Does a downward-sloping yield curve predict a recession?

Charles Goodhart

24 September 2007



A downward-sloping yield curve has commonly been used as a leading indicator of a future recession. For much of last year long rates of interest were slightly below short rates in the US and elsewhere, although recent events have reversed this. Some saw the inverted yields curve as a portent of a coming decline in growth, and may have adjusted their portfolios accordingly. Yet growth has only fallen a little in the US, has strengthened in the euro-zone, and has remained resolutely stable in the UK.

In current research1 I, and my colleague Luca Benati of the ECB, have investigated why the yield curve has appeared to have had such predictive power for future output growth at times in the past, but also why this may now have largely disappeared. We examined the relationship between output, inflation, short term interest rates, and the spread (i.e., difference) between short and long rates, for the US and the UK since the Gold Standard era, and for the Eurozone, Canada and Australia in the Post-WWII period.

In most previous studies of this kind, the relationship between the relevant variables has been assumed to remain constant over time, or has been tested for occasional once-and-for-all breaks. In contrast, we allow the relationships among the key variables to adjust over time, so we can pinpoint more accurately just when the spread had marginal predictive content for output growth, (both in a bivariate, spread/output, and in the multi-variate, four variable case).

Our results suggest that, historically, the additional predictive power of the spread for future output growth –over and above that already encoded in other macroeconomic variables – often appeared during periods of uncertainty about the underlying monetary regime. This is true, for example, of the US during the Volcker disinflation episode. Why might this be the case? If the monetary policy regime is known, and believed to be stable, agents should probably be able to deduce much of the path of future short rates from the trio of (current and past) short rates, output growth and inflation. As a consequence, the marginal predictive content of the spread (MPCS) for future output growth should be zero, because the informational content of the spread moves closely in synch with that of the other variables. And indeed, under the Classical Gold Standard (1879-1914) – the stable monetary regime par excellence – the spread did not exhibit any additional predictive power both in the United States before 1910 (after which time debate over the creation of the Federal Reserve System started, thus injecting some uncertainty about the monetary regime), and in the United Kingdom.

On the other hand, when the monetary regime becomes (at least partly) uncertain, an increased risk premium will have to be added to expectations of future short
rates. According to this view, part of the cause of the subsequent output decline is not that the yield curve is negatively sloping, but that it is insufficiently so, i.e. long rates are above those consistent with the subsequent resolution of the uncertainty, thereby imparting additional downwards pressure on output. The following are some examples in which historical evidence is compatible with this hypothesis:

• United States: 1910-1913, debate over creation of the Federal Reserve System; aftermath of the abandonment of the gold parity in April 1933; 1979-1982, Volcker disinflation episode.

• United Kingdom: 1990-1992, ERM membership, ejection from the EMS, and introduction of inflation targeting.

• Canada: 1979-1981, connection with the US; 1989-1991, introduction of inflation targeting, and conflict between John Snow and the Liberal Opposition.

While we find this hypothesis attractive, there does, alas, appear to be a counter example in our results. In the UK, Canada, and the US there appears to have been some increase in the MPCS in the early 2000s, at a time when the monetary policy regime had been successfully established. Remember, however, that this was a period of low short-term interest rates, upwards sloping yield curves and continuing growth. If our analysis is correct, the implication is that some external force(s) were holding down long yields – compared with the levels that would have obtained based on the information encoded in short rates, inflation, and output growth – and thereby imparting stronger growth. International influences, notably affecting the balance of world saving and investment, are one possibility. Abundant liquidity, ‘climbing the yield curve’ in search for return, is another. But an attempt to establish this must remain for further research. So our conjecture is that the yield curve will only contain unique predictive power for output growth when long rates cannot themselves be reasonably well forecast from current, and prior, developments in short rates, inflation and output. This is most likely to happen at times of uncertainty about the nature and efficacy of a new (and untried) monetary policy regime.




1 ‘Investigating time-variation in the marginal predictive power of the yield spread’, forthcoming in the Journal of Economic Dynamics and Control, available at website .





Topics:  Monetary policy

Tags:  interest rates, inflation, yield curve, output growth

Emeritus Professor in the Financial Markets Group, London School of Economics