by Wesley R. Gray and Tobias E. Carlisle, published 2012
The root of all investors’ problems
In 2005, renowned value investing guru Joel Greenblatt published a book that explained his Magic Formula stock investing program– rank the universe of stocks by price and quality, then buy a basket of companies that performed best according to the equally-weighted measures. The Magic Formula promised big profits with minimal effort and even less brain damage.
But few individual investors were able to replicate Greenblatt’s success when applying the formula themselves. Why?
By now it’s an old story to anyone in the value community, but the lesson learned is that the formula provided a ceiling to potential performance and attempts by individual investors to improve upon the model’s picks actually ended up detracting from that performance, not adding to it. There was nothing wrong with the model, but there was a lot wrong with the people using it because they were humans prone to behavioral errors caused by their individual psychological profiles.
Or so Greenblatt said.
Building from a strong foundation, but writing another chapter
On its face, “Quantitative Value” by Gray and Carlisle is simply building off the work of Greenblatt. But Greenblatt was building off of Buffett, and Buffett and Greenblatt were building off of Graham. Along with integral concepts like margin of safety, intrinsic value and the Mr. Market-metaphor, the reigning thesis of Graham’s classic handbook, The Intelligent Investor, was that at the end of the day, every investor is their own worst enemy and it is only by focusing on our habit to err on a psychological level that we have any hope of beating the market (and not losing our capital along the way), for the market is nothing more than the aggregate total of all psychological failings of the public.
It is in this sense that the authors describe their use of “quantitative” as,
the antidote to behavioral error
That is, rather than being a term that symbolizes mathematical discipline and technical rigor and computer circuits churning through financial probabilities,
It’s active value investing performed systematically.
The reason the authors are beholden to a quantitative, model-based approach is because they see it as a reliable way to overcome the foibles of individual psychology and fully capture the value premium available in the market. Success in value investing is process-driven, so the two necessary components of a successful investment program based on value investing principles are 1) choosing a sound process for identifying investment opportunities and 2) consistently investing in those opportunities when they present themselves. Investors cost themselves precious basis points every year when they systematically avoid profitable opportunities due to behavioral errors.
But the authors are being modest because that’s only 50% of the story. The other half of the story is their search for a rigorous, empirically back-tested improvement to the Greenblattian Magic Formula approach. The book shines in a lot of ways but this search for the Holy Grail of Value particularly stands out, not just because they seem to have found it, but because all of the things they (and the reader) learn along the way are so damn interesting.
A sampling of biases
Leaning heavily on the research of Kahneman and Tversky, Quantitative Value offers a smorgasbord of delectable cognitive biases to choose from:
- overconfidence, placing more trust in our judgment than is due given the facts
- self-attribution bias, tendency to credit success to skill, and failure to luck
- hindsight bias, belief in ability to predict an event that has already occurred (leads to assumption that if we accurately predicted the past, we can accurately predict the future)
- neglect of the base case and the representativeness heuristic, ignoring the dependent probability of an event by focusing on the extent to which one possible event represents another
- availability bias, heavier weighting on information that is easier to recall
- anchoring and adjustment biases, relying too heavily on one piece of information against all others; allowing the starting point to strongly influence a decision at the expense of information gained later on
The authors stress, with numerous examples, the idea that value investors suffer from these biases much like anyone else. Following a quantitative value model is akin to playing a game like poker systematically and probabilistically,
The power of quantitative investing is in its relentless exploitation of edges
Good poker players make their money by refusing to make expensive mistakes by playing pots where the odds are against them, and shoving their chips in gleefully when they have the best of it. QV offers the same opportunity to value investors, a way to resist the temptation to make costly mistakes and ensure your chips are in the pot when you have winning percentages on your side.
A model development
Gray and Carlisle declare that Greenblatt’s Magic Formula was a starting point for their journey to find the best quantitative value approach. However,
Even with a great deal of data torture, we have not been able to replicate Greenblatt’s extraordinary results
Given the thoroughness of their data collection and back-testing elaborated upon in future chapters, this finding is surprising and perhaps distressing for advocates of the MF approach. Nonetheless, the authors don’t let that frustrate them too much and push on ahead to find a superior alternative.
