The conventional view is that portfolio diversification didn’t help in the crisis. At the height of the crisis “risky” assets did move together, but their correlation with a key “safe” asset became strongly negative. These tendencies are now reversing. A carefully balanced portfolio should contain both “risky” and “safe” assets – and a selection of each.
Residents of the United States, please read this important information before proceeding
“Risk on”/“risk off” – and why it happened
The systemic nature of the financial crisis that erupted in late 2008 meant that few regions or sectors in the economy were unaffected. Most business and investment transactions use finance, and when working capital dried up, cash and government promises were king. And as the saying goes, in a crisis you don’t sell what you should, you sell what you can: the urgent need for liquidity – to meet margin calls, or to repay borrowed funds – further caused seemingly unrelated assets to move together. (Note: Barclays’ Investment Philosophy advises against selling at times of turmoil, and does not recommend leveraged investment portfolios as the norm.)
Predicting the best-performing asset class each year is difficult and the cost of getting it wrong can be high
Amongst other things, this resulted in some stomach-turning falls even in defensive stocks. Blue-chips with relatively stable earnings tend to be amongst the most liquid investments, and so are easiest to sell to raise cash (in marked contrast to the illiquid, toxic and opaque credit securities that helped create the mess to begin with).
So in the circumstances, it wasn’t such a surprise to see a failed US investment bank – and the near-failure of a big US insurer – hitting global developed and emerging market stocks, and speculative grade credit, and commodities, and many other risky assets, together. Correlations among riskier assets – developed stocks, emerging stocks, commodities, high-yield credit and emerging market bonds – rose markedly (Figures 1 and 2 – a level of one would indicate complete synchronisation).
Correlations within stock markets – across countries and especially across sectors – also rose during the crisis, to historically-high levels.
Similarly, as the scale of the problems became clear during 2007 and 2008, and the possibility of a major economic downturn if not a second Great Depression seemingly loomed, it was unsurprising to see government bonds do well. Government
creditworthiness in the last resort trumps that of the private sector, and falling interest rates – and eventually, direct central bank buying – boosts bond prices. The correlation between stock prices for example and government bonds became strongly negative, and has mostly stayed so since (Figure 3).
To some extent, this negative correlation reflected the actual movements of funds, as investors switched from stocks to bonds. Of course, not everyone sold their risky assets, or reinvested the proceeds when they did, and prices can and do move even without any associated flows of funds.
Cash, the nominally riskless asset in our framework, also did better than risk assets too, simply by marking time. As an effectively static asset, however, cash was not correlated with other assets to begin with, and its portfolio behaviour couldn’t change.
Short-term correlations among risky assets rose markedly from 2008
The combination of rising correlations amongst and within “risky” assets on the one hand, with falling correlations between “risky” assets and safer government bonds (and to some extent investment grade credit) on the other, had the effect of muting the overall change in intra-portfolio correlations. Indeed, when allowance is made for the strategic weighting of each asset class in balanced portfolios, cross-asset correlations actually fell slightly (Figure 4). This of course takes no account of the scale of the respective moves – stocks are almost always more volatile than bonds, partly because they have longer duration – and fell by more than bonds rose – but it suggests that received wisdom about the crisis is wide of the mark. It became more important, not less, to have a carefully-diversified portfolio – cross-asset correlations actually fell for a while.
Is normal service being resumed?
As you may already have seen from Figures 1– 3, things have moved on. Correlations across “risky” asset classes have been falling back of late, perhaps most strikingly in the case of emerging stocks and commodities (Figure 2). Sub-asset class correlations too have been falling back – Figure 5 shows country and sector correlations within developed stocks, and the correlations of the main sub-asset groups within commodities. The correlation between developed stocks and government bonds – the key relationship in asset allocation, and arguably in investment management more widely – has rebounded dramatically (Figure 3).
Investment conclusions: many eggs still need many baskets
The most dangerous phrase in investing is said to be “It’s different this time”. Markets are driven partly by emotion, by politics and by natural disasters: there will be other crises. We may hope they won’t be as systemic as 2008’s, but history shows that our ability to innovate – to create new opportunities for profit and loss – exceeds our ability to regulate. When the next crisis hits, asset correlations within a portfolio will shift again. We suspect crises are not different this time – they’re subtly different every time.
Meanwhile, the case for diversification is no more hurt by the runaway performance of developed stocks in 2013 (so far) than by the stand-out rise in bond prices in 2008. Investors would of course have done best this year by owning nothing but developed stocks – just as they’d have done best by owning only government bonds in 2008. Hindsight has 20:20 vision. Unfortunately, our ability to predict the winning assets each year is not what we’d like. We give it our best shot, but the changes in the asset classes topping each annual column in Figure 6 reminds us how difficult it is to pick the winners – and the winners only. Diversification is a little like house insurance: peace of mind is worth having, even if your home doesn’t burn down
Falling correlations within risky assets can influence the way in which investors express their views on markets. If sectors and countries are moving less closely together within developed stock markets, for example, the potential ability of fund managers to add value by actively selecting stocks and sectors might rise by comparison with a world in which most sectors and countries are rising together. In the less correlated world, investors may be more inclined to opt for an active fund; in the latter, opting for a passive fund that simply tracks the wider market may perhaps be the obvious thing to do. That said, currently we think that there is still sufficient headroom in developed stocks for passive implementation approaches to continue to deliver attractive risk-adjusted returns from developed stocks.
A final thought. Stocks have the biggest weights in most balanced portfolios, and are one of the most volatile assets. The interaction between stocks and government bonds is probably the most important driver of overall portfolio returns and volatility. Their correlation has rebounded sharply – from being very negative, to being almost positive (Figure 3) – and may rise still further in the months and years ahead. This will likely reflect a further unwinding of crisis-induced flows, and on a very long-term basis we see the “natural” correlation between stocks and bonds as more likely to be positive, as in the more distant past, than negative. Bond yields are an input to many equity valuation models, and lower yields favour higher stock prices, other things being equal. This is not unhealthy: on a long-term view, stock prices trend higher with rising corporate earnings, and a sustained negative correlation with another large asset class would be a drag on returns. There is still plenty of room for risky assets’ correlations to fall back further.