Predictability in finance is rare. It is quite difficult for someone to predict the daily price moves of securities or indices, for instance. However, there is one wager that seems to offer a stunning 100% hit rate: Whenever stocks and bonds sell-off simultaneously to a significant degree, the next day will be flush with sell-side reports or major news articles tying the sell-off to the actions of risk parity strategies.
We saw such episodes of finger pointing in 2013, in 2015, in 2018, and now again in March 2020. “Analysts Point Finger at ‘Risk Parity’ Strategy in Market Rout,” as one Financial Times headline on March 20 put it. It is quite likely that this cycle of sensationalism will repeat. In our view, the claims are often full of unsubstantiated, wrong assumptions and hearsay, and we advise our readers to take them with a grain, or better a bucketful, of salt.
In this edition of Systematically Speaking, we try to address why this analysis keeps recurring, and explain why the story you read on the finance pages is so often inaccurate and incomplete.
Let’s start with why these news articles raise so much attention. To us, the reason is quite straightforward. Combined stock and bond sell-offs tend to occur during times of substantial financial stress, accompanied with fear. When fear is high and things are moving fast, we look for easy explanations of the cause. We want to pin things on some distant, abstract but simple concept—“machines” or “algos” fit the bill perfectly.
Risk parity is often the victim of such hasty conclusions:
“Risk parity leverages stocks and bonds, stocks and bonds are selling off, therefore risk parity strategies must be deleveraging fast and causing all the sell-off.”
There are numerous logical fallacies in this reasoning, the most obvious of which is the “fallacy of the single cause,” a cognitive bias characterized by the tendency to think that a complex phenomenon can have one cause. The fallacy of mistaking correlation for causation applies, too. Financial markets are really complex places. There are myriad different types of investors with different goals, reacting to news and trading with each other and allowing price discovery at the same time. Singling out one type of investor as the cause for a sell-off is often wrong, and in the case of risk parity it is almost surely wrong.
So what is it that the sell-side analysts and media don’t get right?
We believe that the biggest mistake is assuming the size of the risk parity investment domain. By our reckoning, drawing on databases such as eVestment and HFR, assets under management in risk parity strategies amount to about $120 billion.
Now, let’s simplify this and imagine there is one risk parity manager deleveraging her $120 billion portfolio with an existing allocation to equities of 50% ($60 billion) and to bonds of 150% ($180 billion). Say within equities the allocation to U.S. stocks is 10% ($12 billion) and within bonds the allocation to U.S. Treasuries is 15% ($18 billion). Suppose that the volatilities of these markets suddenly doubled and that the manager uses a naïve risk parity approach that weights asset classes in inverse proportion to their volatility: the result is that she halves the allocations. In summary, we are looking at an exposure reduction of $6 billion in U.S. stocks and $9 billion in U.S. Treasuries.
How big are these sales? Let’s get some perspective. In our figure we show the evolution of the daily U.S. dollar trading volumes of S&P 500 Index and U.S. 10-year Treasury futures contracts—the two left-hand charts. For example, during the most volatile day of March 13, the total volume of the S&P futures contracts (not including options) was $900 billion.
Risk Parity is most likely a drop in the Bucket: A Hypothetical Analysis
We assume that the risk parity strategy manages a total of $120 billion and is allocated 50% equity (10% U.S. equity) and 150% bonds (15% U.S. Treasuries). Then we assume that every manager in the strategy halves its gross exposure in a single day. These charts show the hypothetical proportion of S&P 500 and 10-year U.S. Treasury futures trading that would imply on any given day between March 1 and March 20, 2020.
Source: Bloomberg, Neuberger Berman. As of March 20, 2020. The hypothetical data shown is for illustrative and discussion purposes only.
We don’t know exactly when risk parity started deleveraging, but let’s assume the strategy dumped everything in one day. This really would be a worst-case scenario: most large managers break down their rebalancing over a few days, precisely so as to mitigate any potential impact they may have in relatively illiquid markets and also to reduce the chances of whipsaw effects.
The right-hand charts plot the percentage of the daily volume that the risk parity strategy would have contributed were the strategy to deleverage completely on any one day. For example, if risk parity had completely deleveraged on March 13, then the risk parity sales would have contributed 0.6% of the daily volume of S&P futures and 3% of the volume of 10-year Treasury futures.
The lesson is that, even assuming the worst-case scenario of a complete deleveraging in a single day, risk parity would have likely been a drop in the bucket of overall transaction volumes.
Inflows and Outflows
There are yet other reasons why risk parity may not have contributed as much to the sell-off as advertised. One reason is that risk parity strategies react to volatility and correlation, rather than creating it. Strategies first collect evidence that the volatilities and correlations are spiking, and then start trading. We are not aware of any large risk parity manager who rebalances its portfolio every minute, which is the only way it would sell into every market and contribute to a sell-off minute-by-minute. Without first observing the rise in stock-bond correlations and volatility on March 13, why would risk parity managers have decided to deleverage?
It is possible that some risk parity managers experienced significant outflows that might have exacerbated the deleveraging rebalance. We believe the probability of this is rather low, however, because risk parity investors tend to be loss averse in general, and are likely to be running broadly diversified portfolios that mitigate the need to “change horses in midstream.” Moreover, most inflows and outflows for private accounts are likely to occur at the beginning or end of the month, to allow managers to plan for them. There may have been large risk parity clients who for some reason needed to redeem all at the same time on March 13, the worst day of the correlation spike, but we believe such an event is unlikely given the circumstances.
Furthermore, note that on a day like March 13, not only stocks and Treasuries sold off but also other risky asset classes that tend not to appear in risk parity portfolios—things like money market funds, real estate investment trusts, non-exchange traded commodities like palladium, coal and Chinese steel. The deleveraging of risk parity is unlikely to have directly impacted these markets.
Correlation Isn’t Causation
We believe the underlying factor that caused this extreme bout of risk aversion was very simple, and that was the sudden explosion of uncertainties. Pile the pandemic-driven economic shutdown on top of an OPEC price war, rising default probabilities and an already expensive stock market, and it would be a surprise not to get a combined sell-off in financial markets.
The resulting rush to cash and substantial outflows from the “massive passive” ETF complex, alongside an explosion of short interest for gamma hedging as option-implied volatility spiked, and the downsizing of risk in leveraged strategies such as hedge funds, trend followers, equity market neutral funds, portfolio protection strategies and, yes, volatility-targeting strategies like risk parity, inevitably led to market dislocations.
In sum, financial markets are complex places. Singling out risk parity as the cause can be great clickbait but we believe it is wrong to assume that risk parity single-handedly caused the twin stock and bond sell-off. Rather than the cause, risk parity is likely one of the victims here.