Focus
September 27, 2024 | 13:14
The Yield Curve… Again
Economics
Economics
Economics
Economics
Economics
Economics
Economics
Economics
Economics
Economics
Economics
Economics
Economics
Economics
Economics
Economics
The Yield Curve… AgainThe yield curve inversion garnered plenty of headlines and warnings of a U.S. recession. Now, it looks like a false positive as the economy steers a soft landing. |
|
In recent weeks, two-year Treasury yields have been closing consistently below 10-year yields with September poised to average a positive spread (10y-2y) for the first time since flipping negative, or inverting, in July 2022. This un-inversion or ‘reversion’ of the yield curve is said to have recession-signalling ability akin to the curve’s initial inversion 26 months ago. Below, we assess how well the yield curve has done in predicting economic downturns, via both inversions and reversions (Chart 1). Inversions revisitedPersistent yield curve inversions (i.e. 10y-2y averaging a negative value for at least one month) have occurred before the first month of the past five pre-pandemic recessions. The signal ranged from 11 to 23 months ahead, for an average of around 17 months (Table 1). So, once the curve inverted in July 2022, the prediction was for a recession starting December 2023 and no later than June 2024. We’re still waiting for the downturn. Not only did real GDP grow at a 3.0% annualized rate in Q2 (and 3.0% y/y), but most of the six key monthly indicators employed by the NBER to date business cycles are still at their cyclical peaks. That includes nonfarm payrolls, real personal income less transfers, real PCE and real business sales. Admittedly, household employment peaked in November 2023 and is down a net 0.3% over the past nine months, while industrial production peaked in September 2022 and has slipped a net 0.4% over the past 23 months. On balance, the economy appears to be far from an overall recession, arguing the recent inversion’s signal was a false positive, i.e., a curve inversion not followed by a downturn. |
|
Historically, there has been only one other false positive signal. In June 1998, there was the briefest and near-smallest possible inversion, just one month averaging -2 bps followed by zero in July. Meanwhile, there have been no false negatives for the 10y-2y metric, i.e. a recession that was not signalled by an inversion. Although listed in the tables, we do not include the 2020 pandemic-driven recession in our assessments of the yield curve’s downturn-signalling ability. All told, yield curve inversions (10y-2y) sport a 71% success rate in predicting recessions since 1980, including the latest episode. However, 2-year notes started being issued only in 1976. For a longer and/or corroborating assessment of the yield curve’s effectiveness in signalling contractions, we look at the spread between 10-year yields and 3-month T-bill rates (on a coupon-equivalent basis). Currently, the yield curve defined by these nodes (10y-3m) is still inverted at -82 bps. Interestingly, the NY Fed’s recession probability model is based on this metric, with last month’s average inversion (-132 bps) implying 62% odds of a recession in the year ahead. Of course, this will be marked down owing to the Fed’s rate cut. The key takeaway from the model… the greater the degree of inversion, the higher the risk of recession. So, looking at the 10y-3m spread, an inverted curve has occurred before the first month of all seven pre-pandemic recessions since 1969 (Chart 1 again). The signal ranged from six to 17 months ahead, for an average of 12 months (Table 1 again). In the current episode, this metric inverted in November 2022, later than the 10y-2y metric because three-month tenors are more influenced by contemporaneous policy rates. The consequent prediction was a recession starting November 2023 and no later than April 2024. These dates were only a month or two away from what was heralded by the 10y-2y spread. But, like it, this too was a false positive signal (inversion but no recession). Historically, there has been another false positive in 1966 along with two false negative signals (no inversion but recession) in 1957 and 1960. Unlike the 10y-2y metric, it didn’t invert during 1998 (on a monthly average basis) but it did invert during 2019, which we are not counting owing to the pandemic. As such, the 10y-3m spread has a 64% success rate in predicting recessions since 1957 including the latest episode, with a 71% success ratio since 1980 (in line with 10y-2y). Reversions introduced |
Returning attention to the 10y-2y spread, persistent yield curve reversions, in which this metric averages a positive value for at least one month after an inversion, have occurred before only three of the last five pre-pandemic recessions. On those three occasions, the signal ranged from three to 10 months ahead for an average around seven months (Table 2). Interestingly, in the other two episodes in the early 1980s, reversion occurred within three months after the start of the downturn. Reversion has also had one false positive signal, when the yield curve returned to a positive slope after an inversion’s false positive signal in 1998. This results in a success rate of just 50% (not including the latest episode) as a leading recession indicator. Of course, the success ratio doubles focussing on the three most recent downturns, but this is a rather small sample. For the record, with reversion unfolding in September, this predicts a recession starting next April and by July 2025 at the latest. We’ll jot it down in our calendar just in case. |
|
For the 10y-3m yield curve spread, reversions have occurred before only three of the past nine pre-pandemic recessions. On those three occasions, the signal ranged from two to seven months ahead with an average of five months (Table 2 again). Also, in the four preceding episodes, reversions eventually occurred, nearly all within three months after the start of recession. Meanwhile, there was one false positive signal in early 1967 after an inversion’s false positive in late 1966. It did revert during 2019 after inverting but, again, we’re not counting that one owing to the pandemic. In total, reversion of the 10y-3m spread has a 30% success rate predicting recessions but, like the 10y-2y metric, it’s 3-for-3 in the small sample of the last three downturns. This time is differentIn the pre-pandemic period, the yield curve’s power to predict recessions was strong since 1980 for inversions and since 1990 for reversions (forgiving the latter’s dismal pre-1990 performance). However, 2022’s curve inversions pointed to recession by 2024Q2, at the latest, which hasn’t happened yet. We reckon the now-unfolding curve reversions (10y-2y this month, 10y-3m likely by early next year) face a similar ‘false positive’ fate. This is not completely surprising. Since the last ‘true positive’ signal during 2006-2007, a series of developments have inherently flattened the yield curve, making it more prone to inversions and false positive signals. First, the long-run neutral policy rate, or ‘r-star’, has fallen meaningfully over time. For example, when the FOMC first began publishing its projection of the longer-run neutral rate in 2012 the median was 4.25%, which was in line with historic norms. The latest median published last week was 2.875%. Although up 50 bps from its low in 2022, it remains well below what it was at least a dozen years ago. To the extent the neutral policy rate guides market expectations of the profile for short-term interest rates over the coming decade, this major component of 10-year yields will be commensurately lower. Second, the Fed’s adoption of a formal inflation target in January 2012 helped shrink the ‘all-in’ inflation expectations that are embedded in long-term bond yields, both the expected level (now better anchored at 2%) and the perceived risk surrounding it. Third, the Fed’s quantitative easing (QE) efforts that began in the wake of the Global Financial Crisis started including Treasuries in March 2009. These large-scale asset purchases were skewed to longer-term maturities with the intent to pare the term premium and thus reduce longer-term yields relative to shorter-term rates. Although quantitative tightening (QT) efforts reduced the Fed’s portfolio of Treasuries, it remained vastly larger than it was pre-QE (mirroring the shift from a scarce reserves system to an abundant reserves regime). Moreover, QT was conducted via not reinvesting the proceeds from maturing securities, not by outright sales, keeping a longer-term skew to the Fed’s portfolio from fading too quickly. All told, this has likely exerted some persistent flattening pressure on the yield curve. Bottom Line: The yield curve may no longer be as dependable a recession indicator as it used to be. But it’s still worth watching. |