Blog Posts — March 20, 2020

Quant Quake 2020? As Factor Volatility Mirrors Market Volatility, Most Returns Head in the Wrong Direction

by Melissa R. Brown, CFA

Equity investors, needless to say, have faced a brutal market since February 20. But factor-based investors have experienced additional pain from factor returns, both those that would be expected to be compensated and those on which most managers do not have a view (because their long-term returns are close to zero). For those factors, all managers, not just those who focus on style factors, may have been surprised at the dramatic impact the factors may have had on active returns.

Factors that many investors bet on, such as Momentum, Value and Earnings Yield, have produced negative returns in most regions from February 20 through March 18, with the returns for the latter two often three or more standard deviations below average.[1]  (In Japan, the only region to see a positive Value return, performance was three standard deviations above average.) With no diversification in most regions, it’s surprising that value-type measures have fared so poorly. Dividend Yield has lagged in most markets, too—also a bit puzzling, as the Fed sliced rates to near zero.

Market Sensitivity has behaved as expected, with a highly negative return indicating that low Market Sensitivity (beta) stocks have far outpaced their higher-sensitivity brethren. This has been true in every region, with returns ranging from five to nine standard deviations below average. At the same time, while Low Volatility stocks have done better than those with higher Volatility,[2] returns in the US and UK have been relatively muted.

In every region, large-cap stocks have beaten small-caps (that is, the Size factor return was positive), but nowhere has the difference been bigger than in the US, where the return was nine standard deviations above what would be expected over this period.

The only factor to behave as expected in most regions (except Japan and Emerging Markets) was Profitability, where companies with higher return on equity and assets, and other indicators, have generally had better returns.

There are at least three factors in our risk models that meet the criteria of low-or-no long-term risk premium expectations across most managers: Exchange Rate Sensitivity, Leverage and Liquidity.  That is, over time these factors produce no excess return, and their positive and negative returns tend to be small in any given period. Over the period of the market rout, these factors produced returns of higher-than-average magnitude in most regions. Ignoring exposures to them could have meant unexpected hits to portfolios.

Not surprisingly, in a market driven by substantial concerns about the economy, highly levered stocks suffered relative to those with little or no debt. That relative shortfall was most acute in the US, where the return was 14 (!) standard deviations below average in both the broad US and the small-cap models. Exchange Rate Sensitivity’s returns were positive in some regions and negative in others (as expected, since the factor is local-currency based), with the lowest return in the UK (a four-standard deviation shortfall) and highest in Asia ex-Japan (+3.5 standard deviations). Finally, returns to the Liquidity have been highly negative (five to 10 standard deviations below average). We suspect that might be because investors sell what they can sell (i.e., the more-liquid names) when they are trying to exit the market as quickly as possible.

While not all the factor comments apply to all markets, all markets have seen most factors produce outsized returns over this difficult period.

Exhibit 1. Factor Returns from February 20 to March 18, 2020

Source: Qontigo

Note: Highest and lowest factor returns region-by-region are highlighted. An “*” means the return was more than two standard deviations away from the long-term average.

Is this worse than the “Quant Quake” of 2007? Perhaps. It certainly has lasted longer, and the magnitudes of factor shortfalls have been similar. Part 2 of this blog post will look at the daily results in more detail.

[1] We usually use beginning-of-period expected factor volatility for this calculation. Since volatility has risen so much during this period, for this analysis we used the figures as of the end of February.

[2] Market Sensitivity is based on a time-series regression of returns against the market; Volatility is cross-sectional and net of Market Sensitivity.