Correlation
“Nothing is more useful than water: but it will purchase scarcely anything.”
In my post on Policy Agnosticism, I suggested the severity of the next downturn will depend on how policymakers respond, but that it would not be sensible to attempt to front-run their decision. And it would be even more unwise to assume they sustain a given stance (easing or tightening) until an extreme “inevitable” End Game occurs. What does this mean for Strategic Asset Allocation? Do we like hiding at the front end of the curve, collecting more than 5% to wait for greater clarity before taking action? Or putting capital to work?
On the one hand, holding an elevated position at the front end of the curve provides dry powder (compounding at positive real yields) to step in opportunistically, once we see how policymakers act and have stronger conviction in how this cycle will play out. This is consistent with a “Policy Agnostic” approach, and given our absolute return approach, appears to be a prudent strategy. For those benchmarked to risk assets explicitly or implicitly, we imagine this strategy would be far from “safe” in career risk terms, and together with passive / systematic flows, helps to explain why such assets continue to trade at what we think are elevated valuations. These valuations in turn reinforce the case to hide at the front end of the curve for investors like ourselves with incentives more strongly aligned with long-term performance.
On the other hand, although patience is consistent with longer-term incentives, holding dry powder ready to opportunistically act means implicitly shifting to a shorter-term tactical approach. If the macroeconomic landscape is characterised by swings between debt deflation and hyperinflation fears, with policymakers reactively shifting between easing and tightening, moving to a tactical stance could be seen as a necessarily pragmatic response. And potentially offer opportunities to generate positive returns in an overall choppy market with no clear trend. However, we think it would be overconfident to believe that these attempts to “time the market” will be reliably profitable. Setting aside some dry powder to allocate on a tactical basis seems reasonable to us, but it would be unwise to rely excessively on predicting these swings.
If we are keen to put capital to work now, investments with lower Correlations to traditional asset classes would seem to be particularly desirable. We might add these as longer-term “anchor” Strategic Allocations expected to earn income or trend higher in price over the long term, regardless of short-term market swings. Of course, it is not surprising that lower-correlated investments would be attractive for portfolio construction, and this is no silver bullet. Correlations are unstable and time-varying, and calculating a coefficient using historical data and an arbitrary look-back window is less useful for portfolio construction than theory would suggest. This can often result in a false sense of security (and incorrect position sizing if such coefficients are used as inputs into risk models), as Correlations are lower / higher during benign / stressed periods. We think screening for investments with low historic Correlations to traditional asset classes is not sufficient.
Correlations are unstable! S&P 500 vs Gold. Chart: Bloomberg
Nevertheless, the trifecta of portfolio construction is Risk, Reward, and Correlations. It is not possible to construct a portfolio effectively without a view on all three of these. And they each suffer from the problem that calculating coefficients conditioned on backwards-looking historical data is procyclical, and generally results in misleading answers at market inflection points when resilient portfolio construction is most important. The ideal solution would be to take a forward-looking approach, anticipating what Risk, Reward, and Correlations will be in future market regimes. However, it is very difficult to anticipate what specific future scenarios will occur, and so instead it is better to assign probabilities across plausible future scenarios - with different Risk, Reward, and Correlation coefficients applicable in each case. Portfolio construction then requires optimising across these probability-weighted scenarios: allocating more risk to factors / premia likely to perform well across plausible scenarios, less risk to those likely to perform poorly, and diversifying across those that are mixed. And of course, stress testing to ensure the portfolio does not underperform critical thresholds even if low likelihood “tail” scenarios materialise.
When assessing the current set of likely future scenarios, we are confronted with an additional challenge. As the longer-term credit cycle has progressed, and policymakers have become increasingly activist, Correlations between investments are likely to keep increasing. If inflationary dynamics take hold, central bank tightening might weigh on all asset classes (as in 2022). Meanwhile, during the panic of a debt-deflationary episode, the market adage “all Correlations go to one in a crisis” remains apt (as in early 2020). More broadly, “search for yield” dynamics during years of low interest rates led to the financialisation of previously lower-correlation asset classes, with credit expansion providing the liquidity to push prices higher, and sell-side innovation packaging anything “sellable” into an investable “product” (such as ETFs / Securitisations / NFTs). The result of this frantic credit-driven financialisation is that it is increasingly difficult to find assets that, across forward-looking scenarios, are not sensitive to the credit / liquidity “tail” wagging the financialised asset “dog”. After sufficient liquidity has flowed in, even previously De-Correlated investment categories can become increasingly driven by investor positioning.
We currently like a threefold Asset Allocation approach. Firstly, it pays (literally) to set aside an elevated position at the front end of the curve to allocate tactically as the market potentially oscillates between inflationary and deflationary narratives. Secondly, we can try to assign probabilities to forward-looking scenarios, and put some capital to work in risk factors likely to perform well across many of these, while limiting exposure to those factors likely to perform less well. Finally, it makes sense to search for those corners of the investment universe that are likely to remain De-Correlated across future scenarios. Often this involves looking at niche asset classes, which is time-consuming and requires effort. But there is no free lunch, and it makes sense that the promise of De-Correlation is not likely to hold if an investment category is too obvious, or if too much capital has already been allocated. Such opportunities may also have better Risk / Reward, if yields have not been compressed and/or quality declined due to the creation of new “product” to meet investor demand.
Disclaimer: The content in this blog post should not be taken as investment advice and does not constitute any offer or solicitation offering or recommending any investment product.