Modelling correlation: from zig-zag to zig-zig

The hedging property of bonds versus equities in times of stress is something market practitioners have seen reinforced time and again. But the negative correlation between the two markets is not – despite appearances – a law of nature.

Research increasingly suggests the stock-bond link responds to investor sentiment more than previously thought.

With that observation in mind, academics such as Geert Bekaert, a professor of finance and economics at Columbia Business School, are striving for new ways to forecast how the correlation might switch.

The academics point to a huge time variation in the correlation, which these fundamental forces cannot explain. Bekaert describes the puzzle of what causes this bumpiness in the correlation as “the big question” facing asset owners.

The influence of sentiment could potentially mean the stock-bond correlation turns out to be unforecastable, thinks one asset manager.

At the same time, recent events infuse the debate with greater real-world urgency.

When the S&P 500 took a hit at the start of 2018, the bond market fell alongside it, causing some to question how long the adage would hold that when equities zig, bonds zag. Days where the S&P 500 lost 0.5% or more and the 30-year US Treasury bond yield rose 3 basis points or more have become increasingly frequent, analysts from Morgan Stanley point out.

Year-to-date correlations have been mildly negative. But according to one asset manager’s short-term measure, correlation has been fluctuating between –0.5 and +0.2. The correlation peak occurred right around the February spike in equity volatility.

In both academia and industry, the task in hand is made tough because the relationship has switched only a handful of times throughout history, which means any model is prone to overfitting

Asset managers are not taking this as a cue to abandon the principles underlying their asset allocation. Their expectations about correlation have shifted since February, but only modestly – and most see the stock-bond dynamic changing only if inflation materially increases.

But recent events have nudged them, too, towards re-examining how they forecast the stock-bond link.

In both academia and industry, the task in hand is made tough because the relationship has switched only a handful of times throughout history, which means any model is prone to overfitting – reaching false conclusions because the calculations are based on a small number of datapoints.

Historically speaking, the stock-bond correlation has tended to be positive when inflationary factors dominate the market environment and negative when growth factors rule. Correlations in many developed markets have been persistently negative for almost two decades during a time when growth expectations have been the predominant driver.

How investors’ risk perception might play a part in driving correlations is still little understood.

What is becoming clearer, though, is that even without a growth or inflation shock, asset managers could wake up to find themselves in a world where equities zig and bonds zig too. The question remains whether that day can be predicted.