Understanding correlation in relation to diversification

Diversification remains a well-established investment mantra, despite being subject to criticism during and after the financial crisis in 2008. On the surface, such criticism could be justified, given that often no risky asset is left untouched in a market sell-off.

In adverse market conditions, it's often said that "cash is king", and investors can be forced to sell all risky assets. This, for example in 2008, resulted in a situation where equities from developed economies sold off together with emerging markets and high yield bonds, which increased correlations between risky assets globally. 

From a performance perspective, government bonds did really well during and after the crisis, with a risk-off mood in 2008 followed by long-standing support from the central banks across major economies. Today, the total return for government bonds hedged to US dollars has reached almost 40% since 2009 (based on Barclays Global Aggregate Index). Equity performance was exceptional as well, driven by the ample liquidity and economic pick-up over the last eight years. Its US dollar-hedged total return is approaching 170% for the same period as for bonds (based on MSCI World index). So has there been any lasting impact on cross-asset class correlations in an environment of rising bonds and equities?

The short answer is no. What we have seen post-crisis is really a continuation of the new ‘Millennium era’ (after 2000) of negative stock-bond correlations, which generally shifted to lower averages compared to the ‘bad old days’ (before 2000). As shown in the chart below, correlations might occasionally pick up, although we don’t believe the old higher averages are coming back any time soon. Moreover, ‘millennial correlations’ still float around a lower long-term average and in recent years fell deep into negative territory again. This can benefit multi-asset portfolios, which invest across asset classes.

Meanwhile, correlation dynamics for stocks individually also suggest long-term benefits from geographical diversification.

Historically, equities exhibit strongly increasing correlation during periods of turmoil in financial markets. Yet over the long term those correlations tend to revert to historical averages.

The question relevant for a long-term investor for times of increasing correlations: is this a short or long-term pattern?

A comprehensive study by L. M. Viceira, K. Wang and J. Zhou (Global portfolio diversification for long-horizon investors) introduces two factors driving correlations globally: (1) risk premium and (2) fundamentals.

The risk premium effect leads to the short-term spikes in correlations observed across historic market crises. It is driven by news and economic sentiment, which shape risk-on and risk-off environments. If investors buy and sell all risky assets only due to their changing attitude to risk, this can drive short-term correlations higher, which may have little to do with the long-term returns.

The second factor, fundamentals, is more structural and long-term focused. It implies that earnings, together with fundamental economic growth, define the long-term correlation trends in asset returns. Viceira, Wang and Zhou find no increase in co-movements for fundamentals between different geographies, which means that long-term correlations do remain stable. This very factor matters more for longer-term investors and suggests diversification has remained invaluable.

Short-term correlations can increase during hard times. However, as soon as we are able to take a longer-term perspective and look through the temporary spikes in risk aversion, the beneficial effects of diversification come through, as proved by the long-term correlations across the regions and asset classes. 

With this in mind, we prefer to stick to the conventional investment rule: not to put all eggs in one basket. While correlations keep rolling - we stay diversified.

Thanks Martin Dietz for the inspiration and helpful comments, and Charlotte Xu for your valuable input