How worried should we be by a possible sell-off of gilts? It’s easy to tell stories of how this might happen. Reversing quantitative easing and rising interest rates in the US (and possibly the UK) might upset the market, as might looser fiscal policy here. And it’s possible that in both the UK and US tighter labour markets will eventually raise inflation: just because something hasn’t happened so far doesn’t mean it won’t happen.
Let’s, though, take a different perspective and look at the numbers on gilt risk.
Since 1990 the standard deviation of total returns on gilts has been 1.8 percentage points per month, with average returns being 0.6 per cent per month. This implies that we face around a one-in-six chance of losing 1.2 per cent or more in a month or of seeing some kind of loss over 12 months. It also means there’s around a 2.2 per cent chance of losing 3 per cent or more in a month.
There are, though, two complications here. One is that gilt returns aren’t exactly distributed normally. Instead (as with other assets such as shares and commodities) there is a higher chance of big losses. If gilt returns are distributed as a cubic power law for large moves, there’s around a 0.7 per cent chance of losing 5 per cent or more in a single month. (5 per cent is in fact the worst monthly loss since 1990).
Secondly, the standard deviation since 1990 might understate gilt risk. Volatility was very low in the mid-2000s, but has been trending upwards since then. A big reason for this is that as yields fall, prices become more sensitive to given changes in yield.
This suggests we can’t exclude the possibility of significant losses on gilts simply as a statistical fact.
Nobody, however, should own gilts alone. Instead, they are part of a balanced portfolio. Which poses the question: could losses on gilts be offset by profits on other assets, in which case they are not so risky after all?
Correlations with gilt returns
Five-year rolling correlation of monthly returns
A quick glance at the historic numbers suggests so. Since 1990, the correlation between monthly returns on gilts and on the All-Share index has been 0.16, and the correlation between gilts and gold (in sterling terms) has been 0.05. These numbers imply that there’s almost a 50:50 chance that a loss on gilts will be accompanied by a rise in shares or in gold.
But this also means there’s a slightly greater than a 50:50 chance of us losing on two assets at once.
What matters is why gilts sell off. If they do so because investors become more willing to take risk or more optimistic about economic growth (two things that are hard to distinguish), then losses on gilts would very likely be offset by gains on shares. Similarly, if gilts fall because of fears of worldwide inflation, it’s possible (but by no means certain) that their losses would be accompanied by profits on gold.
But there’s another possibility. A gilt sell-off could be triggered by a decline in global liquidity. One way in which this could happen would be if we see a big reversal of QE and significant portfolio rebalancing effects: if investors need to buy gilts from the Bank of England or Treasuries from the Fed, they might – at the margin – sell equities to do so.
Personally, I suspect this is a low risk. Something else, though, does trouble me. It’s that correlations between gilts and both equities and gold have been rising recently. This tells us that shocks that cause gilts to move in the same direction as either have become more common since the mid-2000s.
In the case of gold, this could be because changes in inflation expectations have been quite small while moves in real interest rates have been significant: changing views on inflation would cause gilts and gold to move in opposite directions, but changes in real interest rates would cause them to move in the same direction.
In the case of equities, it might be because changing attitudes to liquidity have been more significant than changes in appetite for risk.
Whatever the reason, though, the implication is the same. There is a significant danger that losses on gilts will indeed be accompanied by losses on other assets.
There are two general messages here. One is that we shouldn’t take simple historic numbers on volatility and correlations at face value. Both change over time, so such numbers might understate future risk.
Secondly, all this is yet another reason to hold cash, despite negative real interest rates (for now). One of its virtues is that it protects us from correlation risk – the danger that asset prices will fall at the same time. This danger might be small, but it is sufficiently nasty for us to guard against.
