Investors Can’t Always Trust the Data, and That’s OK

I was attracted to finance because it promised some order amid chaos. Here was this market, with billions of transactions a day — and yet it managed to set a price for each asset, a price that put a literal number on the value of future risks, or more precisely how much people value those risks in the present day. The world is inundated with information — about individual companies, about the macro economy, about geopolitical risk, about (not to get too meta) prices themselves — and this price incorporated all that, almost instantaneously. This is the definition of market efficiency.

Except for one small thing: This number, this price, has always been a little wrong. Data, as it turns out, has issues.

A roiling controversy in finance is a reminder that any certainty anyone ever had was an illusion. It concerns an academic paper that questions the benefits of factor investing, in which investors make decisions based on “factors” such as a company’s size or how its share price compares to the value of its assets. The theory is that such investments can deliver better returns than the market as a whole.

The paper argues that the data collected and made public by the fathers of factor investing, Kenneth French and Eugene Fama, changed over time — and when the numbers changed, so did the estimates of the factors and their value to your portfolio. True, both the new and old data suggest there are benefits to factor investing, but how much depends on which data are used.

Data Changes Expected Return

This is not just an academic argument. The factor model is taught in business schools and often used to assess market performance and the price of capital. Fama and French are also affiliated with Dimensional Fund Advisors (DFA), a mutual fund company which offers funds that over-weight the factors. DFA staff assist with the Fama/French data in a non-transparent way to outsiders.

Full disclosure: I worked at DFA more than 10 years ago with Fama and French on another, unrelated data project. One lesson that has stuck with me is that all financial data, no matter the source, are very noisy. And by noisy, I mean unreliable. Most estimates made from financial data are extremely sensitive to the time frame that is selected and any assumptions that are made (and assumptions must always be made). No one should ever take an estimate of a financial variable as an actual fact.