How good is ‘Risk Parity’ as a way of allocating across asset classes?
Professor Andrew Clare of Cass Business School presents his research on the impact of a risk parity approach on a multi-asset class portfolio.
How should we allocate our investment portfolio between the growing range of asset classes now available? This is one of the most difficult decisions that any investor has to make. At the same time it is also the most important decision. Research (and common sense) shows that getting asset allocation right is far more important than finding a manager that can outperform their particular market.
Many institutional investors, including large insurers and pension funds, make use of sophisticated optimisation software to help them allocate their funds across different asset classes. The inputs generally require the user to specify the return they expect on any asset class of interest, along with the likely volatility of those returns, plus the correlations between these asset classes. The technology behind this approach to asset allocation is used to identify the ‘optimal’ asset class mix for their retail clients.
The output generally gives reassuring, scientific looking data with precise looking charts of expected outcomes. But ultimately the pseudo-scientific output is only as good as the inputs. So unless you are very good at forecasting asset class returns into the indefinite future, and return volatilities and correlations, the results of your optimisation process, however fancy, will be nonsense. Remember the old saying: garbage in equals garbage out.
In response to the growing dissatisfaction with apparently sophisticated optimisers that simply give us the answer that we have put into them, there is now a growing body of evidence that shows that alternative, almost simple-minded approaches to the same asset allocation problem, produce results that are at least as good as those produced by the most sophisticated optimisers.
One such approach, referred to as ‘1/N investing’, advocates allocating equal amounts of capital to each asset classes of interest. So for example if there are ten (N) asset classes then the approach simply requires that each asset class has a weight of 1/10, that is, 10%. To some extent, any other type of weighting implies that one knows something about the world. In my view, the events of the last few years have shown us that we know only one thing – and that is that we know virtually nothing about the future!
In this blog I want to highlight another approach to asset allocation which is getting a great deal of attention from institutional investors, and which is similar in spirit to the 1/N approach to asset allocation. This is the ‘equal risk’ approach, also known as ‘risk parity’.
The idea is that rather than investing in the asset classes using information from an optimiser, or allocating an equal amount of capital to each one (1/N investing), instead this approach argues that capital should be allocated in such a way that the volatility of each asset clas mulktiplied by its weight in the portfolio is the same. So an asset class with low return volatility would need a higher weight than one with high return volatility.
For example, suppose that an investor was considering investing in just two asset classes, one with volatility of 10% and the other with volatility of 30%. A risk parity approach to the weighting of these asset classes would require that the investor invested 75% in the low volatility asset class and 25% in the high volatility asset class.
At Cass Business School’s Centre for Asset Management Research (CAMR) we have been investigating the impact that a risk parity approach to investing can have on a multi-asset class portfolio*. We collected return data on the sub-components of five broad asset classes: Developed and Emerging market equities, developed economy bonds; commodities; and commercial property. All together we used data on 95 individual asset classes.
We then undertook the following experiment. We calculated the performance statistics for a simple buy and hold strategy for the five main asset classes, where we allocated 20% of the portfolio to each of the five asset classes, that is, we chose the asset class weights based upon the principles of 1/N investing.
Next we formed a portfolio based on these five broad asset classes, where the weights were set at the beginning of each year so that each broad asset class had the same weighted volatility as every other asset class, based on the volatility of that asset class over the previous year.
The first two bars of Figure 1 show the return achieved from both approaches. They are almost identical. The 1/N approach achieved a return of 6.71%pa, while the risk parity approach achieved a return of 6.78%. Not very encouraging really.
However, we also applied the same approach within each asset class. These results, also shown in Figure 1, are more encouraging. In most cases a risk parity approach seems to have enhanced returns, in some cases quite considerably. For example, a 1/N approach to an investment in a wide range of emerging market equities would have produced an average annual return of 5.48%, while a risk parity approach produced a much more impressive return of 9.58%. This result suggests that investors were not being rewarded for exposure to very risky emerging equity markets and that, by comparison, overweighting the less volatile emerging equity markets would have been a more profitable strategy.
Now let’s now consider the risk in these strategies. Figure 2 shows the maximum drawdown statistics for each of the strategies and asset classes shown in Figure 1. A portfolio’s maximum drawdown is the maximum peak to trough decline in its value over a particular time period. This statistic should be of crucial interest to any investor that might need to access their funds in a hurry, or if they are approaching retirement, or if they have retired and are in income drawdown.
The interesting feature of Figure 2 is that when applied across broad asset classes, risk parity can reduce the maximum drawdown compared to a 1/N strategy: from 47% to just 20%. However, within broad asset classes the maximum drawdown is almost the same for both approaches. And in the case of developed economy bonds, it is considerably worse.
Our results indicate that a risk parity approach to asset allocation may not enhance return when applied to a range of broad asset classes, but that it might reduce drawdowns – a key concern for investors. Conversely, our results with regard to, for example, the choice of weights for a global equity portfolio suggest that a risk parity approach to the weights might enhance the return over time, but may still leave the portfolio vulnerable to a large drawdown. In other words it may enhance the average return, but it does not eliminate the downside risk that investors fear most.
Overall, risk parity does appear to offer something to investors, although it may not be the investment panacea that its proponents claim.
A full version of this research is available here.