, I look at low book to market firms. So I said, if you’re in the low book to market group, let’s try to see the guys who are in the low book to market group because there are some reasons why that book value is low versus the ones who are there because their market value is high, i.e., overvalued. So the ones who deserve to be there for accounting reasons, like accounting depresses book values in certain cases when you have lots of R&D, and you have lots of advertising and all. These are things which create assets, but these assets you’re forced to expense on, therefore, these assets don’t show up in your balance sheets and on your book value on the liability side.
Many of my signals were tailored for some of these accounting sort of thing. And the second signals I introduced, which are unique to my signal, was this notion of naive extrapolation. When firms do well, people assume that it’s going to stay on forever. So if you have two firms, both of which have a strong current performance, but one of them has steady performance in the past, the other one’s performance has been variable, the odds are the firm which has variable performance just got lucky and had a strong realization just here and you’re going to see some reversals in the future. So, I also built some signals based on how stable your profitability and how stable your growth has been because these firms– if you look at the ratio like a PEG ratio, talk about earnings, and you talk about growth, you want both the earnings and the growth to have quality. So, I was trying to build signals on that.
So, I came up with this index called G-score. And I basically showed that the G-score strategy is just like Piotroski’s F-score strategy works pretty well if you backtest it. The paper was written in 2005. I think the analysis goes with data up to 2002 or something. But the one difference between Piotroski and my paper is obviously Piotroski is looking at value stocks, which on average outperform the market. If you break his– Let’s say the average value stock beats the market by 5%, he breaks up that 5% into a 15% or 20% and a negative 5%. And he gets a long shot on that. But the short is not that crucial. You’re getting a lot of action from the longs.
On my side, we know that at least at that point of time, the average growth stock underperforms the market by 5%. I’m breaking up that minus 5% into a plus 5% and a minus 15%. So most of the action’s going to come from the short side. To get the maximum bang for buck, at least, that was the idea then, you need to short. Now, as things have gone on, we know that this decade– of the first one and a half decades of the new millennium has been very different, growth stocks have actually done very well. And that’s probably helped the performance of something like G-score, as the folks in Validea have shown. It’s done really well, on a long only side approach. But even there, if you had gone long on just growth stocks, like the book to market ratio, you wouldn’t have done as well as if you had gone long on the book to market ratio conditioned by G-score, which says that, “You know what? let’s focus on growth stocks, which deserve their valuations.” If you will just indulge me, I love using these corny analogies.
What Piotroski does, is Piotroski finds diamonds in the rough. These firms have rough valuations, he finds the diamonds among them. What I do is, I separate out the real diamonds, so these firms all have diamond valuations. I tell you these are the real diamonds and the rest of these are cubic zirconia. That’s my strategy. The real diamonds are the ones who deserve diamond valuations. The rest of the firms have diamond valuations, but these are cubic zirconias. That’s the analogy I use to talk about the differences between the F-score and the G-score approach.
Tobias: It’s a fascinating line of inquiry because it’s reasonably well known that the reason that people invest in the glamour end of the market is because they tend to have these lottery ticket properties where all of the very best companies over time never really get cheap enough to fall into the value bucket. They tend to stay in that glamour and you can think of examples like Walmart, Microsoft, many of those sort of companies, never get cheap on fundamentals. And so that’s why folks behaviorally tend to traffic in the glamor stocks, even though they know as a cohort, they underperform. If you can separate out the diamonds from the cubic zirconia, as you call it, from that group that really is– that’s the Holy Grail of that end of the markets. What does your G-score– What does it do? How does it differ from the F-score?
Partha: There are two fundamental differences. F-score does a deep dive into ratio analysis, because many firms in the values, they’ve been around for a long time, they’re more likely to be in sectors like manufacturing and all those kinds of things. So, your conventional DuPont base ratio analysis actually works very well. And also, he uses a time series approach. He’s comparing the firm to itself because he’s trying to look for signs of recovery. This is a firm which has a bad valuation, but maybe this is different from the rest of the firms because it’s actually showing positive momentum. It’s time series approach and a full DuPont.
