In times of crisis, quant analyst recommends quant

A popular joke in provincial and specialist media is to exaggerate a local angle to the point of absurdity. “South Ayrshire Golf club owner loses 2020 presidential election,” for example, or “New Yorker stepping down as UK PM”.

The same thing happens in broker research, but sincerely.

“Stock betas demonstrated unusual instabilities during the Covid-19 crisis due to emergence of new sources of risk,” begins a note published today from Barclays’ Quantitative Portfolio Strategies team. “This kind of instability has negative effects on systematic strategies.”

It’s true! An unprecedented global pandemic did affect the relative volatility for individual stocks. Or, as Barclays puts it: “The unusual nature of the Covid-19 crisis redefined the nature of systematic risk in the market and therefore reshaped stocks’ beta dynamics at the sector level.”

The note is an update of research first published in the May 2021 issue of The Journal of Portfolio Management (PDF link, page 139, with a Google search workaround if you hit the registration barrier). Then as now, regression analysis of daily market returns finds that when the news headlines are apocalyptic, nearly every sector becomes a binary bet on humanity’s survival:

And inside those sectors it becomes a right old mess:

These seemingly obvious conclusions underlie a serious point. In times of stress it’s commonly advised to seek shelter in low-volatility strategies. This has not proved a fail-safe defence when the stress is wide-reaching, as shown by the 2015 China slowdown panic as well as by the early-Covid crash:

As well as moving the goalposts for how much absolute stock sensitivities might fluctuate each day, Covid was a lesson in how “regime shift” can upend all the quant rules, says Barclays. Relying only on historical data to calibrate beta therefore isn’t clever.

And whenever low-vol can’t be trusted to remain low-vol, things break. Its research zooms in on November 9, 2020, when Pfizer and BioNTech announced breakthrough trial results for their vaccine candidate. The reaction among individual stocks was around 12 times larger than that suggested by their historic beta, which resulted in momentum chasers having their worst day in two decades:

The ideal strategy here would be not to have another pandemic. That’s outside the control of most investors, so Barclays tests various ways to hedge against more “unintended systematic risk exposures”.

Option one is to double sort. The vaccine tape-bomb in 2020 hammered momentum strategies in part because the relative volatility in the long side of their portfolios didn’t match the short side. High-beta stocks had been leading the sell-off for months. Conventional momentum funds became heavily short of the market by accident, so got crushed when it turned.

A smarter way to invest is to arrange all stocks into beta buckets, then arrange each bucket individually by their momentum signals. Stocks that rate highly by momentum within each bucket are grouped together to create a long basket, and vice versa for a short basket. The result is a portfolio that should be equally balanced between long and short.

This proves more defensive, as shown below, but is still not great. One of the problems is that hedging by industry group, a popular way to refine this kind of strategy, has only an incremental benefit when everything is going haywire:

Option two involves ignoring total returns and using a residual measure (as in, a stock’s return relative to beta as a multiple of the benchmark). In theory, residual returns will flatten out marketwide influences and capture only news specific to the stock.

Jump to page 153 of the original paper if you want to see all the working, the upshot of which is that measuring residual returns was just too fiddly. Tiny changes to Barclays’ model had outsized effects so it became difficult to find the signal in the noise.

Option three is where systematic becomes tricky to tell apart from active. The long and short portfolios are weighted by momentum signals, then matched by average betas and sector allocation. Barclays calls this the balanced approach.

But sector taxonomy is a blunt instrument. Hoteliers and online teachers will usually end up in the same industry group, for example, but had very different pandemics. The catastrophic nature of catastrophes means that within each sector, hedging works best when every single stock is individually matched between the long and short portfolios by the most granular measures available.

Does it work? Yes. Is it easy? No. Is it worth the hassle? Debatable. The needlepoint approach to hedging would’ve delivered best protection in the initial Covid crash (chart one below), but also shows the sharpest underperformance in normal times (chart two).

Again, this can’t be wholly unexpected. The world is in co-ordinated meltdown only occasionally, so if the sample period is a couple of decades the established ideas about which sectors see-saw will hold true more often than they don’t.

From all this, Barclays concludes that the best trade-off is to use its balanced approach. Just don’t get too clever, so rely on the broad industry groups.

An alternative conclusion (albeit one rarely championed by hedge funds) might be to just not bother. For every time period cited above, all strategies would have been left for dust by an S&P 500 tracker. Risk managers will of course argue that such a comparison misses the whole point of a hedge, and fair enough. But then, “In times of crisis, quant analyst recommends quant” is no less glibly reductionist than those headlines at the start of this post.

Further reading:
— Inflation is the friend of your trend
— ‘Buy the haystack’ approach still hard to beat