AgPa #38: Value Investing – Fact and Fiction

Fact, Fiction, and Value Investing (2015)
Clifford Asness, Andrea Frazzini, Ronen Israel, Tobias Moskowitz
The Journal of Portfolio Management Fall 2015, 42(1) 34-52, URL/AQR

After busting some myths around momentum, this week’s AGNOSTIC Paper is the sequel for value investing. The authors, the same AQR crew as last week, present several Facts and Fictions around value investing which is actually one of the oldest investment styles out there.

Everything that follows is only my summary of the original paper. So unless indicated otherwise, all tables and charts belong to the authors of the paper and I am just quoting them. The authors deserve full credit for creating this material, so please always cite the original source.

Setup and Idea

What is value investing? Nowadays, there are many different nuances of value but the underlying phenomenon remains unchanged. Securities that look cheap (vs. some fundamental metrics) tend to outperform securities that look expensive. This so called valued premium is a well-researched empirical fact and has been documented in many studies across different sample periods, geographies, and even in other asset classes.

Before we continue, there is one important distinction to understand what is coming. There are two general ways of being a value investor: systematic or discretionary.[1]The words are not perfect because systematic value investors also have discretion about their system, and discretionary value investors may also use systematic procedures. But most people get what you mean by this… Discretionary value investors typically try to deeply understand and evaluate the underlying company. Usually they only invest if they find something that trades below their estimate of “intrinsic value”. This approach is obviously time consuming and you can only do this with a small number of companies. Portfolios of discretionary value investors are therefore typically quite concentrated. Needless to say, the most prominent value investor in this camp is Warren Buffet.

Systematic value investors, in contrast, don’t look deeply into the underlying companies and rather bet on the aforementioned theme that cheap securities tend to outperform expensive ones. For this purpose, they use signals like valuation ratios to construct large and diversified portfolios. As a consequence, the portfolios of systematic value investors look very different although they bet on the same underlying theme.

This article (and the underlying paper) almost exclusively focuses on systematic value investing which is also known as the value factor. But some of the issues are also quite general and certainly relevant for both approaches.

Data and Methodology

Similar to last week, there is not so much to say about specific data or methodology. The paper is an overview about value investing and some misconceptions around it. That said, the authors use a lot of data and references to back up their arguments. To be as transparent as possible, they use the well-known and publicly available value factor from Kenneth French’s website and an improved version from AQR for most of their analyses.

Important Results and Takeaways

Fiction: Value investing requires concentrated portfolios

The first fiction is a direct result of the two general value-approaches that I mentioned before. Buffett and his partner Charlie Munger are famous for their view that “Diversification is protection against ignorance.” and that “It makes little sense if you know what you are doing.”

I will go into the details if this statement is really true and whether diversification makes sense even when you know what you are doing.[2]Okay, I couldn’t resist. Diversification in the sense of just buying many things is indeed not very useful. The key is to buy different things that are independently attractive and I am pretty sure Buffett also understands and applies that whenever possible… In fact, there isn’t really much to say about this fiction. There are different ways to implement a value strategy and if executed properly, both deserve their spot in the investing landscape. What is not true, however, is that you must deeply understand businesses and run a concentrated portfolio to earn the value premium. Systematic approaches historically also delivered outperformance and often did so at lower cost and less active risk.

Fiction: Value has low turnover and is thus passive

This fiction is an easy one and a direct consequence from common misunderstandings of passive investing. I have written a lot about the academic idea of passive investing, the global market portfolio, various “passive” investment products, and I refer more interested readers to those articles. With respect to (systematic) value, the authors explain that some people define it as passive because of its low turnover and because it follows a rules-based process.

If you read my previous posts, you know what is coming. From an academic (and logical) point of view, passive investing means holding the market portfolio. This is the value-weighted portfolio of all available securities. Restricted to the stock market, passive investing therefore means holding a global market-cap weighted index fund, for example an ETF on the MSCI ACWI.[3]The MSCI ACWI is of course just a (sufficently good) proxy for passive equity investing because it doesn’t include all available stocks. No matter how much turnover, rules-based or instinct-driven, any portfolio that deviates from this benchmark is a form of active management. Given that value investors overweight cheap securities, they are not passive and will never be.

Fact: Fundamental indexing is similar to systematic value investing

With the rise of ETFs and increasing popularity of factor strategies, some asset managers used this tailwind to create ETFs that attempt to deliver simple and transparent implementations of things like value, momentum, or defensive. Such products became widely known as smart beta or fundamental indexing.

