Robo advisors have raised hundreds of millions of dollars in funding and are getting incredible media attention. They have built fantastic, highly usable interfaces and will no doubt change the advice industry (in fact, they already have).

However, when asked about the potential problems with online portfolio management platforms in Scottrade's 2014 advisor survey, advisors focused on the mechanized method of client interaction as the chink in the robo armor. Eighty percent of advisors say the robos’ biggest weakness is “lack of human interaction” and 46% say that “lack of service” plays a role.

But “lack of service” is essentially the same human interaction element; therefore virtually all RIAs focus on the lack of human interaction as the main weakness. They are probably right, and a number of experts, including Michael Kitces and Joe Duran have predicted that robo technology will work best with advisors not against them, for that very reason.

However, those RIAs surveyed by Scottrade failed to mention a very important point. Almost all robo advisors were founded in 2009 or later and they have never seen a meaningful downturn in the financial markets. This is a major issue, for the following reason: Mean-variance methodologies are highly susceptible to market shocks.


Robo advisor methodologies are based on mean-variance optimization, which focuses on trailing variance as a measure of risk. Trailing variance is calculated from relatively recent history of the financial markets, usually a few years. Variance in financial models is simply an average squared difference between the average return and return on every date observed over some historic period of time.

Thus, it should be clear that when the market is trending up, as it has over the past few years, the average return is positive and deviations from that average are relatively small. That is why in periods of prolonged market tranquility, trailing risk measures have dramatically understated market risk.

Having worked in risk management prior to the 2008 crisis, I can still remember how all variance-based measures such as standard deviation, tracking error and Value-at-Risk showed historically low levels of risk leading up to the 2008 crash. It turned out that only the realized volatility was historically low, but the level of risk was quite high. 

This historically low level of risk predictions coming from variance-based measures right before the Lehman collapse prompted much dismay, including the insightful story, "Risk Mismanagement" by Joe Nocera in The New York Times magazine on January 2, 2009. Consider the following chart that shows a 5-year trailing standard deviation of the S&P 500, a measure highly related to the type of variance used in mean variance optimization.

Note that risk, according to this measure, is very low just prior to the 2008 crash, it increases dramatically toward the end of 2008, and peaks toward the end of 2011. The reason for this is that years that precede market crises have been unusually tranquil. This makes portfolios that are managed with variance-based risk measures susceptible to a dramatic reversal.


Furthermore, there is a rise in correlation, which exacerbates the situation by reducing whatever diversification is built into the portfolio. These are reasons why the Federal Reserve all but stopped mentioning Value-at-Risk (a variance related measure) in their risk assessments and now talks about risk in terms of stress tests.

A great economist, Hyman Minsky, explained in his excellent paper, The Financial Instability Hypothesis, that risk arises as a result of overleveraging in good times. Since realized variance is low during good times, this finding assures that variance based measures of risk will tend to suggest to buy high and sell low around crisis events.

In addition to the issue of understatement of risk prior to a crisis, variance-based measures overstate risk following it. You can see that risk in our chart keeps climbing all through 2009 and in fact keeps climbing through 2011. As a result,  investors that go to robos after the crisis will get portfolios that are now excessively light on stocks.

The current wave of robos have great management teams and, in all likelihood, will eventually address these concerns. In the meantime, however, these points represent a huge opportunity for advisors who are prepared to add value by creating a thorough risk management plan for their clients’ wealth.

Daniel Satchkov, CFA, is president of RiXtrema a risk modeling and consulting firm which helps advisors link portfolio crash-testing directly to financial planning. 

Read more: