Blog Posts — August 16, 2019

Do Yield Curve Inversions Impact Factor Performance?

A client recently asked an intriguing question: “Do style factors behave differently after a yield-curve inversion?” Intuitively, one might think they would. Particularly if a recession were looming, investors might turn to stocks that are less volatile, lower beta, more profitable, less leveraged, etc. We studied the impact of recessions and expansions on factor returns earlier this year and didn’t find much (see here), but thought a different angle might produce different results. Alas, that was not the case, but style-based investors who would be unlikely to make big changes to their strategies might take comfort in the results.

The curve has inverted three times since the inception of Axioma’s US model: in June 1989, July 2000 and August 2006. So we began by calculating six-month factor returns subsequent to each inversion. To determine whether the return in the six-month period was significantly different from the long-term average, we calculated the p-value[1] of the six-month return versus the average. The p-value shows the probability that returns in the period are significantly different from the long-term average, and low value suggests the current period is truly statistically different from the long-term average.

The results of these tests (and the low number of test periods) did not enable any firm conclusions about which factors might see their performance differ. Several factors did show significant return differences in the first two instances of inversion, but none showed differences in both. No factor’s return was significantly different from the long-term average in the 2007 instance.

Even ignoring the statistical significance issue, only three factors showed differences from the long-term average that had the same sign in all three periods. Exchange Rate Sensitivity, Growth and Size returns were higher than the long-term average, but Size returns were still negative in the latter two periods.

Standout differences included the highly negative return to Market Sensitivity and Volatility in 2000, and the extremely positive returns to Earnings Yield Profitability that year. The return to Size was much higher than average in 1989, and positive rather than negative.

Despite some big return differences, based on these results we would not recommend changing any style strategy just because the yield curve inverted. (Note that we also looked out a year after inversion and the conclusion was the same.) One reason that differences may seem insignificant is that volatility can change substantially as recession worries mount and, statistically, what would have been significant assuming the same volatility becomes insignificant. Still, it is always a good idea to test!

US Model Style Factor Returns

Source: Axioma

[1] Tests used a two-tailed t-test assuming unequal variance