web analytics

Quantitative Value Investing is Broken

This article first appeared on RealClearMarkets.com February 17th 2021

All truth passes through three stages.  First, it is ridiculed.  Second, it is violently opposed.  Third, it is accepted as being self-evident.”  Arthur Schopenhauer

Our last column, Yes Virginia, Valuation Matters, caused a bit of controversy on Twitter regarding the nature of Applied Finance’s Quantitative Valuation™ research and investment approach versus that espoused by the quantitative value approach.  It is a recurring theme in the marketplace that our Quantitative Valuation™ approach is confused with quantitative value investing, something we work diligently to avoid.  While Quantitative Valuation™ and quantitative “value” agree regarding the importance of objective repeatable factors and rigorous statistical testing, the approaches ultimately have very different worldviews to explain stock returns and construct portfolios.  Quantitative Valuation™ is deeply steeped in traditional security analysis and valuation to develop detailed and actionable company-specific opinions about intrinsic value, wealth creation, and corporate stewardship.  Applied Finance’s Quantitative Valuation™ identity is very clear – we systematically determine whether a company is undervalued through proprietary research we developed in 1995 and applied continuously through today to model each firm’s: economic profit, capital investment, risk, and competition.

We believe it is ill-advised to build a portfolio without a credible, time tested approach to understand the intrinsic value of the companies being bought or sold, While the market is efficient with respect to some information, it can be exploited using valuation and wealth creation concepts.

With this big picture in mind, we will expand on the following criticisms we have with quantitative “value” investing.

  1. Quantitative “value” investing has no identity.  While it began with the central idea to identify securities mispriced relative to book value, it has devolved into a factor hunt with no unifying theory to guide its development.
  2. The evidence standard used in quantitative “value” studies is low quality, consisting of observations subject to look-ahead bias and data snooping, with poor out-of-sample results.
  3. The theoretical motivation used to justify the “investment” factor reflects an incomplete understanding regarding valuation and wealth creation.
  4. Quantitative “value” investing as a discipline suffers from the fundamental research mistake of confusing correlation and causality.

Are Quant “Value” Investing Factors Relevant?

Quantitative “value” investing lacks a clear identity.  While its origins in the Fama/French 3 Factor Model stemmed from buying cheap stocks (high book to price) expecting them to be undervalued, it has devolved into hunting for factors that backtests show historically generate alpha, with little to no direct link to valuation.  In addition, there is no theory guiding the development or search for factors, which opens the door to an overreliance on empirical results regardless of whether they make sense. Currently, the non-market factors du jour in some form are: Size, Book to Price, Profitability, Investment, Price Momentum, and Volatility.  The naïve valuation theory underpinning this work is evident in the investment factor which favors companies shrinking their capital base over those that are reinvesting.  Based on in-sample research from 1963 to 2015, this is an empirical truth.  From a theoretical perspective, however, asserting that companies reinvesting in their business will generate negative stock returns relative to those shrinking their capital base regardless of their economic profitability violates the NPV rule, in addition to any reasonably complete theory of valuation as we will explore later.  Similarly, price momentum has nothing to do with a firm’s value.  Again, historic observation shows price momentum is a useful factor to explain cross-sectional stock returns, but is there a theory that links price momentum to intrinsic value?  Without a unifying theory, it is too tempting for this approach to simply argue for more factors that explain away periods of underperformance.  Where does it lead?  When does it stop?

Since 2014, quantitative value investing as a strategy has added 4 widely accepted factors to its arsenal – profitability, investment growth, momentum, and volatility.  That is a significant amount of change for what should be a stable, long-term approach to investing.   As Mike Tyson famously says, “Everyone has a plan until they are punched in the mouth.”  For “value” factor investing that punch has been dreadful underperformance over the last 1,3,5, and 10, years.  Rather than acknowledge the “value” factor is broken, and an unrealistic representation of intrinsic value, quantitative “value” investors chose to add more and more factors to explain underperformance while keeping “value” as a prominent component of how they communicate their investment philosophy.  Whatever quantitative value investing represents at this point is ultimately not important.  As detailed in our recent research, (Valuation Beta™ – Addressing the Inadequacies of Book to Price through Valuation, Stewardship, and Leverage), after incorporating the Intrinsic Value Factor™, Financing Yield, and Leverage, the common quantitative value investing factors generate statistically insignificant alpha.  This includes recent “innovations” such as capitalizing intangibles and creating composites of various cheapness variables, rather than using book to price alone.