Especially the latter part is relevant for value. The authors show that once you use some fundamental variables to determine the weights of your portfolio, which is the whole idea of fundamental indexing, you are implicitly running a systematic value strategy. The main problem the authors have with such investment products? Fundamental indexing is often sold as new and innovative approach to portfolio construction although it was historically just strongly correlated to the well-known value factor. The authors thus conclude that investors should be careful about investment processes that claim do to something new but are ultimately just different interpretations of a well-known phenomenon.[4]Needless to say, ETFs on more creative indices often come with higher management fees…

Fact: Profitability signals improve value investing

Everyone who run at least one value screen for himself knows that. Many companies with a low multiple are probably cheap for a good reason. Fundamental problems, scandals, or just bad prospects for the industry – you will typically find something that justifies a low multiple. Against this background, it seems obvious to go beyond valuation ratios and also consider some fundamental quality characteristics in the investment process. Both systematic and discretionary value investors like Buffett adapted accordingly over time.

The authors show that a combination of the value and profitability factor generated much better risk-adjusted returns than value alone. If you also add momentum, it gets even better. In my opinion, this is quite intuitive. When cheap stocks already outperform expensive stocks without additional factors, cheap stocks with high profitability and strong momentum should do even better. I think this isn’t too surprising.

Fiction: Value is redundant in modern factor models

This fiction requires some background on the academic literature about factor models. I will give my best to not bore you too much with theory. Value is one of the ingredients to Fama & French’s (1992) famous three-factor-model (together with the overall market and size). As more and more researchers studied factors and market anomalies, Fama & French (2015) expanded their model to five factors and included a profitability and investment factor.[5]I discussed profitability in the previous fact. The investment factor captures the empirical phenomenon that companies with conservative asset growth outperformed those with more aggressive investment. According to the authors, one criticism of the five factor model is the fact that value, statistically speaking, no longer adds information in this setup. Once you appropriately examine stocks by their exposure to the market, size, profitability, and asset growth, you (statistically) no longer need to look at value.

This sounds already a bit shaky, right? The authors agree and present two compelling arguments why these results are most likely misleading. First, against all criticism, the Fama-French five-factor-model ignores momentum. Fama and French often argue that there are few plausible economic theories to justify a momentum premium and don’t want to add a factor solely based on the idea of irrational mispricing. However, from last week’s post we (hopefully) know that the evidence for momentum is actually quite compelling. Momentum is extremely well-researched and most practitioners agree that it is a valid factor.

The second argument is more technical. In their model, Fama & French construct the value factor based on Price/Book multiples with considerable time lags. This is good scientific practice to avoid look-ahead biases in the data. In practice, however, it is quite unrealistic that investors use the Price/Book multiples from six months ago to evaluate a stock. Furthermore, today’s point-in-time databases also help to use unbiased data. In another paper, the authors thus construct an improved value factor (HML Devil) and use it to clarify this fiction.[6]HML Devil comes from the title of the other paperThe Devil in HML’s Details. Have I already mentioned that those guys are my seemingly unattainable idols with respect to titles?

Once you add momentum and use the improved value factor, value is no longer redundant. So the problem doesn’t seems to be some non-existing redundancy of value, but an incomplete factor model of Fama & French.

Fact: Value also works in other asset classes

A general theme of the last two articles (and my website in general, hopefully…) is robust statistical evidence. The most effective tool to achieve this are out-of-sample tests. That means to test existing ideas in completely different setups. With respect to factors, different asset classes are very appealing environments. According to the authors, however, some people question that value is applicable to other asset classes. In some way, this is understandable. You will easily find valuation ratios for stocks, bonds, and equity indices, but what exactly makes a commodity or currency cheap?

Well, researchers and practitioners came up with various value-signals for almost all asset classes. The authors again cite their own research which shows that value historically worked among individual stocks, equity indices, bonds, and commodities. The assets are different but the underlying theme is always the same – cheap securities tend to outperform expensive ones. In fact, the authors even show that value premiums within different asset classes are stronger correlated than the respective long-only benchmarks. Apart from profitable investment opportunities, this again suggests that the underlying theme (cheap beats expensive) appears to be very robust.

Fact: Value is best measured by a composite of signals

To bet on cheap stocks you obviously need to measure what cheap actually means. There are unlimited plausible ways to do this. Fama & French started with Price/Book, but Price/Earnings, Price/Cash Flow, EV/EBITDA, and whatever other fundamental multiple are also fine within a value process.[7]Fama & French already used several other fundamental valuation ratios in their famous paper, but they are still best known for the Price/Book value factor… I think it is quite intuitive that we should rely on more than just one value measure to be as robust as possible.

In fact, the authors argue that if you frame it the opposite way, the idea that there should be one single best measure of value “[…] seems at best remote and most likely false.” This is of course also related to the aforementioned out-of-sample tests and robust statistical evidence. If you have an idea and test it with data, it shouldn’t break down when using different ways to measure the same phenomenon. With respect to value, one particular multiple will of course look better or worse in a backtest. What is important though, is that the results are generally similar (cheap outperforms expensive).

