Risk Parity and Volatility Targeting as a Danger – Let’s Get Real with Some Numbers

There is growing talk that volatility targeting and risk parity are the dangerous new “portfolio insurance” strategy of the decade. In the post-’87 crash period, the view was that portfolio insurance sowed the seeds of market destruction by creating a market decline feedback loop. As an option replication strategy, portfolio insurance automatically increases risk exposure in up markets and cuts exposure in down markets to form an option-like pay-off of returns. The feedback loop meant that a market sell-off would be exacerbated as equity prices declined. Selling would only lead to more selling as the insurance adjustment kicked in. This came to a head in the ’87 crash as portfolio insurance strategies generated wave upon wave of selling in a down market. Given the systematic behavior in an illiquid market, portfolio insurance sellers became a more significant portion of futures index trading. Whether cause or contributor, portfolio insurance was never the same after October 1987, and there is now a fear of other strategies or algorithms that may have feedback loops to further a market decline.

Risk parity and volatility strategies have feedback loops that may exacerbate a market sell-off tied to changes in volatility. The idea is simple, albeit different from portfolio insurance. If there is an increase in volatility for strategies with a volatility target, positions will have to be cut. Similarly, if risk parity exists across a set of assets, an increase in equity volatility will lead to a selling of equity exposure relative to other markets. Hence, volatility increases will lead to a market sell-off that could further increase volatility, which will then contribute to further selling.

This good story should be a concern, but it is still just a story. It is true, but it is a matter of significance. It needs to be translated into numbers and market impact. The impact or feedback associated with the story is related to the selling size relative to the market size. We can measure size through a simple model.

Let’s assume an investor wants to target a 60/40 stock/bond blend at 8% using mini-SPY contracts and Treasury (TY) note contracts. The current combination will generate a volatility of 5.98% using a volatility of 9% for equities and 5.25% for bonds, with a correlation of .1. The portfolio volatility would be lower if the stock/bond correlation is negative.

We can calculate the futures necessary to replicate this 60/40 mix per $100 mm. The portfolio will have to levered by 164 equity and 147 bonds contracts per $100 mm to go from a 5.98% vol to an 8% vol target. If the volatility for stocks and bonds increases by 1% point and the correlation stays the same, 72 equity and 64 bond contracts per $100 mm will have to be sold. Therefore, for this simple volatility targeting program, we can determine that 720 equity mini-contracts per billion dollars will change to hit the portfolio target. This would be .17% of daily volume, given about 414,000 equity index contracts daily. You would need about $6 billion of equity exposure or about $10 billion in 60/40 volatility target programs to represent about 1% of daily volume. This tells us the sensitivity in contracts and dollars of exposure to a change in volatility. If volatility increases significantly, the selling pressure will be even greater.

So, how much is in volatility-targeted programs?

That is the critical question. If the answer is $100 billion, then an adjustment of 1% higher equity volatility will be 10% of daily volume if the adjustments are all made simultaneously. Of course, this is from a straightforward stylized example.

Is this volume from volatility adjustment strategies enough to create a feedback loop? I don’t know. But it gives us a better idea of what to expect if we know the extent of trading. We do know that some hedge funds may have tens of billions tied to volatility adjustments.

Of course, we know that market liquidity is elusive. It will not be present when we need it. We can count on the discipline of models. We cannot count on volume on the other side of these trades.