Factor investing with value strategies

Price is what you pay. Value is what you get.

Warren Buffett

Background on value investing

Value investing is an investment strategy that involves picking stocks that appear to be trading for less than their intrinsic value (in other words, “cheap” stocks). It involves using fundamental analysis to identify stocks that have been systematically undervalued, which is the epitome of the “buy low, sell high” adage.

One of the most influential persons behind the idea is Benjamin Graham. The idea is encapsulated in his influential 1949 book, “The Intelligent Investor”. In the book, he outlines several criteria to identify value stocks, which can be broadly classified into two groups: accounting-to-price (P/E -> E/P, P/BV -> B/P, Div/P) and accounting-only (debt ratio, current ratio, earnings growth). An investor would screen stocks based on some pre-determined criteria and the filtered stocks should be more likely to be winners than losers. This screen can, in essence, limit the universe of investment opportunities so she can do further due diligence before making the decision. She may decide to invest in just a handful of the list after doing additional research, and that’s okay.

Note that the screen can be simplistic and may disagree with qualitative assessments of the stocks, especially in modern setting where today’s accounting numbers may not fully reflect the potential of the business.

Factor investing with value strategies

In today’s terminology, factor investing can be considered a form of data-driven investing that removes narratives and emotions for investment choices and instead rely on rules/algorithms that turn signals into actions. In other words, factor investing doesn’t choose stocks based on their individual merits but instead on some common characteristics that systematically predicts higher returns. If there are 100 stocks in a portfolio, as long as there are more winners than losers and the portfolio returns beat a fair benchmark (fair can be an ambiguous word – we’ll discuss this further below), then it’s a good investment strategy. In factor investing, pick characteristics, not stocks.

In factor investing, we pick characteristics, not stocks.

So why does factor investing work this way? It’s a byproduct of asset pricing research whose objective is to understand determinants of returns beyond the mathematically- and/or economically-motivated capital asset pricing model (CAPM) which predicts that variations in returns are driven only by systematic risk (which is measured as beta); all other variations are random and can be diversified away.

Unfortunately, that prediction hasn’t held up very well to empirical scrutiny. Researchers have documented that other common characteristics, such as size (Banz, 1981), E/P ratio (Basu, 1983), B/P ratio (Rosenberg, Reid and Lanstein, 1985), S/P ratio (Barbee, Mukherji and Raines, 1996), CF/P ratio ( Lakonishok, Schleifer and Vishny (1994) ), and OCF/P ratio (Desai, Rajgopal and Venkatachalam, 2004) are also associated with higher returns, even when accounting for their CAPM beta.

This has led to a movement among academics to extend asset pricing models beyond the CAPM by establishing theoretical frameworks on asset pricing/valuation that incorporate such characteristics explicitly into the model. One of the most influential model is the Fama and French (1993) three-factor model that incorporates portfolios constructed on market cap and B/P ratio as two additional factors (Fama and French call it the ratio B/M ratio). The two additional factors are called the size factor and value factor. It has since been extended several times to incorporate other factors, such as profitability, investment – and momentum in their latest version, Fama and French (2018).

Can factor investing be considered skills?

There are actually many asset pricing models and academics haven’t really reached a consensus on what’s the “best” model. But at the end of the day, we may never need to. Asset pricing factors are generally long-short (zero-cost) portfolios constructed from characteristics that are associated with systematically high returns (we call them factor-mimicking portfolios), which makes them investment strategies in their own right. What academics refer to as asset pricing models are essentially what we define as commonly accepted determinants of returns, to be used as benchmark.

Factors that are outside of a model are often referred to by academics as anomalies, and they generate alphas relative to an asset pricing model. The same alphas are regarded as skills in the world of investment management. In other words, what we classify as skills depends on how our choice of benchmark, which brings our discussion back in full circle to the earlier point of what constitutes fair. If we assign a benchmark that doesn’t incorporate some factor investing strategy, when the strategy is successful, it will appear as an alpha relative to the benchmark. If we do, however, the alpha will no longer be there since it is part of the benchmark.

In addition, since factors are in principle transparent rules, it means their returns can potentially be replicated by anyone, both individuals and professional. In fact, there’s evidence that academic publications can destroy alphas as investment strategies become more widely known and trade volume diminishes returns (we call this a crowded trade). See, for example, McLean and Pontiff (2015).

