Factors: Not Driving, Just Along for the Ride

“There are more things in heaven and earth, Horatio, than are dreamt of in
your philosophy.”

— William Shakespeare, Hamlet, Act 1, Scene 5

In our white paper “The Limits of Theory” we wrote that “company stock price movements drive factor returns; factor returns don’t drive company stock price movements.” Events since the start of the year have brought this statement to life all too vividly, and merit spending a few moments illustrating how the factor-driven view of the world does not provide meaningful insight into what has been happening.

Consider the data in Table 1. Here we list 10 securities and show two pieces of data for each: its dividend yield at year end, and its return from 12/31/19 through 3/20/20. Many investors view dividend yield as a kind of valuation metric. The securities are listed from lowest dividend yield (i.e., “most expensive”) to highest (“least expensive”).

Figure 1 shows what this data looks like visually, with dividend yield on the horizontal axis and the subsequent returns on the vertical axis.

The chart below includes a trendline that best fits the data and the slope of that line indicates that higher dividend yield was associated with lower returns from 12/31 through 3/20. Think of the chart as a simple, single-factor analysis. The results are clear: for whatever reason, higher dividend yield was associated with poorer returns. We would describe this situation by saying that “the dividend yield factor had a negative return.” Why might that be the case? The model is silent on this question. Factor models don’t really try to explain why a factor has a positive or negative return. Rather, they proceed on the assumption that we can identify what factors investors care about, and then simply measure what the returns to those factors are over any given period. Real world factor models, of course, have many more factors than just dividend yield; they include things like profitability, size, financial leverage, earnings quality and more. But the point is still the same: models assume that whatever factors they include are the things investors “price.”

As the quote from Hamlet at the beginning of this note points out, the world is a far more complicated place than any factor model can possibly capture. Consider what has been happening since the start of this year. This is a presidential election year in the U.S. and for a while in February it looked like a socialist was going to sweep to victory in the Democratic party primaries. Just as that risk began to recede, a new virus, which many thought had been contained in China, began spreading uncontrollably to other countries around the world. As it became clear that the virus was putting global growth at risk, oil producing countries tried to negotiate reductions in supply to prevent the drop in demand from pushing prices lower, only to have a price war break out between Russia and Saudi Arabia as each tried to demonstrate its dominance. The price of oil fell by 50% in just a few days, just as major developed economies began telling non-essential workers to stay home for an as-yet undetermined amount of time to try to stop the spread of the coronavirus. Empty shelves became a common sight in grocery stores as consumers began stocking up for an extended stay at home. Long-term bond yields fell to record low levels in the U.S., with the 10-year Treasury bond at one point yielding just 0.31%. Stock markets lost roughly a third of their value in just a few weeks—a record-setting (for speed) bear market.

With that remarkable set of events as context, consider Table 2. The data is the same as in Table 1, but we have made two changes: we filled in the names of the securities, which are in fact the sector indices within the MSCI World Index, and we sorted them not by dividend yield but by their return.

Now, given what we all know about the real-world events laid out above, the returns in this table make a lot of sense. It is hardly surprising that a 50% drop in the price of oil almost overnight amid an accelerating slowdown in growth would push energy stocks down 55%. Or that financial stocks would suffer from a record-setting low in the U.S. 10-year yield. Or that consumer staples stocks would perform the best on a relative basis in an environment where growth seems to be heading over a cliff but consumers are stocking up on food. Or that health care and technology stocks would hold up better than industrials and materials stocks.

The point is, there are obvious reasons why the sector returns line up the way they do in Table 2 and those reasons have nothing to do with the dividend yield of one sector versus another. The relationship depicted in Figure 1 is a great illustration of the notion that “correlation does not equal causation.” Investors were not pricing factors in recent weeks; they were pricing significant events taking place in the world.

In closing, we will simply quote again from “The Limits of Theory:”

[M]arkets don’t reward or punish abstract factors; they reward or punish companies because of how well or poorly their business is doing, and that in turn creates what we end up measuring as “factor returns.” But we should never lose sight of the fact that those factor returns are a derivative. They are not the starting point.

In understanding the performance of any investment strategy over recent weeks, it is important to pay attention to how real economic events drove that performance, rather than fall back on a set of abstract factor returns as if they were somehow responsible.