In the previous 2 posts, we looked at clues from super-investors and index investing for how we will build our successful investing process. Next we turn to another market model for clues to our investing strategy. The Fama and French model is a simple model to describe returns of assets in markets. It is a great model in that it explains a lot of the returns of stocks and bonds, for example. And, it also allows us to understand the market in such a way as to load up on the types of securities that generate higher returns. But, like all models, it is a simplified version of reality that does not describe the entire picture. So, keep in mind that virtually all models fall short of reality, but they can still be useful in understanding how systems work.
The Fama and French model was originally introduced with 3 factors: beta, value, and size. The model explained returns of stocks, bonds, and other securities in terms of these factors. Let’s talk about value and size first.
In the Fama and French 3 factor model, value is a factor that indicates that the security is cheaper than it should be. In the early days, price to book was used. So if you construct a theoretical portfolio where you buy low price to book stocks (value) and short high price to book stocks (overpriced), you observe excess returns in this portfolio if it is composed of many securities and averaged over many years of investment to smooth out the random fluctuations that happen in all investments. Similarly, an excess return was observed when designing a theoretical portfolio that loads up on small companies and shorts big companies.
The explanations for these effects are numerous. One example is that behaviorally, people prefer to invest in large, well-known companies while also preferring exciting, high growth companies to boring value companies in their attempts to get rich faster, so those kinds of stocks are overpriced relative to the converse.
Beta is a measure of market risk in this model. Beta measures how correlated a security is to the overall market (the market has a beta of 1). Something with a positive beta is correlated to the market. A negative beta means that the security or basket of securities tends to move in the opposite direction of the market (anti-correlated). While a beta near 0 means that the security tends to be uncorrelated to the overall market. Betas that are large in magnitude (greater than 1 or less than -1) mean that the security has big swings when the market swings, while betas that are smaller in magnitude (between -1 and 1) mean that the security has swings that tend to be smaller when the market moves in one direction or the other. In this model, the larger the beta, the more risk you take on, and the more return that you earn.
It was also noticed a bit after the introduction of this model that stocks which have moved up for a certain period of time tend to continue moving up for a similar length of time. A new factor called momentum was introduced. Theoretically, if a long portfolio of stocks that have moved up for, say 3 months is bought for 3 months, and at the same time a portfolio of stocks that have moved down for 3 months is shorted for 3 months, then one will see an excess return over time. This factor has been changing since it was discovered and early funds sought to exploit this momentum factor. There is debate about whether it will continue to exist, since it is now known and people are seeking to exploit it. It may not be as powerful as it once was and could change at any time. Adding to this complication, momentum does reverse unpredictably, and often leads to big drawdowns.
Further, it was discovered that there is actually a beta anomaly. It seems that there are assets (like stocks) where one can earn an excess return despite them being lower in beta factor than the overall market. A new factor was introduced to explain this called the quality factor. If one constructs a theoretical portfolio of high quality companies and shorts a group of low quality companies, then one also finds an excess return.
So, the Fama and French 5 factor model emerged from the 3 factor model, and explains much of the returns in baskets of assets (large averaging over many fluctuating things) held over long periods of time (large time averaging of fluctuations). To summarize, the 5 factors are: Beta (market risk or non-diversifiable risk), value (aka value minus growth), size (aka small minus big), momentum, and quality. This model is great at eliciting some explanations for returns in aggregate. But like all models that imperfectly explain complex systems, it has some shortcomings, which we should discuss.
First off, beta and its cousin volatility are not risk. True risk is related to 2 ideas: 1) that you will not see an adequate return on your capital, and 2) that your capital will not be returned to you. These are 2 sides of the same coin – value destruction. Specifically, it’s the value that you have stored up in your portfolio through many years of saving and hard work. It is actually quite hard to find an easy to use, objective measure for risk of value destruction. You often don’t realize the risk you were taking until the value destruction happens. So, to have a proxy for risk we often pick beta or volatility instead. If your mentality is short term trading, risk and volatility do converge because the market movements are more random on shorter timescales. However, on long timescales, volatility is less important as the forces of compounding and growing earnings exert their influence.