There are many practical decisions involved in designing factor investing strategies.

The silver lining, however, is that there are many practical decisions involved in designing factor investing strategies. For example, do I use annual versus quarterly financial statements to construct signals? Do I use a single or multiple characteristics to form signals? Should I rank signals based on raw data or make some adjustments? What weights do I assign to stocks when I form my portfolio? How frequently do I rebalance my portfolio? Many factor investing funds do not explicitly tell investors how they process signals to form their portfolios, and that’s probably a good thing. When insights can easily become public, there’s less incentive to generate them, as encapsulated by the Grossman-Stiglitz paradox in their seminal 1980 paper titled “On the Impossibility of Informationally Efficient Markets”. Here’s the key insight from the paper:

If competitive equilibrium is defined as situation in which prices are such that all arbitrage profits are eliminated, is it possible that a competitive economy always be in equilibrium? Clearly not, for then those who arbitrage make no (private) return from their (privately) costly activity.

Grossman and Stiglitz (1980)

Since we can’t grant temporary monopoly rights like other types of intellectual property, there would be no incentive to innovate if innovators cannot capture private gains to offset their private costs while others free-ride. Sometimes, opacity is necessary for the market to function. Proprietary skills in how factor investing strategies are designed are thus investment edges.

Back to value investing…

Value strategies are often based on signals calculated from ratios of some accounting numbers relative to stock price, but we will see later that the classification of “value” in factor investing is multifaceted, mainly depending on what one defines the opposite of value is. In other words, if a value stock is “cheap”, then is a not-value stock “expensive”? Are there alternative explanations?

Notice that Benjamin Graham’s accounting-only criteria involve aspects which could be considered “quality” but doesn’t explicitly include profitability. In a future post, I will talk about other accounting-only criteria which underpin other popular factor investing strategies as well.

Value stocks (high B/M) are considered against growth/glamour stocks (low B/M) – see, for example, Lakonishok, Schleifer and Vishny (1994). In the early days, fundamentalists argue that value stocks are riskier (thus require higher discount rates, earning higher rates of return) and the value premium is compensation for bearing risk, while behavioral economists argue that investors are overly pessimistic about value stocks and overestimate future earnings of growth stocks. Both of these arguments are consistent with empirical evidence and cannot be distinguished without deeper investigation.

This highlights the challenge in asset pricing research that we often observe prices and returns which can be influenced by many reasons, from investment fundamentals to human psychology, or even mispricing. It is worth highlighting that in the end, we only observe price, not value. The efficient market hypothesis notion that price = value requires many assumptions to be true, most notably that investors are very capable, act rationally and markets are fully liquid. Violations of such assumptions can lead to price-value deviation and thus profit-making opportunity. We saw a glimpse of this in my earlier post on momentum investing.

In any case, researchers try to discern the sources of these anomalies. For example, Doukas, Kim and Pantzalis (2004) posit that if value-growth divergence is caused by disagreement among investors regarding future earnings, then the returns should be related to disagreement. Indeed, they find that disagreement identified via dispersion in analyst forecasts can explain the returns differential.

Factors are often related to one another. For example, Yan and Zhao (2011) relate the value-growth divergence to the post-earnings announcement drift (referred to as earnings momentum in the earlier post on momentum investing). They argue that value stocks are subject to higher information uncertainty (their proxies: low size, age, analyst coverage, turnover; high dispersion in analyst forecasts, returns volatility, cash flow volatility). This is because they usually draw less media attention, and weaker demand for information leads to higher uncertainty. Upon earnings announcements, two forces come into play: (1) delayed reaction effect from processing information on stocks with higher information uncertainty, and (2) risk premium effect to compensate for the uncertainty itself. For good news, value stocks experience both effects and thus should experience greater earnings momentum than growth stocks. A similar result is documented by La Porta et al. (2012), but as often seen in asset pricing research, interpretations can differ.