My approach, I don’t use a full DuPont Analysis because, at least when I was looking at the paper like 15 or 16 years ago when I was working on it, many of these firms, and I haven’t gone deep dive into it, so I don’t know if it’s true right now. Many of these firms are not necessary– They’re very often in the early stages of their life. Not everything’s a Walmart. If you look at many of them, they’re also like firms which have gone IPO in the last five years. And so you have much more young firms and so the operations by nature are extremely unstable. So, year by year comparisons are actually deceptive. You could have a firm whose losses could be worsening, but it’s actually doing well, because it’s like trying to build market share or something like that. Doing time series comparisons are a little fraught. So what I did was, I did within-industry cross-sectional comparisons. That is among the low book to market firms in this industry, and I think I used the SIC code or something, which firms are doing better on the signal and which firms they’re doing worse on that signal. That’s the first thing I did.
Second thing is because of the nascent nature of the operations, I didn’t do a deep dive into ratios. I just looked at, “Are you profitable? How are you doing on earnings? How are you doing on cash flow?” This is just very, very basic profitability. And then, I had these signals related to these naive extrapolation. How stable is your earnings growth? Or how stable is your earnings, your ROA or something. And use that. If you’re high on stability, you get a 1. I think I made it continuous, but I don’t really remember the exact– if you’re high or above– no, if you’re above the industry median, you get a 1. If you’re below the industry median, you get a 0 or something is how I did it. So actually, G-score is a little more computationally intensive. The F-score is really easy to do, but it’s not difficult, but it’s a little more computationally intensive.
And the last thing is, I introduce these accounting-based signals. Are you investing in R&D? Are you investing in advertising? Are you investing in CAPEX? For two reasons, number one, in the case of advertising and R&D, it’s clear there’s an accounting reason why it depresses your book to market ratio. So, you want the low B firms or the high-end firms. In addition for something like CAPEX, these are firms which are being valued for growth. You want firms which are investing in growth. So, things like R&D, advertising, and CAPEX, means you’re a firm, which is doing stuff to ensure that the future is going to be bigger and brighter and more profitable than the present. Even if your present is depressed, you know that this is a harbinger of good things to happen in the future. That’s the justification for these three kinds of signals.
I think Piotroski had nine signals. I believe the G-score has eight signals, but fundamentally the construction is very similar. Basically, what you want is your backtesting, you want it to look like the skyline of Manhattan. You want to see a bunch of upward bars with very few negative bars. So, if you look at 20 years, you want the strategy to rarely have massive negative returns because when you have massive negative returns, and you have massive positive returns, it’s very difficult to say that this is mispricing. It’s probably just risk. You take on more risk, you’re gonna get more return.
So we try to rule out the mispricing, a risk-based explanation, in addition to doing asset pricing tests, but just looking at the prevalence of losses in your strategy, and the losses hardly ever happen. That’s the first thing.
Second thing is, most of the returns are concentrated around future earnings announcements. If you look at the performance of the next year, almost 30% or 40% of the returns comes around the three-day trading windows of the next four quarters, which means that consistently these firms are surprising positively, at least your longs are surprising positively around future earnings announcements, while your shorts are surprising negatively around future earnings announcements, which tells you that it’s not risk. It’s something to do with fundamentals, which the market has not impounded, but your strategy in a sense has.
Tobias: How are you assessing the stability of the earnings?
Partha: I think if I’m not mistaken– so this is a bit of a challenge, because you need to have time series to do that. If memory serves me right, and I wrote the paper 16 years ago. And unlike other people, I have no coauthor to blame, it’s just me. So, I think I looked at the standard deviation of quarterly earnings for eight quarters. For example, I would look at something like earnings divided by assets or something, and look at how– I just calculate a simple standard deviation of that across eight quarters. And again, people might say, “How can you compare that so-so variable?” Remember, I’m doing the comparison across industry. You certainly can compare a ratio of that saying that among all firms in this particular SIC code, this firm had above-median variability, and therefore that’s a bad thing. This firm had below median variability, and therefore it’s a good thing. So that’s the way I code it.
