Risk Parity

How does risk parity work?

Risk parity is a form of portfolio allocation that has become popular among investors over the past decade. The strategy provides a way to build portfolios without relying on projections for returns since the process aims to diversify risk. Moving away from forecasting returns is not a new concept. It is very similar to volatility targeting, where investors rely on the fact that volatility today has some predictive power of volatility tomorrow. Traditionally, investors have used mean-variance optimization or the Markowitz efficient frontier to build optimal portfolios with the highest expected return for the lowest expected volatility. However, this approach has many limitations; the estimates that go into the optimization drive the output so you can have a case of garbage in garbage out. Over the years, various modifications of the mean-variance approach have been proposed to improve portfolio construction. Risk parity is one such modification.

Risk parity aims to allocate portfolio risk equally across assets or individual risk factors. The risk parity portfolio is often the combination of stocks and bonds that maximizes the Sharpe ratio. Remember that this ratio is a measure that penalizes higher returns if they are associated with higher volatility. As such, risk parity portfolios tend to have a far higher bond allocation than the traditional 60/40 portfolio. The holdings tend to be around 80 percent to bonds. Part of the reason for this is that the 60/40 portfolio gives an illusion of diversification when viewed from the lens of risk. Since equities have 3 to 4 times the volatility of bonds, most if not all of the portfolio risk ( > 90 %) is concentrated in stocks and not diversified at all. While risk parity improves risk concentration, portfolio returns tend to be far inferior to portfolios with higher risk concentration. As such, leverage is typically applied to the portfolio to reach required or desired rates of return.

The critical thing to note here is that leverage is applied to the entire portfolio. In doing so, an investor can preserve the risk-adjusted characteristics of the portfolio while increasing the expected return and, subsequently, the volatility. One of the features of this approach is that assets with lower volatility will have higher allocations. Many of the dynamics described in the volatility targeting explainer series apply here. Specifically, prolonged periods of low volatility (and correlation) will lead to higher allocations and increased sensitivity to changes in volatility and correlation.

Furthermore, since leverage is applied equally to the entire portfolio, an increase in volatility would force these strategies to unwind positions across asset classes. In that respect, risk parity is a bridge between asset classes and can be a source of contagion risk within financial markets. However, their size is not that large, ranging from US$200-600bln; the use of leverage, typically between 3 to 5 times their AUM, makes their potential market impact far larger.

How do we track risk parity positioning?

We introduce a simple model to track positioning to monitor risk parity strategies. This model invests across 27 futures contracts across four jurisdictions and asset classes. The four asset classes include fixed income, inflation-protected securities, commodities, and equities. The strategy is not a pure risk parity strategy as we set the risk budget to 15 percent for inflation-protected and 23 percent for the other asset classes. Note that these weights align with S&P's risk parity 2.0 indices methodology. In addition, we set the target volatility to 10 percent and rebalance it daily. We choose to rebalance daily, not because we think this is common practice, but rather to see how allocation changes every day. In practice, there are likely thresholds that, when breached, would trigger rebalancing.

Today, we can see that risk parity allocation remains low historically. The total exposure of the risk parity strategy is around the 10th percentile going back to 2016. COVID-19 forced many strategies that use volatility as a toggle for exposure to unwind positions. These strategies will keep artificially low positioning until those volatile days fall out of the estimation window. In the case of our model, we expect exposure to increase in March of next year since we use a 2-year look-back window.

We also conduct a similar exercise to the one done in the volatility targeting post where we use the past year's returns and project allocation out to 1-month. This exercise aims to get a rough distribution of possible exposure over the coming month. The projection shows that exposure is stable in the coming month.

As always, thanks for reading

Tiago Figueiredo