Should we really be worried about an inverted yield curve?
The shape of the yield curve in the bond market — usually proxied by the 10-year government bond yield, less a short-term policy rate — has been used as a lead indicator of recession risk in the US and other developed economies for several decades.
In fact, this one simple indicator sometimes seems to exert a mesmeric influence over investor sentiment about future economic activity. While this degree of credence is far too simplistic, inversions in the curve have certainly preceded many previous recessions, and it is dangerous for investors to ignore them.
Policymakers and academic economists also pay close attention to the information contained in the yield curve. The recent inversion in the US curve is the first since 2007, and it will be high on the agenda at the Federal Open Market Committee meeting next week. Opinion on the committee about the usefulness of the curve is somewhat split, but there is probably a majority of members in favour of taking the latest signal seriously.
In the US, whenever the yield curve has inverted in the past 60 years — with only one exception in the late 1960s — a recession has followed. That is some record! Even more remarkably, the yield curve seems to have offered guidance about future economic activity since the 1850s. It has also worked in more recent cycles in many other economies.
Despite this remarkable history, it is not fully understood why the curve tends to invert before recessions. Most economists point to the fact that the curve systematically flattens and eventually inverts during periods of monetary policy tightening. When short-term policy rates rise, expected future short rates, and bond yields, are more stable, because they reflect the market’s view of the underlying equilibrium interest rate, which changes only very slowly.
Accepting that the yield curve is generally a good indicator of the stance of monetary policy, then it is not too surprising to discover that it contains leading information about economic activity. This is just a reflection of the normal lag between policy and the output response that appears in most macro models of the economy.
The Fed routinely uses the yield curve to calculate the probability of recession 12 months ahead. One of its most important models is that published by the New York Fed, shown in the box below. This uses “probit” estimation techniques that relate the yield curve slope at any given time to the appearance (or not) of recessions 12 months later, as determined by the National Bureau of Economic Research recession dating committee.
Although the statistical significance of these methods is inevitably limited by the fairly small number of recessionary episodes, the regularity of the inversion of the yield curve ahead of recessions is very striking. The latest calculation shows that the yield curve slope in early June implies a probability of 29 per cent that a recession will occur sometime within 12 months.
There are, however, reasons for questioning whether today’s recession risk is exaggerated in this calculation.
Long bond yields represent the expected forward path for short rates plus a risk premium, or term premium. Because of reduced inflation risk, this term premium has dropped by around 1.5 percentage points since the early 2000s, which implies that the yield curve is more inverted, for any given path for expected short rates, than it might have been in previous eras. Research has not clearly established whether this drop in the term premium reduces the reliability of the recession signal from the yield curve.
Separately, the Fed staff also calculate recession probabilities from the so-called excess bond premium in the credit market, a measure of investor sentiment unconnected to bankruptcy risk. On this model, the latest recession risk is only 13 per cent. Although quite fashionable, this model has not been around for very long, and its main test in real time came in 2016, when a large widening in credit spreads caused it to predict a recession risk of 60 per cent. Following a dovish shift in Fed policy, no recession ensued.
A third approach that is supported by some FOMC members relies on the impact of the yield curve on the supply of credit in the economy, especially from the shadow banking sector. Research published in 2010 by the New York Fed indicates that an inverted curve reduces the margins earned on new credit creation by financial intermediaries, which reduces their incentives to expand their balance sheets through additional lending. The resulting reduction in credit supply leads to elevated recession risks.
In summary, FOMC members will enter their crucial policy meeting next week with their yield curve and credit spread models indicating that recession risks are running at 13-29 per cent, which is somewhat more elevated than normal. There may be mitigating factors, connected to the unusually low term premium, which might be distorting these signals. Nevertheless, the committee would be unwise to ignore the information in the yield curve, and is quite unlikely to do so.
The yield curve and recession probabilities in the US
The Fed publishes different models that assess the probability of recession within 12 months; while the New York Fed publishes a widely-followed model based on the “term spread”, or slope of the yield curve. The latest estimate shows a recession probability one year ahead of 29 per cent.
An alternative model that has become fashionable recently is based on “excess” credit spreads in the bond market. These recession risks are updated monthly by the Board’s staff. The latest result shows a recession probability of 13 per cent.
It is possible that the yield curve models are being distorted by the very low term premium, or risk premium, that currently exists in the long end of the bond market.
This may have resulted in lower long bond yields, and, therefore, a more inverted yield curve, than might have existed in previous eras, notably before about 2002. That might make the current yield curve inversion less worrying than it seems on the surface.