Chart 1 depicts these commonly used quantitative value investing factors and their t-stats (in parentheses) to determine if the alpha from each factor is significantly different than zero versus the Applied Finance 5 factor model.  Notice that no common quantitative “value” factor is statistically significant, over the 22-year period measured from 1998 to 2020.  As a point of reference, quantitative “value” investing began with a live track record for its variables in 1992, after Fama/French released their 3-Factor model.  Beginning in 1998, Applied Finance began producing monthly live out-of-sample intrinsic value estimates on a consistent basis.  This 22-year overlap covers over 75% of the live, out-of-sample period since Fama/French launched quantitative value investing in 1992 with the “value” factor.  During this period, quantitative “value” investing has generated alpha from time to time, but it is dominated as an investment approach by Quantitative Valuation™, which focuses on comprehensive valuation and wealth creation concepts.

Chart 1. Annualized Factor Alpha vs Applied Finance 5 Factor Model

An alternative perspective is to ask – Do the popular quantitative “value” investing asset pricing approaches explain away the information embedded in Applied Finance’s Intrinsic Value Factor™, and the answer is clearly no.  Chart 2 below depicts the alpha (and corresponding p-values in parentheses) to the Intrinsic Value Factor™ after applying the following asset pricing models:  Fama/French 3 Factor, Fama/French 5 Factor, AQR 6 Factor, and the newest flavor which we will call 6 + volatility.  Regardless of model, unlike the quantitative “value” investing factors depicted in Chart 1, Applied Finance’s Intrinsic Value Factor™ generates statistically significant alpha.  Without a doubt, over the past 22 years in a live out-of-sample context, Applied Finance’s Intrinsic Value Factor™ has dominated what is considered “state of the art” finance investment theory.

Chart 2. Annualized Factor Alpha: Intrinsic Value Factor™

The Nature of Evidence Versus Observations

Another area where Applied Finance significantly differs from the “quantitative value” school is the interpretation of historical evidence.  Namely, when performing financial research with historical data, researchers already know how history plays out and that consciously or unconsciously guides variable selection.  This takes place via look-ahead bias and/or data snooping.  How a variable performs outside this known window determines whether it is useful for investment management.  For example, when Fama/French unveiled the “value” factor in 1992, the variable’s performance from 1963 to 1991 was extraordinary (as depicted in Chart 3).  A portfolio of the most attractive book to price stocks generated 4.87% annual alpha measured against the CAPM as depicted in Table 1.   However, after 1992 its performance has not lived up to its pre-1992 hype, with its annual alpha dropping to 0.88% as depicted in Chart 4 and Table 1.  Note that these results reflect a mix of Large, and Small capitalization stocks.

Table 1:  Expensive vs Cheap – Book to Price Portfolio Alpha (CAPM)
Large + Small

Focusing on Large capitalization stocks, the primary investable universe, the results are much worse as depicted in Charts 5 and 6, and Table 2.  During the observation period of 1963 to 1991, attractive Large Capitalization stock portfolios formed based on book to price generated annual alpha of 3.54%.  During the live, out-of-sample period, from 1992 to 2020, which best represent the returns to these strategies’ investors would have actually obtained, Large Capitalization stocks returned an annual alpha of -0.56%.

Table 2:  Expensive vs Cheap – Book to Price Portfolio Alpha (CAPM)
Large

Outside of the observation window, on a live out-of-sample basis, the “value” factor has been a spectacular failure for large capitalization stocks, and barely effective across all stocks.  The popular media narrative that “value” investing is having a temporary poor run is simply not true.  The “value” factor has performed poorly or failed on a risk adjusted basis since being live and out-of-sample.  If Fama/French began their research in 2020, the “value” factor most likely would not exist.

The disparate performance of data that is studied in retrospect versus its subsequent performance is such a significant problem in finance research that Harvey, Liu, and Zhu published a paper in 2015, titled …And the Cross-Section of Expected Returns, discussing the problem in detail.  The point they make is that it is too easy to bring accidental bias, and/or mal intent when performing research on known data.  They argue, that as in the “hard” sciences, the gold standard to evaluate any variable is to develop it from a strong theory and determine its efficacy through live, out-of-sample testing.  They acknowledge that live, out-of-sample testing is rarely available. As a result, any variable showing statistical significance but supported by weak theory should be ignored or viewed with extreme skepticism until it proves its worth out-of-sample.

Historical observations are a low-quality form of evidence to assume something works going forward.  How safe is a vaccine developed and tested in the laboratory but not yet verified through animal or human trials?  This “in the lab” versus “outside the lab” concept applies to financial research as well.  We view this reliance on “in the lab” historical data as low-quality evidence to support an investment strategy, especially absent a strong, clearly accepted theory to motivate its use.  Until a variable is tested “outside the lab”, the evidence consists of mere observations.

