Finance & Investing

Do Portfolio Managers Underestimate Risk by Overanalyzing Data?

New research questions whether “smart” beta is always smart.

July 25, 2016

| by Louise Lee


Illustration of a man moving a life-sized rook on a chess board | iStock/JDawnInk

Some potential investment catastrophes can be missed by “smart beta” analysts, no matter how deep they dive. | iStock/JDawnInk

While new technologies and globalization have brought many changes to the practice of money management and investing, the one thing that has remained constant for most investors is the desire to beat the market. Recently, many portfolio managers have seen smart beta strategies as easy ways to do better than average.

Smart beta strategies are based on calculating some attribute of each stock that is fairly simple to measure, such as the ratio of the company’s book value to its market value. The investment portfolio is then tilted toward stocks with high measures of this smart beta and away from stocks with low measures. The hope is that this tilt increases the portfolio’s return without increasing the risk.

In recently published research, Paul Pfleiderer, the C.O.G. Miller Distinguished Professor of Finance at Stanford GSB, argues that smart beta strategies may not be the free lunch investors are seeking. These strategies may expose the portfolio to risk the investor isn’t aware of, says Pfleiderer, whose research was written with Terry Marsh of the University of California, Berkeley.

The problem lies in the way smart beta strategies are identified. In most cases the strategies are based on statistical discoveries. Researchers fish through ever-growing streams of data seeking common traits of companies whose stocks outperformed the market in the past on a risk-adjusted basis.

Pfleiderer claims it is important to attempt to identify the reason these stocks historically have had higher returns. Were investors systematically making mistakes in valuation that were related to the putative smart beta? Were there institutional rigidities that created these higher returns, ones that can be exploited by nimble investors? In these cases and under the assumption that the mistakes and rigidities persist into the future, smart beta is indeed smart.

Of course, it is always possible that the historical findings are spurious. If one fishes long enough in a sea of data, one is virtually guaranteed of finding misleading patterns and relationships.


If one fishes long enough in a sea of data, one is virtually guaranteed of finding misleading patterns.

The main worry that one should have in tilting toward a smart beta is that the higher returns are actually earned as compensation for hidden risk. The downsides of some risks can be infrequent but catastrophic and may not show up in historical returns, says Pfleiderer. A good example is what happened in the last financial crisis, when housing prices declined throughout the United States — something that many did not consider a possible outcome since it had never happened in the several decades of historical data they used to identify risks.

In their analysis Marsh and Pfleiderer show that the consequences of ignoring hidden risks can be quite severe. They are generally much greater than what is lost when one erroneously assumes smart betas are due to risk and therefore fails to fully exploit the free lunch.

Pfleiderer acknowledges that it’s difficult for investors to uncover all the important economic factors behind stocks’ performance and that it’s tempting to make investment decisions based purely on statistical patterns, especially ones that appear robust. If you do that, “be aware that you might be taking on some added risk,” he says.

For media inquiries, visit the Newsroom.

Explore More