In Part 1 of this three part series, I introduced the concept of factors and discussed how they may be used to help generate higher risk-adjusted returns. Since the early 90’s and the growth in popularity of the Fama and French 3-Factor Model, there have been hundreds of additional factors that have been ‘discovered’ by researchers. The process of designing and testing the model to prove a factor is real requires careful statistical analysis. Researchers review other’s papers and often find that the reported significance is incorrect or that there was some flaw in the data, the model definition or both.

According to Research Affiliates, in a post from January, there are hundreds of factors, but according to their research they identify six that provide an opportunity to outperform the market. The list of six is not definitive. Other firms and college professors would debate the list. Still, this is as good a list to start from as any I have seen.

What Makes a Good Factor?

I have read dozens of papers on factors. I ran my own statistical analysis and simulations and have done some experimental implementations with trading factor strategies. Without getting into the weeds on analysis, let me highlight a few of the key things I look for.

First, there must be some reasonable economic or behavioral explanation behind the factor. According to the Super Bowl Indicator, if an American Football Conference (AFC) team wins the market is believed to be heading down. I haven’t studied the data, but this one gives a gross example of a factor that I wouldn’t even consider. What economic rationale could account for this? Statisticians call this kind of relationship spurious correlation. If you look long and hard enough for some variable that looks to be linked to another you may just find them. Without a real link between them, they cannot be expected to hold in the future.

Implementation is a major question. Often times the research papers do not account for trading costs or slippage. Trading costs is obvious. If it costs $7 for a trade, that has to be deducted from the profits. I can give a concrete example of slippage. Let’s say my algorithm says ‘buy’ based on the closing priced on day 1. On day 2 the market opens and I trade at the market. The price my order trades at will not be the same as the price I wanted to buy at. The difference in price is part of the slippage between the strategy and the implementation. Based on my analysis and limited testing, I estimate slippage, even on broadly traded ETFs with monthly turnover, could be as high as 30 to 40 bps (basis points). (A basis point is one hundredth of 1%, so 30 bps is .003%.) If a factor strategy has less than 3-4% alpha on testing, I would probably not even attempt to execute it.

Finally, taxes have to be considered. If a trading strategy generates short term gains and we are in a taxable account, then excess profits can be eaten by the additional taxes that would have to be paid compared to a purely passive index approach. For this reason, active strategies may be more appropriately placed in taxable accounts.

Six Factors

As noted earlier, for a discussion on common factors let’s start with the list from Research Associates. I’d like to briefly describe the factor, the economic rationale and offer my opinions given my own research.

Value

The value factor derives from the idea that a stock has some intrinsic value, but is trading at a price below this level. Fama and French used the market value of the stock and the book value of the company for their definition of value. There are additional definitions as well. One of my favorite factors for a value proxy is cash flow yield, the operating cash flow plus investing cash flow (a negative number) divided by the total shares outstanding. One behavioral explanation for value providing excess return is that less glamourous companies can get overlooked and the price can get depressed. Caution is warranted. If the value of a company is based on earnings and earnings growth, then the metric used for value may not capture the intrinsic value correctly. It is common to find stocks in the value segment that are financially distressed and/or are experiencing low growth. When markets start to crash, value stocks often crash first. This seems counter-intuitive, but the idea is that risky, distressed or financially troubled companies are not ‘safe’ and when the market goes bad there is a ‘flight to quality’.

Size

The Fama and French 3-Factor model included the small capitalization factor (SMB). This factor is also called the size effect. Small companies tend to outperform the larger companies. There is quite a bit of debate on the economic rationale for why this should be the case. Some argue that the size factor is hiding some unquantified risks. For example, if the stock is thinly traded with poor liquidity this could explain why size works. According to Research Affiliates, the size factor has the greatest value-add with +3.19% for a long-only portfolio. Size return could be explained by investors under estimating the growth potential for smaller companies.