The Intrinsic Value Factor™ data used in our “deconstructing” value research and our asset pricing model is consistent with the Harvey, Liu, Zhu gold standard given the length of its consistent, live, out-of-sample duration.  Our data is derived from a theoretically sound valuation framework, consistently produced live and out-of-sample monthly for over 22 years.

Applied Finance launched in 1995 with one objective – create a testable, rigorous, detailed company analysis framework that consistently and systematically answers two questions:

  • What is a firm’s economic performance?
  • What is a firm’s intrinsic value?

We began our research through observations in 1995.  Specifically, we observed and performed research on financial and economic data spanning 1970 to 1995.  At the time, our thinking was influenced by the effects of inflation in the 1970’s and the tax policy debate from the 1980’s.  We understood what happened and reflected that knowledge in our models but never imagined a world of zero to negative interest rates or the “winner take all” economics exhibited by many technology companies today.  This only solidifies our belief that live, out-of-sample data is the only way to truly evaluate the efficacy of an approach.  While we built a valuation model that explained the known results from 1970 to 1995 well, we had no idea how it would explain subsequent years. We were confident in the strength of our theory and complete valuation approach to adjust to a changing world, however, as ultimately valuation reflects – Economic Profit, Capital Growth, Risk, and Competition.

Incidentally, our 25-year observation period is similar in scope to the Fama/French 1963 to 1991 28-year observation period.  From 1995 into 1996, we created our approach to measure a company’s economic performance and intrinsic value.  We called our approach the Economic Margin® framework, (Frank Fabozzi published a detailed discussion of the Economic Margin® approach in 2000).  Among notable attributes of our early work: capitalizing R&D, measuring returns to intangible assets, avoiding terminal value assumptions, sector and industry-specific standardization, capitalized leases, inflation-adjusted financial statements, international accounting differences, market-derived costs of capital, and company-specific size and leverage risk factors among many others.  We initially applied our research through corporate consulting, where numerous interactions with finance and operations personnel helped us sharpen our insights into the nature of R&D investing, international cost of capital, and inflation adjusted financial statements.

By July of 1998, we finished calibrating our model ending our (“in the lab”) observation period.  Starting in August of 1998, we produced and distributed our Economic Margin® and Intrinsic Value Factor® data monthly and then weekly.  Since 1998, we have systematically estimated over 20 million intrinsic value estimates for companies worldwide.  Each week we calculate approximately 20 thousand intrinsic value estimates that like all the data before it, is time-stamped and delivered to our clients.   It is interesting to compare the live, out- of- sample performance of the Intrinsic Value Factor™ vs book to price (the “value” factor).  The results for the Large and Small cap universe are depicted in Charts 7 and 8, while the results for the Large cap universe are depicted in Charts 9 and 10.   The performance domination of the Intrinsic Value Factor™ over book to price is consistent and complete over the past 22 years on an absolute and risk adjusted basis.  Book to price’s performance in the Large Cap space over this period is particularly poor, but the Intrinsic Value Factor™ performs as expected with undervalued stocks significantly outperforming overvalued stocks.

Table 3:  Book to Price vs Intrinsic Value Factor ’98-’20 (CAPM Alpha)
Large + Small

Table 4:  Book to Price vs Intrinsic Value Factor ’98-’20 (CAPM Alpha)
Large

The above Charts/Tables and our Valuation Beta™ study reference only the live out-of-sample valuation data we produced since 1998.  As a result, the valuation data used throughout our studies is of the highest quality.

Conversely, quantitative “value” investing relies heavily on historical observations, subject to look-ahead bias and data snooping, to justify their theories and changes to their models.  They cite historic data observations nonchalantly as being evidence, we disagree.  Thirty years ago, screening on Yahoo! Finance did not exist, but now anyone can screen on “state of the art” factors for free and construct a world class quantitative “value” portfolio.  Does this change the usefulness of conclusions derived from historical data?

Applied Finance’s Intrinsic Value Factor™ has 22 years of live, out-of-sample data.  Given the ongoing changes to quantitative “value” investing from the Fama/French 3-factor model, quantitative “value” investing has continuously reset the start point of live out-of-sample data. There is no shortcut to reach 20 + years of live, out-of-sample data, it is only obtained by having the conviction to produce and follow a strategy for 20+ years through good and bad times.  Regardless of how many years of experience practitioners of quantitative “value” investing have, each time a model undergoes a significant change, the track record for live out-of-sample data for that model resets to zero.

