Darkness in the Valley of Value

By
Posted on June 4th, 2018

Recently, I had the pleasure of hearing Kenneth French discuss creating the framework for evaluating factors.  A factor is simply a characteristic that explains some of the variation in returns of an investment.  These factors can be used to build a portfolio that has higher expected risk-adjusted returns.

In the 1950’s, academia began getting involved in finance, and Harry Markowitz ushered in the first model[i] used to evaluate portfolios. Rather than using a collection of individual investments, Mr. Markowitz’s model had one factor, the riskiness of the portfolio compared to the model, which is Beta.  For example, an all stock portfolio, with 480 companies in the S&P 500, likely has a Beta of approximately 1.  In short, if someone wants a return that matches the market; they will have to take the same amount of risk.  There are no free lunches!

Mr. French and Eugene Fama coauthored the seminal paper[ii] that moved the investment world beyond using only Beta as the factor to explain equity returns. They came up with two additional factors: Company size and “value” characteristics. This became known as the “3 Factor Model”.  These factors impact Beacon Hill’s decision to tilt towards Value companies and to include smaller capitalization stocks.

While these factors help explain the variation in returns, Fama and French later added two additional factors[iii] to the 3-factor model; Profitability and Investment.  They creatively called this the “5 Factor Model”.

5 Factors

  • Market Risk – a measure of volatility of a stock in comparison to the market
  • Size– smaller companies
  • Value– high book value/market value
  • Profitability- returns of firms with high operating profit
  • Investment– defined as increase in assets

Do these factors persist?  Why?

The most persistent factor is the Value factor, which has widely underperformed the recent years.  Why would it be persistent?  Most point to the basic thought that companies identified as Value companies aren’t the type that generally are much fun to talk about- no fun at a cocktail party!  We have a economic model for why the factor persists.  Without this justification, there may be quirks that show statistical significance but have no validity.

Race for other factors

We have traditionally tilted our portfolios to include a bit more value and smaller capitalization stocks. Should we use additional factors? Why not 10… or 100? There has been an explosion of factors that product providers are racing to create and profit from.  Some may be valid, and some may not.  As they say, “if you torture data long enough, it’ll say what you want it to”.   For example the “Super Bowl Indicator” shows that the S&P 500’s performance for the next year will be positive if a team from the AFC wins.  It’s been accurate 80% of the time.  But we all know that’s just a coincidence and it would be foolish to invest based on that!

We’re frequently pitched by investment wholesalers on the latest of hundreds of factors that emerge.  French’s discussion focused on a framework for analyzing if factors will persist.  A key point from this discussion, was that data- without a compelling model- is simply noise. This overlaps with our skepticism in using additional factors in our analysis. There is another factor that has drawn other investors and our attention in recent years – Momentum.

Momentum

The factor that we’ve been researching the most of late is Momentum.  In short, stocks that are going up tend to continue in that direction for a period, and vice versa.  We measure the significance by the T-stat; a basic the rule of thumb is anything over 2 indicates some significance.  From 1927-2016, a t-stat of 5.5 shows a high level of significance that momentum helps explain variation in returns.  However, when we’ve analyzed products attempting to capture this, they have all been underwhelming.  Which has puzzled us.

Mr. French posits that while Momentum does statistically explain returns in the data set from 1927-2017 (t-stat of 5); he does not believe that there is a “Model” for why momentum should exist. I asked him if there wasn’t some model he wasn’t aware of.  For example, investors generally invest in the mutual funds that had the best performance last year.  As prior year performance is a contra indicator to future performance they, as a whole, significantly underperform the aggregate returns of mutual funds.  Therefore, they are losing return due to a momentum effect.  He, however, didn’t buy it.

While we may disagree on that, his data did answer our overarching question of why we can’t see positive performance when we review products.  In short, while the Momentum factor is explanatory for the time period 1927-2016; it has no explanatory power for the subset of time from 2002-2016,  which is when these products have been created.

So, is it a viable factor that has just been out of favor?

Bayes’ Theorem: How to analyze New Factors

Bayes’ theorem gives us a way to quantify a hypothesis (in this case that a factor is statistically significant), based on prior assumption of its probability and new evidence.

In this case, Momentum appears statistically significant- but with no “Model” (which I still disagree with).

However, more important to our current portfolios, if “Value” didn’t bring us risk-adjusted returns for the last few years, it will take 400 years until our models tell us it’s not a factor worth investing in.  In fact, it will take 400 years!  We genuinely hope we don’t have to wait that long!!

 

 

[i] Markowitz, H.M. (March 1952). “Portfolio Selection”. The Journal of Finance. 7 (1): 77–91.

[ii] Fama, Eugene F.; French, Kenneth R. (1993). “Common Risk Factors in the Returns on Stocks and Bonds”. Journal of Financial Economics. 33 (1): 3–56.

[iii] Fama, E. F.; French, K. R. (2015). “A Five-Factor Asset Pricing Model”. Journal of Financial Economics. 116: 1–22.