Momentum

Subsequent to Fama and French publishing the 3-Factor Model, Narasimhan Jegadeesh and Sheridan Titman published a paper that showed that buying stocks with the highest performance and selling the stocks with the worst performance generated excess returns. In 1997, Mark Carhart created the 4-Factor Model. It extended the 3-Factor model and added a factor for momentum called MOM. MOM is the return of a stock over the last year excluding the last months return. The rationale for this factor is thought to be behavioral. As stocks appreciate, the investor enthusiasm grows and additional investors drive the price up above the intrinsic value. Eventually the stocks in this group will decline. This process is gradual, but research has shown than within 2-3 years the excess return tends to reverse.

Because this factor is grounded in irrational human behavior it seems to be one of the most persistent factors. As the stock market rose to its highs in 2000, I couldn’t believe how far the PE ratio’s rose. Based on discounting future earnings and even after accounting for low interest rates, to me the market was overvalued by > 50% and it persisted for years. It was at this point where I begrudgingly really started to let go of this idea that markets are entirely efficient.

Low Beta

Low Beta is a very interesting factor. Andrea Frazzini and Lasse Pederson wrote a paper in 2013 that showed that buying stocks with low beta outperforms stocks with high beta. Beta is the risk of the stock measured against the market. The economic rational is interesting. The idea is the investors ‘reach’ for return to the point where the highest risk stocks produce less return for each unit of risk. One way I like to think about this is to understand the difference between average return and compounded return. A stock with a high average monthly return that has high risk can end up producing a low return over many months (the compounded return). Investors see a good month or two of performance, pile into the stock only be exposed to excessive risk.

Profitability (Quality)

Profitability can be measured in a number of ways. Operating profit and return on equity are common measures. More broadly, quality factors speak to the quality of a company’s earnings. Many factors have been developed and implemented. For example, Brightwood Ventures uses an iShares ETF of Quality stocks (symbol: QUAL) that invests in companies with high return on equity and consistent earnings (low earnings variability). Other quality scores look at the financial stress of the company. One of my favorite researchers on quality is Robert Novy-Marx. One of his key quality measures is gross profitability. He argues that high gross profitability is a strong indicator for future earnings – a key element to valuation using discounted cash flows. He argues that high earnings are indicative of a company that has strong market power. These are companies with high market share and exceptional value in their products. His premise is that these companies tend to have more persistent earnings and less competitive pressure. I’ll talk more about Quality in Part 3. My research shows that combining quality with a value factor based on cash flows generates significant returns and benefits from low correlation between value and quality.
For a more detailed discussion of quality, including a review of measures for quality, see Quality Investing by Robert Novy-Marx.

Investment

Economic valuation models imply that higher investment implies lower future expected return. A behavior argument for lower future returns could also be that managers are investing in company projects that generate returns below the required rates. This could be the case if managers are seeking to expand their power and not optimizing shareholder value. Fama and French are now in the process of creating a 5-Factor model that includes operating profitability and investment factors to explain stock returns.

Conclusion

We have covered six common factors used to explain stock market performance. These factors are based on fundamentals of underlying companies. Factor models have been developed for other assets and the factors may include macro factors as well (e.g. interest rates, GDP, investor sentiment). The sheer number of factors and analysis required to determine if a factor has value can be very challenging. Gauging the factor should consider the economic or behavioral rationale and the true cost to implement the strategy such as trading costs and slippage. After tax returns should be absolute measure for any portfolio strategy that incorporates factor strategies.

Investment in factor based strategies using ETFs and smart beta are experiencing high inflows in capital. In the final Part 3 article for this series, I’ll discuss how smart beta funds can be used in investor portfolios, including some of the key issues and risk.

——–

Jegadeesh, N.; Titman, S. (1993). “Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency”. The Journal of Finance. 48: 65. .

Carhart, M. M. (1997). “On Persistence in Mutual Fund Performance”. The Journal of Finance. 52: 57–82.