The importance one puts on the results obtained from historical observation is a personal decision.  We agree with Harvey, Liu, Zhu that the quality of results derived from historical observation is of significantly lower quality than results obtained from a live, out-of-sample period.  Equipped with over 22 years of live out-of-sample data, Applied Finance has set the standard for providing evidence to explain stock returns.  Stated simply, evidence can only be ascertained through live data, anything less is an observation.

Deconstructing Value: Correlation vs Causality

A common problem with research not based on strong fundamental theory is finding statistically significant results that stem from correlation to other variables unknown to the researchers.  This is the case for book to price and various other cheapness factors used throughout the quantitative “value” literature.  Accounting book value is an inadequate proxy for a firm’s intrinsic value.  As mentioned above, intrinsic value requires understanding – economic profitability, capital growth, risk, and competition.  Accounting book value reflects none of those intrinsic value attributes.  Regardless, book to price and intrinsic value to price have some correlation.  Specifically, intrinsic value can be expressed as a function of book value in the following manner for an all-equity financed firm:

Intrinsic Value = Book Value + Present Value (Future Economic Profits)

For instances where the present value of future economic profit is small, intrinsic value and book value will be similar.  As discussed in the growth section below, when firms with positive Economic Margins® invest, their intrinsic value will grow relative to book value.  Conversely, if firms have negative Economic Margins® and invest their value will shrink relative to book value.  Book value is a backward-looking static measure that fails to reflect ongoing wealth creation and destruction.

Whether a backward-looking variable such as book value is superior to a forward-looking intrinsic value estimate is an easy research question to address.  To do so, we compare portfolios of stocks sorted on book to price and intrinsic value to price (Intrinsic Value Factor®) to understand the return properties of each variable and determine if one variable adds significantly to the other in explaining future stock returns.  Table 5 and Chart 11 depict these results.  We sorted all US stocks in our database into portfolios consisting of bottom 30%, middle 40%, top 30% sorts of book to price and the Intrinsic Value Factor™ and rebalanced them monthly.  We find high book to price stocks add positive alpha, but only when supported by undervalued stocks as ranked by the Intrinsic Value Factor™.  High book to price stocks that are fairly or overvalued as determined by the Intrinsic Value Factor™ generate negative alpha.  Conversely, low to middle book to price stocks corresponding to undervalued stocks as determined by the Intrinsic Value Factor™ generate positive alpha.  In summary, regardless of a stock’s book to price ranking, undervalued stocks as determined by the Intrinsic Value Factor™ generate positive alpha.  Book to price or the “value” factor, however, works because of its correlation to the Intrinsic Value Factor™.  Book to price is a failed, flawed cheapness concept and it is deceptive and unproductive to refer to it in a value context.

Chart 11 and Table 5: Deconstructing Value 
’98-‘20

Understanding Wealth Creation

In addition to advocating out-of-sample data to evaluate a variable’s performance value, Harvey, Liu, and Zhu also advocate that any variable needs to have a strong economic theory to support its research and use.  In developing the investment factor used in quantitative “value” investing models, Fama/French derived their investment factor from a dividend discount model.  They structured their model such that if a firm invests additional capital, that investment generates no future dividends, only reducing firm value.  A testable implication of that model is that companies investing in their business generate lower future stock returns than those shrinking their asset base.  Empirically they found that firms investing in their business, had lower returns than those did not.  This is a very naïve view of the corporate investment process, as it misses the interaction of investment and future cash flows.  The point of making investments is to generate additional future cash flow.  The underlying worldview represented by this research is systematically biased against investing in the greatest wealth compounding companies throughout time.  This is not a recent “tech” company issue.  Tech and non-tech firms such as Amazon, Microsoft, Alphabet, McDonald’s, Walmart, and Pfizer, all generated incredible wealth by aggressively reinvesting into their businesses.  This worldview is at odds with the basic principles of valuation – investments with returns above a firm’s cost of capital will increase a firm’s value.  In a curious thought experiment, in the early 1990’s, would the world’s leading quantitative “value” investing portfolio managers advise Jeff Bezos not to expand beyond selling books?

The problem with how Fama/French modeled valuation is they missed the distinction between wealth-creating and destroying investments.  Firms investing in projects with returns greater than their cost of capital create wealth.  Firms investing in projects with returns below their cost of capital destroy wealth.  Table 6 illustrates the problem simply.  Indeed, high investment firms, with low levels of operating profitability have poor returns, relative to low investment, low profitability firms.  However, firms with high investment and high profitability have the highest rates of return from 1998 to 2020 of all the possible profitability/growth combinations.  (For a detailed discussion, please review our Valuation Beta paper cited above.)