Tobias: Are you applying the strategy in– is it being practically applied by anyone?
Partha: Well, I know for a fact– Okay, the first thing is, people ask me, do you apply it yourself? The answer is, I don’t. I tell you why I don’t, because I just don’t have the time to do this kind of stuff and frankly, I don’t have the money to do this kind of stuff. Professors are well paid, but not that well paid. But more importantly, if you invest a lot in individual stocks, like something like this, you need to be monitoring this thing on a pretty active basis, and I just don’t have the time and the bandwidth to that. So, I just buy index funds or whatever which line up with this, whenever I have to make decisions on my retirement accounts and so on and so forth.
I know for a fact that, in addition to– so somebody highlighted this Validea thing to me a few years ago, saying, “Hey, Partha, I found that you are listed as a guru among some really, really big names.” And I said, “Come on, you’re joking.” I found that really cool that these guys put me in a list along with the Buffetts and some really, really big names, and I was very kicked to see that. But the other thing is, when you look at many reports, and I come across some reports from some buy side investors, they do mention this strategy from time to time. And so, I do know that this stuff is being used, but the thing is it’s in the public domain. People can use it and I don’t need to– it has nothing to do with me.
The other question people ask me is, “If your strategy is so good, why aren’t you running a fund? Why are you working as a professor?” My answer is different people get motivated by different things. I really like my job. I like teaching, I like doing my research, and I like highlighting these things. And I’m not that motivated by the actual financial aspect of it. But I do believe that this stuff actually works because when we think about fundamental analysis, to me, it’s this very strange alchemy or amalgam of market efficiency and market inefficiency.
It relies on market inefficiency because you say that firms do get out of whack, they move out of position, but it also relies on market efficiency because you assume that they’re going to come back to the real value. So, it’s just belief in long-run market efficiency, but short-run market inefficiency. In some cases, it’s not gonna work. There are many people who say that Amazon is incredibly overvalued, or Uber or Tesla are incredibly overvalued. But some valuations are likely to be stuck in that thing for whatever reason. Again, I have this analogy here. You cannot take things too literally.
Suppose you’re a chemist and you’ve studied chemistry, and you have– I’m gonna go back to diamonds. You have 10 grams of diamonds in your right hand. And somebody says, “I have 10 grams of coal right here. It weighs the same. It’s the identical same chemical composition. It’s really inefficient that the market is valuing this 10 grams of diamonds at a million dollars, and this 10 grams of coal at 10 cents. So, I’m going to go long coal and I’m going to go short diamonds.” That’s not going to work because that inefficiency is baked in. And if you don’t agree with that, just try giving a significant other a coal ring instead of a diamond ring. It’s not gonna work.
Leaving out situations like that, fundamental analysis believes that there is something which causes stock price to deviate from value, but eventually they find that value. And if you can find that deviation systematically earlier and better than other people, you can make some money on it. And I believe in that. Certainly, I’m not one of those University of Chicago guys who say that– I’m a Harvard guy, by the way. You hear this joke about this Harvard MBA and the Chicago MBA who are walking on the street, and they found a $100 bill. And the Chicago guy says it’s not possible, because markets [unintelligible [00:20:08]. The Harvard guy says, “Okay, I’ll make the markets efficient,” and he’ll pick up the $100 bill. So, that’s my approach basically.
You can find out more about Tobias’ podcast here – The Acquirers Podcast. You can also listen to the podcast on your favorite podcast platforms here:
For more articles like this, check out our recent articles here.
Don’t forget to check out our FREE Large Cap 1000 – Stock Screener, here at The Acquirer’s Multiple: