Is The Yield Curve Still A Reliable Signal Of Recession Risk?

In the current debate about recession risk, some commentators have rolled out the yield curve argument or variations thereof. On first glance, this line of analysis looks like a slam-dunk refutation of the forecast by some that another economic contraction is now fate. But such arguments based heavily (or exclusively) on the yield curve risk overplaying their hand. It’s true that the yield curve has been a reliable predictor of recessions for half a century, as many studies assert. Indeed, the literature on this topic is now quite extensive and persuasive. But in the dark art of developing macro forecasts, one can never assume that a predictor’s track record—even one as strong as the yield curve’s—seals its fate for repeat performances. It’d be wonderful if we could point to one indicator as a dependable predictor, but macro’s just not that simple.

But let’s back up for a minute and recognize the yield curve’s allure. It’s well known that when short rates rise above long rates, that’s been a strong signal that the economy was headed for a recession in the near term. This track record is quite clear by reviewing the yield spread for, say, the 10-year Treasury less the 3-month T-bill. For additional clarity, this signal can be transformed with a probit model so that the spread is converted into a probability estimate of future recession risk.
For example, using the procedure for parameter estimates outlined in a 2006 New York Fed paper (“The Yield Curve as a Leading Indicator: Some Practical Issues”), I calculated recession risk in an Excel spreadsheet for the 12-month-ahead time horizon based on monthly yield data history for the 10-year Note and 3-month T-bill (see “The Probability Model” sidebar on p. 3). Here’s the resulting chart, which is updated through the end of 2011:

Clearly, the history of the yield curve as a predictor of recessions is quite good. No wonder, then, that if you create a model that piggybacks off of the yield curve in some degree you’re going to generate similarly favorable results these days. There’s virtually no recession risk via the yield curve model at the moment given that the spread was positive by 186 basis points as of January 13 (i.e., the 10-year yield was 1.89% and the 3-month T-bill was just 0.03%). That’s a relatively high reading in the context of the last 50 years and it suggests that a new recession is nowhere in sight.
The problem is that it’s unclear if the yield spread indicator has been compromised by the unusual run of monetary policy in recent years. The Federal Reserve has reduced short rates to virtually zero since late-2008 and it appears that zero will prevail for the foreseeable future. Under those conditions, an inverted yield curve isn’t possible.
It’s an open debate if the yield curve remains a viable predictor of the business cycle under the current conditions. In particular, the question is whether the economy can still contract when short rates are at zero? Only time will tell. Meanwhile, a simple reading of history suggests the odds are quite low. But even under the best of circumstances, relying on one indicator is dangerous. At some point, every predictor fails. So far, the yield curve appears to have sidestepped that fate, but it’s unlikely that it will shine indefinitely. Whether it’s failing now is unclear. In any case, it’s folly to put too much weight into the yield curve (or any other predictor) for estimating recession risk.
A more robust model is one of combining multiple predictors and generating probit models for each to estimate the outlook for the economy. Ideally, the pool of predictors will represent a diversified mix of economic dynamics to minimize redundancy and the risk of failure. It’s reasonable to anticipate that one or two predictors will stumble at some point, but it’s unlikely that all will fail simultaneously (assuming you choose a truly diversified mix). Over time, it’s likely that we’ll see different predictors fail at different times. Fortunately, there’s no shortage of useful factors to consider to manage this hazard. If anything, there’s too many predictors, which is why empirical testing keeps econometricians busy.
The good news is that there’s a fair amount of corroboration from other predictors for the yield curve’s current prediction that economic growth will continue. Nonetheless, some analysts disagree. But any forecaster who’s relying on the spread alone for expecting that the modest expansion will roll on may be courting trouble. It’s premature to conclude that the economy is doomed for another round of contraction, but the odds that we’ll muddle through are well below the near-certainty levels implied by the yield curve alone.