They begin their search with an “academic” approach to quantitative value, “Quality and Price”, defined as:
Quality, Gross Profitability to Total Assets = (Revenue – Cost of Goods Sold) / Total Assets
Price, Book Value-to-Market Capitalization = Book Value / Market Price
The reasons for choosing GPA as a quality measure are:
- gross profit measures economic profitability independently of direct management decisions
- gross profit is capital structure neutral
- total assets are capital structure neutral (consistent w/ the numerator)
- gross profit better predicts future stock returns and long-run growth in earnings and FCF
Book value-to-market is chosen because:
- it more closely resembles the MF convention of EBIT/TEV
- book value is more stable over time than earnings or cash flow
The results of the backtested horserace between the Magic Formula and the academic Quality and Price from 1964 to 2011 was that Quality and Price beat the Magic Formula with CAGR of 15.31% versus 12.79%, respectively.
But Quality and Price is crude. Could there be a better way, still?
Marginal improvements: avoiding permanent loss of capital
To construct a reliable quantitative model, one of the first steps is “cleaning” the data of the universe being examined by removing companies which pose a significant risk of permanent loss of capital because of signs of financial statement manipulation, fraud or a high probability of financial distress or bankruptcy.
The authors suggest that one tool for signaling earnings manipulation is scaled total accruals (STA):
STA = (Net Income – Cash Flow from Operations) / Total Assets
Another measure the authors recommend using is scaled net operating assets (SNOA):
SNOA = (Operating Assets – Operating Liabilities) / Total Assets
OA = total assets – cash and equivalents
OL = total assets – ST debt – LT debt – minority interest – preferred stock – book common equity
STA and SNOA are not measures of quality… [they] act as gatekeepers. They keep us from investing in stocks that appear to be high quality
They also delve into a number of other metrics for measuring or anticipating risk of financial distress or bankruptcy, including a metric called “PROBMs” and the Altman Z-Score, which the authors have modified to create an improved version of in their minds.
Quest for quality
With the risk of permanent loss of capital due to business failure or fraud out of the way, the next step in the Quantitative Value model is finding ways to measure business quality.
The authors spend a good amount of time exploring various measures of business quality, including Warren Buffett’s favorites, Greenblatt’s favorites and those used in the Magic Formula and a number of other alternatives including proprietary measurements such as the FS_SCORE. But I won’t bother going on about that because buried within this section is a caveat that foreshadows a startling conclusion to be reached later on in the book:
Any sample of high-return stocks will contain a few stocks with genuine franchises but consist mostly of stocks at the peak of their business cycle… mean reversion is faster when it is further from its mean
More on that in a moment, but first, every value investor’s favorite subject– low, low prices!
Gray and Carlisle pit several popular price measurements against each other and then run backtests to determine the winner:
- Earnings Yield = Earnings / Market Cap
- Enterprise Yield(1) = EBITDA / TEV
- Enterprise Yield(2) = EBIT / TEV
- Free Cash Flow Yield = FCF / TEV
- Gross Profits Yield = GP / TEV
- Book-to-Market = Common + Preferred BV / Market Cap
- Forward Earnings Estimate = FE / Market Cap
the simplest form of the enterprise multiple (the EBIT variation) is superior to alternative price ratios
with a CAGR of 14.55%/yr from 1964-2011, with the Forward Earnings Estimate performing worst at an 8.63%/yr CAGR.
Significant additional backtesting and measurement using Sharpe and Sortino ratios lead to another conclusion, that being,
the enterprise multiple (EBIT variation) metric offers the best risk/reward ratio
It also captures the largest value premium spread between glamour and value stocks. And even in a series of tests using normalized earnings figures and composite ratios,
we found the EBIT enterprise multiple comes out on top, particularly after we adjust for complexity and implementation difficulties… a better compound annual growth rate, higher risk-adjusted values for Sharpe and Sortino, and the lowest drawdown of all measures analyzed
meaning that a simple enterprise multiple based on nothing more than the last twelve months of data shines compared to numerous and complex price multiple alternatives.
