Blog Posts — July 6, 2020

Tracking error of Russell 2000 vs. Russell 3000 soars to 25-year high…and yours could too

Equity markets have mostly recouped the losses of the downturn that started in February of this year, but at different rates. Notably, the broad market Russell 3000 index ended the first half down just 3.5%, whereas the small-cap Russell 2000—unable to benefit from the strength in such names as Amazon, Apple, Microsoft, Tesla and others that drove the Russell 3000—is down more than 13% year to date.

Large-Cap Strength Drove Small-Cap Weight Down…
As the high-profile technology-related names increased in Russell 3000 weight, the Russell 2000 fell from 7.3% of the broad index to 6.3%. At the same time, the risk characteristics of many of the biggest names also changed. The 10 largest names in the Russell 3000 currently account for 22.7% of the weight, but just 17.6% of the risk, whereas the same cohort was less than 19% of the weight and 20.7% of the risk of the benchmark back in December. Most contributed more to benchmark risk than would be expected given their weight six months ago, while the opposite is true currently. And the risk profile of Info Tech as a sector changed; it now looks more like a traditional “staples” sector (see, for example, “The shifting sands of sector risk…When low-volatility sectors become high volatility – and vice-versa”.)

…and Predicted Tracking Error Up
As many of the biggest names became less risky than their weights would imply, the tracking error of the Russell 2000 to the 3000 increased even more dramatically—to at least a 25-year high of over 18% (Exhibit 1). The last time the tracking error was even close to this level was at the height of the internet bubble in 2000. It also exceeded its usual range of 5% to 10% (with an average of about 7.7%) during the global financial crisis. For the past seven or eight years the tracking error has been below average, ranging from around 5%-7%. So, this sudden increase looks quite dramatic.

Exhibit 1. Tracking Error*, Russell 2000 vs. Russell 3000 and Risk Breakdown

Source: FTSE Russell, Qontigo

*US4 medium-horizon fundamental model  

We dug a bit to try to identify the drivers of the big increase in active risk from a factor perspective and identified at least two major drivers: factor exposures and factor covariances.

Many Active Factor Exposures Are At or Near Extreme Levels…
For our active Russell 2000 portfolio, we see that many current exposures are at or near the high or low ends of their historical ranges, most notably those for Exchange Rate Sensitivity, Growth, Leverage, Liquidity, Profitability, Size and Volatility (Exhibit 2). We have written about changes in factor exposures, see “Market Sensitivity Exposures: “And the ‘New Normal’ is” and “Who Blew Up My Tracking Error” for some examples. In these posts we also addressed changing correlations, which we continue to see.

…and Factor Volatility Remains High
At the same time, most factors’ volatilities are extremely high—in general, above the 90th percentile relative to where they have fallen over the past 25 years.

One feature of our performance analytics software integrates the impact of changing volatilities and correlations within each risk model factor block: the percent of active risk from style covariance. This measure tells us how much diversification we are generating from the factor block. The percent of active risk from style covariance has also climbed since February, when it was negative (Exhibit 3). By June it reached a 25-year high, indicating that factors were not providing diversification. We have not seen the same thing for the industry factor style block—though it has risen slightly this year, it remains close to zero.

Source: FTSE Russell, Qontigo

Note: Black dot denotes current exposure.

Exhibit 3. Percent of Active Risk from Style Covariance, Russell 2000 vs. Russell 3000

Source: FTSE Russell, Qontigo


It is highly unlikely that a small-cap manager would have an all-cap benchmark (those that do should pay extra attention), but this situation is quite illustrative of the possible unexpected risk effects on many active portfolios driven by:

  • the changing nature of benchmarks,
  • what appears to be a lasting shift in factor volatilities,
  • and large changes in factor correlations.