June 20, 2024

Higher correlation of multi-asset returns - temporary legacy of Covid, or permanent change?

Indrani De, CFA, PRM

Head of Global Investment Research

Robin Marshall, M.A., M.Phil

Director, Global Investment Research

Mark Barnes, PhD

Head of Global Investment Research, Americas

Alex Nae, M.Sc

Quantitative Research Analyst, Global Investment Research

Correlations of asset returns (CAR) clearly play a key role in asset allocation decisions, given the potential benefits from portfolio diversification. But these correlations are not directly observable, and must be estimated from underlying data sets. In addition, probability theory informs us there may be a link between volatility in returns and measured correlation. Specifically, in periods of higher volatility in returns, measured correlation will be higher, even if the underlying drivers of correlation have not changed.

This has led some to conclude “correlation breakdowns may reflect time-varying volatility of financial markets rather than a change in the relationships between asset returns.”  But despite these statistical issues, changes in key macro drivers may also explain changes in CAR, and indeed the underlying volatility in returns, which drives the higher correlations.

In this paper, we attempt to identify key drivers of correlations in returns, using FTSE Russell multi-asset index data since 2000.

Key takeaways: 

  • Higher correlation of US asset returns has persisted since Covid
  • The persistence may be due to higher-for-longer rates and inertia in core inflation
  • Only US HY credit shows no structural break in correlation
  • Inflation may be a strong driver of the higher correlation in returns, even if the relationship is non-linear
  • Less stable correlations suggest investors need vigilance in making asset allocation decisions

Points of differentiation:  

  • Analysis of multi-asset returns using FTSE Russell multi-asset index data
  • Linking correlation of returns analysis to asset class characteristics
  • Detailed assessment of the role of inflation in driving higher correlation of returns
  • A consistent narrative for why, and how, the correlation of returns has changed since 2000

What our research means for investors?

The paper provides investors with a thorough guide to the evolution of the correlation of US asset returns since the 1990s, and why correlations may be less stable since Covid, and what that means for portfolio models built around the notion of low, and stable, correlations of returns.

  • In an article we published last year we pointed out the observed correlation of US stock and government bond returns had increased sharply in 2022-23, as central banks raised interest rates and inflation rose sharply. This ended a long period, starting in the mid-1990s, in which the correlation of global stock and government bond returns had been relatively stable, low and even negative (during the deflationary shocks of the GFC and Covid). Indeed, evidence since 2021-22 confirms that relying on historical correlations to predict future correlations may give poor results.

    In this paper, we start by looking at the longer-term correlation of returns between US Treasuries and the Russell 1000 index of large-cap US equities. We then look at what correlations might be expected, given the structural differences between asset classes (from “risk-on” assets like ordinary equity, to cross-over assets like high yield credit, to higher quality investment grade credit, to pure “risk-off” assets, like government bonds).

    We then present empirical evidence on the correlation of US multi-asset class returns since 2000 and identify the key drivers of these correlations. We also assess the period since Covid in more detail, finding evidence of a series of “perfect storms” that have boosted CARs in different sub-periods. We then offer possible macro-economic and policy rationalisations for these high correlations.

  • Several factors help to explain why the correlation of returns between government bonds and equities was both relatively low and stable from 2000 to the outbreak of Covid in 2020-21.

    Firstly, low inflation rates became embedded in G7 economies, helped by globalisation and low inflation in tradeable goods. Structurally low inflation meant that investors focussed on cyclical fluctuations in real growth and corporate earnings as the principal drivers of equity returns, as well as on nominal bond yields. In cyclical slowdowns, monetary policy easing by the Federal Reserve drove lower US Treasury yields (and superior bond returns). Conversely, equities outperformed during growth recoveries, while government bond yields increased.

    Secondly, central banks adopted inflation targets, generally close to 2%, to consolidate low inflation rates and inflation expectations. Thirdly, relatively high labour migration rates and labour supply elasticity meant Phillips curve trade-offs between unemployment and inflation became more favourable. As a result, a very benign economic environment developed in the G7 (known as the Goldilocks era), during which growth and inflation were neither too hot nor too cold. The US saw stable and apparently predictable nominal GDP growth of about 6%.

    The combination of these factors drove strong support for asset allocation models, like 60/40, or the so-called Norway model, that are built on a 60% target weighting in equities and a 40% target weighting in bonds and credit. These models assume that bonds are reliable diversifiers for risk assets like equities, and that the relationship showing a low correlation of returns would persist. It is well documented that the Goldilocks era ended abruptly with the Global Financial Crisis (GFC) and the resulting global deflationary shock in 2008-09. Nominal GDP contracted sharply, as Figure 3 shows, and a deep recession unfolded. But despite the enormous macro-economic and financial uncertainty caused by the GFC, there was no structural break in the correlation of US asset returns during the 12 years that followed the crisis, with the correlation remaining close to -0.30.

  • Substantial market and policy cross-currents may explain why there was no structural break in CAR after the GFC. On the one hand, in the initial stages of the GFC, correlations briefly turned strongly negative, as safe haven flows into Treasuries drove yields towards historic lows, while equities sold off as the scale of financial dislocation became clearer. Corporate earnings and GDP growth expectations were also revised sharply lower.

    There was enormous financial and economic uncertainty until the Fed implemented a zero interest rate policy (ZIRP) and quantitative easing (QE). Once the Fed embarked on this major monetary stimulus, equity and broader asset returns both recovered, and the correlation of US equity/bond returns reverted to the post-2000 mean near zero. It is also important to note that inflation stayed low in the post-GFC era.

  • Finally, the much higher correlations of returns since 2021 raise the question of whether a confluence of events, or “perfect storm” has driven higher correlations over the period? After the initial sell-off in equities and the US Treasury rally when Covid emerged, the correlation of returns recovered strongly once the Fed cut rates to zero and adopted QE asset purchases (including credit) in 2020-21.