(I) Bond/equity correlation:
At our recent quarterly investment offsite, an interesting debate arose on the correlation between bonds and equities. One participant suggested bonds and equities in the US had been positively correlated in recent years, while another was firmly of the view that they had been negatively correlated. How could they disagree about historical facts? Who was right?
Both, in actual fact. They were each talking with a different time horizon in mind. Perhaps surprisingly, this can amount to the opposite conclusion even when considering historical data.
Figure 1 plots the total return of the S&P 500 stock index against the performance of long-dated US Treasuries over the past 10 years. Over short periods of weeks or months, the two lines regularly seem to move in the opposite direction – a negative correlation.
As growth risks predominate (as in the 2008 financial crisis or early this year, bonds have performed well and equities, poorly. The reverse is true as growth expectations pick up, as in 2009. Measured using rolling daily or weekly correlations, bonds and equities have clearly been negatively correlated on average over this period.
But what concern is this to an investor with a multi-year investment horizon such as a 40-year old saving for retirement? If we take the 10-year period, both bonds and equities have both provided (very similar) substantial returns and handsomely outperformed cash: a positive correlation.
When considering the overall period, in particular post-2008, an absence of inflation along with very low interest rates (and other central bank policies such as QE) has supported strong returns from both bonds and equities.
So correlation is not always as straight forward a concept as it might seem and we need to consider the time horizon we are referring to when talking about it.
(II) Correlations are not constant
As noted, real interest rates have been at historically low levels for a prolonged period now and this has been an important factor behind the strong excess returns from bonds and equities over cash in recent years.
If this policy backdrop were to change significantly over the coming years then it will have profound implications for returns on financial assets. On a multi year view, a material rise in real policy rates (say 300-400bps for example) would have a very negative effect on many mainstream government bond market returns given current valuations.
We might have less certainty over how equities would fare, as such a monetary tightening would only occur in a strong nominal growth environment, but reasonably we might expect them to struggle in the US from current valuations, certainly over periodic phases.
In this hypothetical scenario, the long-run correlation may remain positive with returns relative to cash driven by the direction of real policy rates (and probably inflation risk premia), while the effect on short-term correlation may change to turn positive as well, or perhaps oscillate in an extremely volatile fashion.
Figure 2 highlights the changing nature of correlation over time.
Much of the period until the mid-1990s provided a very different correlation environment for bonds and equities, when returns were positively correlated. This was particularly true in the 1970s, where returns for both assets were driven by big shifts in inflation trends.
For the past 20 years, the shorter-term correlation has mostly been negative but there have been periods when the correlation has reversed into positive territory as during the 2013 taper tantrum and on occasion more recently (as shown in figure 3, which shows weekly correlations).
The chart makes is clear that correlation patterns are not constant through time, with important implications for asset allocation and portfolio construction.
(III) The nature of risk
We are firmly of the view that the real risk investors should care about is a prolonged (or permanent in the strictest sense) impairment of capital and not short-term volatility. Indeed, this philosophy is central to our Episode approach to investing.
James Montier shares this view but also provides a useful classification of risk (in the permanent sense) as coming from at least one of three sources: (i) operational or business risk; (ii) financial/balance sheet risk; or (iii) valuation risk.
The first two categories clearly apply to equities but should not apply to sovereign bonds in countries with their own central bank (N.B. as the euro crisis showed, re-financing debt is a risk if you don’t have a lender of last resort), at least in nominal terms.
However, the third classification applies very much to all assets, including bonds: the price you pay very much alters the risk properties of the same asset.
In this way, it should be clear that statistical measures of risk – and perceptions of risk – taken from recent experienced price behaviour are often likely to be diametrically opposed to insights provided by consideration of value and the nature of (permanent) risk that we should be concerned about.
The recent sell-off in long-term G7 bonds is illustrative. In recent months, valuations of long-dated bonds in Europe, Japan and the UK in particular became very extreme in terms of the path of future events required to justify the prevailing market price.
Thirty or forty years ago, bonds were perceived to be incredibly risky. The prevailing sentiment accorded those assets high real rates of return (redemption yields). Subsequent returns were even more spectacular than expected as inflation and real interest rate structures declined more than anyone imagined.
Today, most mainstream government bonds are perceived to be safe, supported by central bank and pension fund purchases and a capitulation of thought in favour of theories of secular stagnation. The conclusion is simple: not only are expected returns extremely low but some government bonds have become much riskier too.
The value of investments will fluctuate, which will cause prices to fall as well as rise and you may not get back the original amount you invested. Past performance is not a guide to future performance.