The authors also mention a very important point that is relevant for all investment processes. If there are no plausible reasons why one measure is better than the other, taking the average of all is typically the best way to go. It is tempting to pick the multiple that generated the best backtest, but this is obviously data mining and very dangerous. In fact, the authors show that a simple average of four value-signals lead to a much more consistent performance through time. If there are several plausible ways to measure the same thing, use them all.

Fact: Value is quite weak among large caps

While momentum was special because it historically worked somewhat better among large caps, value is the exact opposite. Research and realized performance of investment products shows that value performed much better among small caps. In fact, Small Cap Value is now a standalone segment within equity markets.

The authors split Kenneth French’s value factor by market capitalization to illustrate this fact. The value premium among large caps was historically much lower than for small caps and during some periods even statistically insignificant. However, once you again combine value and momentum, this pattern disappears. A 60/40 portfolio of value and momentum historically delivered similar risk-adjusted returns among both large and small caps. But standalone value, and that’s the fact, indeed performed much better among small caps.

Fiction: The value premium should disappear because there are no plausible explanations

The grand final is again the standard question. If we all know about value investing since Graham & Dodd in the 1930s, why should it still work? If you read the previous posts, you already know the answer. Because there are plausible reasons why a value premium exists and because it is not always easy to stick with it! As for all factors, explanations either come from the risk-based/rational or behavioral camp. Researchers still debate which side is more relevant, but for investors it is probably sufficient to know that there actually are some plausible reasons out there.

Risk-based theories of value suggest that the factor compensates some non-diversifiable risk. The fact that value occasionally moves through painful and long drawdowns (like in the early 2000s, or currently since 2018) supports this idea. Apart from that, researchers also came up with the idea that value compensates distress risk. As I mentioned before, many cheap companies are cheap for a reason. So the value premium could be a compensation for holding companies that are more likely to go through major distress. While plausible at first glance, the authors provide two arguments against this idea. First, value also worked among commodities and I think we all agree that it is quite difficult for a commodity to go bankrupt. Second, if value compensates distress risk, the factor shouldn’t work better in combination with the distress-mitigating profitability factor. Within the risk-based theories, the “going through painful periods” idea therefore appears more reasonable.

The leading behavioral explanation is the idea of investor overreaction. This is a well-documented pattern from behavioral finance and suggests that (some) investors tend to overreact on new information. The idea is that through this mechanism, (some) investors are too optimistic about growth stocks and too pessimistic about value stocks. Growth therefore becomes too expensive, value becomes too cheap, and value investors make money from a delayed correction. Proponents of risk-based theories usually criticize that this explanation requires the strong assumption that investors don’t learn from their or previous generations’ mistakes. This is of course true, however, I believe it is fair to doubt investors’ ability to learn when comparing the period around 2000 with the boom among non-profitable growth stocks or cryptocurrencies after the COVID crash in 2020.

In reality, both explanations are probably active at the same time. More importantly, both provide plausible reasons why the value premium shouldn’t disappear in the future. But even if the value premium disappears, value would still be attractive within a multi-factor portfolio. The authors repeat the simple multi-factor optimization that they also did for momentum and show that even at a negative premium of -1%, value would still receive a weight in a multi-factor portfolio because of its attractive negative correlation to momentum.

To summarize this article on value, I think we need a little more humility than for momentum. There is empirical evidence for a value premium, but the out-of-sample results for the simple Fama & French factor over the last years are arguably disappointing. To resurrect (systematic) value investing, you need to do view the strategy in a multi-factor context and rely on better definitions. After adjusting for that, however, value remains a very valuable investing strategy.[8]Value remains valuable… Believe it or not, this was not on purpose. We have robust empirical evidence and plausible reasons why the premium exists and should continue to do so.

Due to the still ongoing drawdown since 2018, the value factor currently looks very attractive. AQR and Robeco frequently share their estimates of the value spread, a simple measure of the valuation-gap between cheap and expensive stocks.[9]I should mention that both AQR and Robeco offer systematic value strategies. Despite strong value-performance in 2021 and 2022, this spread still stands at the >90% quantile of its own history. This indicates that the value factor itself is currently very cheap and given all the evidence for a value premium, I therefore believe it is worth a closer look right now![10]Full disclosure: I implemented a systematic value tilt in my own portfolio at the beginning of 2022 and still hold it.

This content is for educational and informational purposes only and no substitute for professional or financial advice. The use of any information on this website is solely on your own risk and I do not take responsibility or liability for any damages that may occur. The views expressed on this website are solely my own and do not necessarily reflect the views of any organisation I am associated with. Income- or benefit-generating links are marked with a star (*). Please also read the Disclaimer.