The yield curve – a predicator of recession?

12 Feb 2019

We can all agree that the ability to accurately forecast whether the economy is about to slide into a recession is a valuable commodity. If you can predict a recession coming early enough, then you can reallocate your portfolio accordingly and position yourself in advance for the cross-asset class correction in prices that tends to occur during a downturn.

However, it may surprise you to learn that many investors and even some policymakers believe that the yield curve is the best place to look for a reliable gauge of the prospects of recession. In fact, there is academic literature which dates back at least two decades which has demonstrated the predictive power of the yield curve slope – with the basic message that when the curve inverts, it is time to worry. Indeed, research recently published by economists at the Federal Reserve Bank of San Francisco concluded that:

The term spread—the difference between long-term and short-term interest rates—is a strikingly accurate predictor of future economic activity. Every U.S. recession in the past 60 years was preceded by a negative term spread, that is, an inverted yield curve. Furthermore, a negative term spread was always followed by an economic slowdown and, except for one time, by a recession

We don’t dispute the facts, and we don’t think it’s unreasonable to worry a little more about recession risks today than you might have done a year ago. However, we believe that the faith placed in the slope’s forecasting prowess in some quarters is inappropriate.

 

Assessing the yield curve’s credentials

A sensible place to start is with an objective assessment of the slope’s credentials, which means locating the particular expression of the yield curve slope that theory suggests should have some predictive power and then testing the relationship between that particular measure of the slope and the incidence of recession several quarters in the future – as opposed to running a battery of tests to uncover the spread between any two points on the yield curve that appears to best explain the incidence of recession.

What is interesting is that while it is possible to make arguments – although not necessarily altogether compelling ones – that the slope of the yield curve at the front end of the curve or movements in particular drivers of the term premium which influence the slope at the back end could send signals about the risk of recession, it is not at all obvious what the precise mechanism is that justifies looking at the spread between the three-month and 10-year yield, which is supposed to be the best-performing recession indicator.

It is possible to construct a financial instrument that should accurately portray market expectations of the prospects of a recession. The value of an exotic binary option that pays out in the event of a recession at or before some specified date and zero otherwise should convey that information. However, the returns on that hypothetical instrument bear no relation to those on the government bonds from which the yield curve is derived. Bond prices do respond to economic fundamentals, but there are no grounds for believing that there should be a stable relationship between percentage point changes in the perceived risk of recession and a given move in bond prices that delivers a given move in the slope.

 

The yield curve – a flawed mirror?

In practice, we believe that the yield curve acts like a flawed mirror, offering an imperfect reflection of those fundamentals with a recession signal shrouded in noise that captures the other factors that are relevant to the fair value of bonds, but contain no information about economic prospects.

We think it makes more sense to look at those fundamentals directly – to analyse the data and macroeconomic forecasts – if you want to assess the risk of recession rather than rely on their imperfect reflection. Indeed, we think that the only reason that you would rely on the slope as a recession indicator is if you believed that investors were far better at “doing economics” than economists are.

Even then, it ought to be the case that market prices cannot perfectly reveal the views of these superior forecasters because otherwise there would be no incentive for investors to spend precious resources on constructing those superior forecasts.

 

Bond prices – determined globally rather than locally

There are theoretical reasons to question the relevance of the signal currently being sent by the yield curve slope that reflect both the causes and consequences of the current curve shape.

If you want to put your faith in the slope as a recession indicator, then you need to believe that the shape of the yield curve responds to local fundamentals – that is, developments in the local economy.

However,  even before the financial crisis, it was increasingly apparent that the prices of the highest quality advanced economy sovereign bonds were increasingly co-determined, driven by common global factors, facilitated in part by institutional investors that can participate in many markets.

The fact that many of the world’s largest central banks have accumulated huge portfolios of these bonds through quantitative easing (QE) programmes since then has likely further elevated the importance of those common global factors. Local QE has ended up having a global impact on bond prices. If local bond prices are determined at the global level, it follows that the local slope of the yield curve is unlikely to send reliable signals about the local economy.

 

A negative feedback mechanism embedded in the yield curve

Moreover, the yield curve’s relationship with the economy embeds a negative feedback mechanism that works to undermine the signal the slope sends.

Central banks have always relied on the fact that lower yields stimulate spending. Indeed, the relationship between short to medium-term yields and spending is integral to the transmission mechanism of a traditional cut in the policy rate; the relationship between long-term yields and spending is integral to the transmission mechanism of QE.

So when medium- and long-term yields fall, sending a signal that the risk of recession has increased,  the yield curve is actually injecting a stimulus into the economy that makes a recession less likely.

 

It’s the macroeconomic fundamentals that matter…

Our point is not to argue that there is no information in the yield curve slope. The stylised facts quoted above clearly suggest otherwise. We simply claim that the yield curve is a flawed mirror and you are better off looking at fundamentals themselves rather than their imperfect reflection if you want to understand where the economy is heading.

As a matter of fact, we are perfectly willing to concede that there are risks on the horizon, although we humbly submit that the US economy is not the obvious place to start worrying about a recession, no matter what the US yield curve might imply.