But wait, there’s more!
The QV authors also test insider and short seller signals and find that,
trading on opportunistic insider buys and sells generates around 8 percent market-beating return per year. Trading on routine insider buys and sells generates no additional return
short money is smart money… short sellers are able to identify overvalued stocks to sell and also seem adept at avoiding undervalued stocks, which is useful information for the investor seeking to take a long position… value investors will find it worthwhile to examine short interest when analyzing potential long investments
This book is filled with interesting micro-study nuggets like this. This is just one of many I chose to mention because I found it particularly relevant and interesting to me. More await for the patient reader of the whole book.
Big and simple
In the spirit of Pareto’s principle (or the 80/20 principle), the author’s of QV exhort their readers to avoid the temptation to collect excess information when focusing on only the most important data can capture a substantial part of the total available return:
Collecting more and more information about a stock will not improve the accuracy of our decision to buy or not as much as it will increase our confidence about the decision… keep the strategy austere
In illustrating their point, they recount a funny experiment conducted by Paul Watzlawick in which two subjects oblivious of one another are asked to make rules for distinguishing between certain conditions of an object under study. What the participants don’t realize is that one individual (A) is given accurate feedback on the accuracy of his rule-making while the other (B) is fed feedback based on the decisions of the hidden other, invariably leading to confusion and distress. B comes up with a complex, twisted rationalization for his decision-making rules (which are highly inaccurate) whereas A, who was in touch with reality, provides a simple, concrete explanation of his process. However, it is A who is ultimately impressed and influenced by the apparent sophistication of B’s thought process and he ultimately adopts it only to see his own accuracy plummet.
The lesson is that we do better with simple rules which are better suited to navigating reality, but we prefer complexity. As an advocate of Austrian economics (author Carlisle is also a fan), I saw it as a wink and a nod toward why it is that Keynesianism has come to dominate the intellectual climate of the academic and political worlds despite it’s poor predictive ability and ferociously arbitrary complexity compared to the “simplistic” Austrian alternative theory.
But I digress.
Focusing on the simple and most effective rules is not just a big idea, it’s a big bombshell. The reason this is so is because the author’s found that,
the Magic Formula underperformed its price metric, the EBIT enterprise multiple… ROC actually detracts from the Magic Formula’s performance [emphasis added]
Have I got your attention now?
The trouble is that the Magic Formula equally weights price and quality, when the reality is that a simple price metric like buying at high enterprise value yields (that is, at low enterprise value multiples) is much more responsible for subsequent outperformance than the quality of the enterprise being purchased. Or, as the authors put it,
the quality measures don’t warrant as much weight as the price ratio because they are ephemeral. Why pay up for something that’s just about to evaporate back to the mean? […] the Magic Formula systematically overpays for high-quality firms… an EBIT/TEV yield of 10 percent or lower [is considered to be the event horizon for “glamour”]… glamour inexorably leads to poor performance
All else being equal, quality is a desirable thing to have… but not at the expense of a low price.
The Joe the Plumbers of the value world
The Quantitative Value strategy is impressive. According to the authors, it is good for between 6-8% a year in alpha, or market outperformance, over a long period of time. Unfortunately, it is also, despite the emphasis on simplistic models versus unwarranted complexity, a highly technical approach which is best suited for the big guys in fancy suits with pricey data sources as far as wholesale implementation is concerned.
So yes, they’ve built a better mousetrap (compared to the Magic Formula, at least), but what are the masses of more modest mice to do?
I think a cheap, simplified Everyday Quantitative Value approach process might look something like this:
- Screen for ease of liquidity (say, $1B market cap minimum)
- Rank the universe of stocks by price according to the powerful EBIT/TEV yield (could screen for a minimum hurdle rate, 15%+)
- Run quantitative measurements and qualitative evaluations on the resulting list to root out obvious signals to protect against risk of permanent loss by eliminating earnings manipulators, fraud and financial distress
- Buy a basket of the top 25-30 results for diversification purposes
- Sell and reload annually
I wouldn’t even bother trying to qualitatively assess the results of such a model because I think that runs the immediate and dangerous risk which the authors strongly warn against of our propensity to systematically detract from the performance ceiling of the model by injecting our own bias and behavioral errors into the decision-making process.