Now, let’s take a look at how the classical value strategies fare in Thailand. As mentioned earlier, signal generations involve a lot of practical decisions and many versions of the same signal exist. The factors with q suffix are formed using quarterly financial statements, and the number suffix represents holding periods. A strategy with 6 at the end involves initiating a portfolio at the end of each month (so there are portfolio vintages), to be held for 6 months each. As the first portfolio matures, it’s time to start another, and at any given point in time, there are 6 outstanding portfolios. I test 11 value strategies over 36 definitions between January 2000 and May 2021. Here’s what I find:

• 10 of out 11 strategies work. The only strategy that doesn’t work is the BM strategy that uses end-of-December book value to end-of-June market cap (often referred to as the HML Devil, from the title of the paper – The devil’s in the HML details). This is rather surprising because in Asness and Frazzini (2013), it works better (at least in the US, where researchers find that the value premium has vanished in latter years). The argument is that traditional BM that uses end-of-December market cap ignores recent price movements. I’m not sure what’s causing this difference, but it might be something related to what high BM ratio captures in Thai setting (distressed firms? firms under the radar?).
• Ratios from quarterly financial statements work better. Many value factors perform better with signals generated from quarterly financial statements. This may have something to do earnings announcement effect that takes shape via earnings momentum. I’ll be taking a look at this issue in more details in subsequent work.
• Many value strategies are highly correlated. For example, the correlation between EP, CFP, OCP, EM, DP are at least 0.49. This is perhaps not that surprising, because the numerators and denominators are very similar (cash flow = profit – investment, and dividend is one form of cash flow). The most uncorrelated strategy is BM, which is the classical value factor and remains an active topic of research until today.

Just like momentum, value is everywhere, and there are many ways of defining it. In closing, I’d like to remind the reader that the discussion of value investing here is limited to factor investing applications only, where we pick characteristics, not stocks. There are a lot of nuances and subtleties that simplistic quantitative analyses of structured data such as accounting numbers cannot pick up. This is particularly relevant in modern setting where today’s accounting numbers may not fully reflect the reality of the business (think IFRS 16 and recent issues related to accounting of supply chain financing), as well as intangible assets (typically not on balance sheet) which has played an increasingly important role in the economy. Research on intangible assets and the implications on asset pricing is still nascent but has recently received greater attention in light of tech stocks’ performance. I’ll dedicate a post to talk about this at some point in the future.

References

• Banz, R. W. (1981). The relationship between return and market value of common stocks. Journal of Financial Economics, 9(1), 3-18.
• Basu, S. (1983). The relationship between earnings’ yield, market value and return for NYSE common stocks: Further evidence. Journal of Financial Economics, 12(1), 129-156.
• Rosenberg, B., Reid, K., & Lanstein, R. (1985). Journal of Portfolio Management, 11, 9-16.
• Lakonishok, J., Shleifer, A., & Vishny, R. W. (1994). Contrarian investment, extrapolation, and risk. The Journal of Finance, 49(5), 1541-1578.
• Desai, H., Rajgopal, S., & Venkatachalam, M. (2004). Value‐glamour and accruals mispricing: One anomaly or two?. The Accounting Review, 79(2), 355-385.
• Barbee Jr, W. C., Mukherji, S., & Raines, G. A. (1996). Do sales–price and debt–equity explain stock returns better than book–market and firm size?. Financial Analysts Journal, 52(2), 56-60.
• Fama, E. F., & Kenneth, R. (1993). French, 1993, Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33(1), 3-56.
• Fama, E. F., & French, K. R. (2018). Choosing factors. Journal of Financial Economics, 128(2), 234-252.
• McLean, R. D., & Pontiff, J. (2016). Does academic research destroy stock return predictability?. The Journal of Finance, 71(1), 5-32.
• Grossman, S. J., & Stiglitz, J. E. (1980). On the impossibility of informationally efficient markets. The American Economic Review, 70(3), 393-408.
• Chan, L. K., Jegadeesh, N., & Lakonishok, J. (1995). Evaluating the performance of value versus glamour stocks The impact of selection bias. Journal of financial Economics, 38(3), 269-296.
• Doukas, J. A., Kim, C., & Pantzalis, C. (2004). Divergent opinions and the performance of value stocks. Financial Analysts Journal, 60(6), 55-64.
• Yan, Z., & Zhao, Y. (2011). When two anomalies meet: The post–earnings announcement drift and the value–glamour anomaly. Financial Analysts Journal, 67(6), 46-60.
• Porta, Rafael La, Josef Lakonishok, Andrei Shleifer, and Robert Vishny. “Good news for value stocks: Further evidence on market efficiency.” the Journal of Finance 52, no. 2 (1997): 859-874.
• Asness, C., & Frazzini, A. (2013). The devil in HML’s details. The Journal of Portfolio Management, 39(4), 49-68.