Table 6: Operating Profitability & Investment Rate Portfolios ‘98-’20
Large + Small

Failing to properly partition firms leads to the generic, but incorrect conclusion regarding investment growth, advocated by Fama/French in constructing their 5-Factor model.  This is an unrealistic representation of the world, which perfectly illustrates the Harvey, Liu, Zhu criticism of empirical work requiring a strong theoretical foundation to evaluate in-sample data.  Yet virtually every major quantitative “value” firm has adopted this incorrect worldview – indiscriminately penalizing all firms that are investing in their business.  An investment consultant recently told me – “all these quant value firms seem to imitate each other to the point it is difficult to differentiate one from another”.  His statement reminded me of the Bugs Bunny skit where Bugs has a guitar and says – “This is my imitation of Ricky imitating Frankie, imitating Elvis”.

Applied Finance strongly disagrees with quantitative “value” investing’s treatment of growth.  Our Economic Margin® Framework provides a complete valuation perspective to estimate a company’s intrinsic value.  Intrinsic value is the intersection of: Economic Profitability, Capital Growth, Risk, and Competition.  In the Economic Margin® framework, to the extent firms invest in their businesses with Economic Margins® greater than zero, their intrinsic values will increase, consistent with the NPV rule.  The blanket assumption that all growth leads to negative returns is a blind spot of quantitative value investing which have historically ignored security analysis.

Quantitative Valuation and Quantitative Value – Differentiation

Given the philosophical, theoretical, and research process differences between Applied Finance’s Quantitative Valuation™ and quantitative “value” investing, naturally the investing approaches are equally different.  Applied Finance’s Quantitative Valuation™ approach reflects a stock’s attractiveness regardless of the underlying market oscillations between “growth” and “value” stocks.  Quantitative Valuation® properly executed is indifferent to slow vs. fast growth companies.  All that matters is the intrinsic value impact of a firm’s growth and what is the market price to own it.  Again, using live, out-of-sample data, this is readily apparent in the composition of our longest-running large-cap strategy (started in 2004), the Valuation 50™.  In 2014, the Valuation 50™ was approximately evenly invested in value/growth stocks as defined from a book to price perspective.  Since then, as growth has become increasingly favored by investors pushing up the price of “growth” relative to “value” stocks, most of the Valuation 50™ marginal transactions have been in “value” stocks. Today, the strategy consists of approximately 70% “value” and 30% “growth” companies.  Interestingly, our most recent addition to the portfolio in August 2020 was an undervalued growth stock, highlighting how true valuation is not defined by a simplistic “growth/value” dichotomy.  This ability to understand valuation has served our investors well. The Valuation 50™ strategy has outperformed the S&P 500 since its inception in 2004, and over the past 10 years, while quantitative “value” strategies have significantly underperformed.

We think there is little doubt the investment research profession will embrace the models and methods Applied Finance developed 25 years ago.  Existing approaches rely on overly simplistic metrics to represent an increasingly complex economic reality and have failed relative to a complete valuation approach over the past 22 years.  Finance professionals are very competitive, and they will adjust their thinking and copy successful approaches.  For any investor evaluating these new future potential models, it is instructive to always remember that it is impossible to shortcut the 20 plus years of live out-of-sample data required to evaluate the efficacy of a complete valuation framework.  Further live, out-of-sample data reflect the model, not the modeler.  For example, make a significant change to a model to capture a rising interest rate environment and the track record gets reset to year zero.  Applied Finance does not claim to own valuation, but it is the only investment manager that has invested the time and developed the expertise to have harnessed Quantitative Valuation™ since 1998 through domestic and international market booms and busts, economic expansions/recessions, and sadly a 100-year pandemic.  Expertise and track records matter

Quantitative “value” investing is an amazing commercial success, reflecting the desire for a consistent repeatable process by investors to “stay the course” for the long-term.  As presented in our Valuation Beta™ research, Applied Finance’s Valuation and Stewardship factors have distinctly different and demonstrably superior investment properties than those commonly advocated by quantitative “value” proponents.   Responsible advisors will take the time to understand the differences to determine if this approach makes sense for their clients.

For updates on our investment strategies and more articles like this, join our Valuation Edge™ newsletter.

Messrs. Obrycki and Resendes Co-Founded Applied Finance in 1995.

Authors

RECENT COMMENTARY

Valuations in the Energy & Utility Sectors

Looking at the most recent aggregate Valuation data, both the Russell 1000 and Russell 2000 look expensive relative to their long-term levels. On a sector basis, large and small cap Energy stocks, and large cap Utility stocks all seem attractive, while small cap Utilities appear expensive. […more]