This century has been good for investors in traditional asset classes in our view. Not only have bonds and equities returned 4.2% and 5.2% per annum respectively to end May 2022, but correlation between the two has been negative overall.
'Classic' portfolios such as ‘60/40', representing notional allocations to equities and bonds respectively, therefore, have felt a substantial tailwind.
We can see this nicely in our 'Battenberg cake' chart in Figure 1 plotting monthly returns of world stocks and world bonds since 2000. In the top right-hand quadrant, both engines in our classic portfolio are firing nicely. We see the negative correlation effect - stocks up, bonds down and vice versa - in the top-left and bottom-right quadrants. These three quadrants account for around four months in five. 'Troublesome' months are the ones shaded red - where both equities and bonds generate losses and correlation, by definition, goes positive. A total 18% of months, or roughly one in five, constitute this red bucket. What's worse, over this time period, not once has this happened four months in five, as it has in 2022.
Source: Man Group/Bloomberg/MSCI
The problem is brought into focus when we take a longer-term view. Figure 2 illustrates that the negative correlation between bonds and equities really is a phenomenon of this century; positive correlation is more the norm when we look back multiple centuries.
Source: Man Group
So, bonds and equities rising and negatively correlated to each other would seem to be a match made in heaven, but it is a fairly recent phenomenon. Further, with bonds and equities declining 7.7% and 11.2% respectively to end May, 2022 has been a stark reminder that things can go wrong quickly.
What can pension funds do about the correlation problem?
Two techniques that can be used to improve the performance of 60/40 portfolios in backtests:
1. Momentum-overlays, which in effect use trend-following techniques to reduce the exposure to a market if the price declines over the timescale of a few weeks, because we believe that the price is likely to continue falling.
2. Volatility overlays, which borrow from trend-following techniques to scale positions inversely with volatility to keep a more constant risk profile over time.
These two techniques help solve the correlation problem. On the momentum side, positive bond/equity correlation is only a concern when prices fall; multi-asset and pension fund investors are happy when both assets rise at the same time. Using a trend-following system to take off some exposure when prices fall can potentially mitigate losses. On the volatility side, increasing bond/equity correlation leads to lower diversification in the portfolio and hence higher volatility. We believe if we scale our positions inversely with volatility then we can reduce risk when markets are volatile and cut off that pesky left tail.
More can be done, however. History tells us that we tend to see a flight to high quality government bonds in times of stress, even when yields are low - we saw this at the height of the coronavirus crisis in February/March 2020 and when concerns around war in Ukraine were most acute at the end of February this year. Bonds perform a valuable role in multi-asset portfolios in this context.
The opposite effect does not hold true: when bonds sell off we do not typically observe a flight to equities. In fact, we have seen cases where bond sell-offs can cause equity sell-offs. We saw this contagion effect in 2013's 'Taper tantrum' and have observed it in spades in 2022. Bond-driven sell-offs represent a significant threat to multi-asset portfolios, but we still need bonds in multi-asset portfolios for their flight-to-quality properties.
The answer to this conundrum is responsive correlation risk overlays which allow us to maintain long bond exposure, but cut this exposure when contagion risk is high.
When it comes to responsive yet robust correlation estimation, the more data points the better. Our experience using daily data is that correlation changes too slowly to be of use in rapidly changing market environments. We find that we have to higher frequency data to get the responsiveness we need; it allows us to cut portfolio exposure quickly when those crucial hedging properties of bonds and equities are missing.
Implementing correlation overlays using a systematic approach
In general, a systematic approach encourages us to be rigorous in researching our solution through the necessity to write down methodologies, and then it implements that solution in a perfectly structured way. It also allows us to implement our overlays without emotion; behavioural finance teaches us that human being do not always behave rationally, particularly in a crisis situation. Systematic approaches enable backtests, to see how our overlay may have helped (or hindered) in the past.
It is this last facet of the systematic approach that needs careful attention when considering correlation. How reliable can a backtest of a responsive correlation overlay be? The data underlying Figure 2 is mostly only available monthly. Daily data is available at most back to the 1980s, and good intraday data only for the last decade or so. During this time, we have not seen many significant declines in bonds. How can systematic traders evaluate a responsive correlation risk overlay when the problem is not readily visible in historic data?
Correlation risk models therefore have to be forward looking. We cannot rely on better backtest metrics to justify their inclusion. Common sense has to be at the heart of the process, and we need to look no further than 2022 for justification of this.
Graham Robertson is head of client portfolio management at Man AHL (part of Man Group)