Second, the model does not say much about individual companies, only baskets of them. We can dig into this a little deeper, by, for example, looking at the value factor. The Fama and French model must handle a lot of data over many years, so model researchers need an easy handle on value. There are a number of such measures, such as price to earnings, price to free cash flow, and price to book value. Fama and French chose price to book. All of these valuation models fall short, however. In addition to picking low price to book companies that are true value gems, for example due temporary bad news, there will also be value traps that are chosen in the data. Value traps are companies that look like they are trading at deal prices when using a simple measure, like price to book. However, on further analysis of the company itself, one can often easily see that the company is trading at a lower valuation because of serious problems with the company from which it will likely never recover. This is a major problem with any single stock market measure. In isolation, it never tells the whole story and could be good or bad. But with these models, it becomes untenable to analyze tens of thousands of companies individually over 100 or more years to get meaningful data, so they take a shortcut to get some analysis done.
You also see this when index fund fanatics discuss how dividend paying companies are no better than non-dividend paying companies or companies that buy back shares are no better than companies that eschew such practices. On a level, they are right, because they are lumping all companies together without differentiating. On another level, they are seriously wrong, because dividends can be a powerful signal of quality and needing to pay dividends can prevent waste of money within the company on non-productive activities which won’t yield a better return than the investor could receive elsewhere. Share buybacks can be meaningless if a company does it indiscriminately whether the company is overvalued or not, but can be seriously accretive to shareholder value if disciplined buybacks occur only when the company is undervalued. These fanatics would counter that you can’t know when a company is really undervalued or when a dividend is really a signal of quality. However, there are some serious statistical outliers that prove them wrong. Warren Buffet, Walter Schloss, Peter Lynch, and those from the Benjamin Graham value investing school of thought are 10-20 sigma statistical outliers on this front. The odds of a cluster of such statistical outliers grouping together under one school of investing is even more improbable. I would personally say that with a bit of financial literacy, you can absolutely trounce index investing over long periods of time. But like anything, it involves work, diligent study, and investing discipline over many years to do so.
Finally, there are a few quirks to these models. There appear to be interactions between the factors. For example large cap growth companies do worse than the model predicts in aggregate, and small cap value companies do much better. Low beta stocks, which by the way, often load higher on quality, do better than predicted compared to extremely high beta stocks. Another quirk is that loading on one factor is not usually possible. For example, loading on the value factor will typically load on negative momentum factor, since value stocks are often in that range because they have decreased in price over the last time period. Quality, size, and beta co-vary, also. So, these factors cannot strictly be isolated. It’s not that we can’t handle this situation statistically, but when purchasing investments in the real world we cannot construct a portfolio of pure value and 0 loading on momentum very easily.
While the Fama and French model falls short of analyzing and explaining individual companies, it does provide us with some meaningful clues about what might be important to investing. I want to focus on the value and quality factors in particular, because they suggest you can earn outsized returns if you can assess value and quality in a stock. We are not so much concerned with beta specifically, since there are anomalous returns in low beta stocks, and beta isn’t really risk anyway for a very long term investor. Low beta stocks load higher on quality anyway, so we want to focus on the quality factor. The size effect has diminished in recent years, and really shows itself more in the interaction with value anyway (small value shows a stronger positive return effect, and large growth shows a stronger negative return effect than the sum of value + size would suggest). It is also speculated that some of this is due to small companies having larger spreads, poorer reporting, more fraud, and other factors we cannot really counter as individual investors. In addition, more recently, index investing in small companies using, for example the Russell 2000, causes a lot of money to flow into small cap stocks with limited capacity for such flows of capital, perhaps leading to an overpricing of the small factor more recently. Momentum is unpredictable and can reverse at any time. It can work for a while and then suddenly stop working as many momentum funds found when they lost massive amounts of money in the years during and after the great financial crisis in 2008.
When we combine this analysis with our analysis of index investing and past successful super-investors, we see that value and quality seem to be the common factors that are emerging as important when picking stocks. We want to pick quality and value stocks that we can buy and hold for a long time or even forever with extremely low turnover. In the next post, we will put all of the ideas together from successful super-investors, index funds, and the Fama and French 5 factor model to discuss the methodology for our investing strategy.
Leave a Reply