Other notes and unanswered questions
“Quantitative Value” is filled with shocking stuff. In clarifying that the performance of their backtests is dependent upon particular market conditions and political history unique to the United States from 1964-2011, the authors make reference to
how lucky the amazing performance of the U.S. equity markets has truly been… the performance of the U.S. stock market has been the exception, not the rule
They attach a chart which shows the U.S. equity markets leading a cohort of long-lived, high-return equity markets including Sweden, Switzerland, Canada, Norway and Chile. Japan, a long-lived equity market in its own right, has offered a negative annual return over its lifetime. And the PIIGS and BRICs are consistent as a group in being some of the shortest-lifespan, lowest-performing (many net negative real returns since inception) equity markets measured in the study. It’s also fascinating to see that the US, Canada, the UK, Germany, the Netherlands, France, Belgium, Japan and Spain all had exchanges established approximately at the same time– how and why did this uniform development occur in these particular countries?
Another fascinating item was Table 12.6, displaying “Selected Quantitative Value Portfolio Holdings” of the top 5 ranked QV holdings for each year from 1974 through 2011. The trend in EBIT/TEV yields over time was noticeably downward, market capitalization rates trended upward and numerous names were also Warren Buffett/Berkshire Hathaway picks or were connected to other well-known value investors of the era.
The authors themselves emphasized that,
the strategy favors large, well-known stocks primed for market-beating performance… [including] well-known, household names, selected at bargain basement prices
Additionally, in a comparison dated 1991-2011, the QV strategy compared favorably in a number of important metrics and was superior in terms of CAGR with vaunted value funds such as Sequoia, Legg Mason and Third Avenue.
After finishing the book, I also had a number of questions that I didn’t see addressed specifically in the text, but which hopefully the authors will elaborate upon on their blogs or in future editions, such as:
- Are there any reasons why QV would not work in other countries besides the US?
- What could make QV stop working in the US?
- How would QV be impacted if using lower market cap/TEV hurdles?
- Is there a market cap/TEV “sweet spot” for the QV strategy according to backtests? (the authors probably avoided addressing this because they emphasize their desire to not massage the data or engage in selection bias, but it’s still an interesting question for me)
- What is the maximum AUM you could put into this strategy?
- Would more/less rebalancing hurt/improve the model’s results?
- What is the minimum diversification (number of portfolio positions) needed to implement QV effectively?
- Is QV “businesslike” in the Benjamin Graham-sense?
- How is margin of safety defined and calculated according to the QV approach?
- What is the best way for an individual retail investor to approximate the QV strategy?
There’s also a companion website for the book available at: www.wiley.com/go/quantvalue
I like this book. A lot. As a “value guy”, you always like being able to put something like this down and make a witty quip about how it qualifies as a value investment, or it’s intrinsic value is being significantly discounted by the market, or what have you. I’ve only scratched the surface here in my review, there’s a ton to chew on for anyone who delves in and I didn’t bother covering the numerous charts, tables, graphs, etc., strewn throughout the book which serve to illustrate various concepts and claims explored.
I do think this is heady reading for a value neophyte. And I am not sure, as a small individual investor, how suitable all of the information, suggestions and processes contained herein are for putting into practice for myself. Part of that is because it’s obvious that to really do the QV strategy “right”, you need a powerful and pricey datamine and probably a few codemonkeys and PhDs to help you go through it efficiently. The other part of it is because it’s clear that the authors were really aiming this book at academic and professional/institutional audiences (people managing fairly sizable portfolios).
As much as I like it, though, I don’t think I can give it a perfect score. It’s not that it needs to be perfect, or that I found something wrong with it. I just reserve that kind of score for those once-in-a-lifetime classics that come along, that are infinitely deep and give you something new each time you re-read them and which you want to re-read, over and over again.
Quantitative Value is good, it’s worth reading, and I may even pick it up, dust it off and page through it now and then for reference. But I don’t think it has the same replay value as Security Analysis or The Intelligent Investor, for example.