As prices continue to soar and the Russia-Ukraine conflict makes inflation pressures even worse, the Federal Reserve is sharpening its tools to fight the run-up in prices. And now news that the 10-year/two-year curve inverted has led to calls that a recession is around the corner.

Although uncertainty over the economic outlook has increased as inflation sits at a 40-year high, we don't view the inversion of the yield curve, on its own, as signaling any near-term danger.

Indeed, our augmented term spread model of recession risk in April, using the 10-year/three-month (10Y/3M) curve, has a 6% chance of recession out 12 months. And our term spread indicator in our dashboard of leading growth indicators signals positive momentum so far. This is in line with our forecast for 3.2% economic growth this year and an unemployment rate, already at 3.6%, expected to reach its precrisis rate of 3.5% in second-quarter 2022.

Chart 1

We don't currently believe the inversion is a recession signal for several reasons:

Only one part of the curve, the 10-year/two-year (10Y/2Y), inverted at the start of April, the first time since August 2019, but only for two days.  Moreover, other traditionally followed term spreads, like the 10Y/1Y, 10Y/3M, and near-term forward rates, all remained in positive territory. (At the National Association for Business Economists on March 21, Fed Chair Jerome Powell, perhaps to allay inversion worries, boasted the near-term forward yield spread, which was far from inversion, was more accurate in predicting recessions than the others.) Combined, this gives mixed signals on the predictability of the inversion of the 10Y/2Y curve.

Looking back to 1982, we found that the daily signal falsely predicted recession numerous times depending on the recession indicator used (see table 1).  For example, the 10Y/2Y curve falsely signaled a recession 153 times over a 12-month horizon, 29 times the next 12-18 months, and six times over an 18-24 month horizon. Even the near-term forward curve falsely signaled a recession 40 times over a 12-month horizon, 30 times over the next 12-18 months, and four times over an 18-24 month horizon.

The 10Y/3M curve that we watch reported fewer false positives, one of the reasons we use it when measuring recession risk. But daily signals for the 10Y/3M curve still falsely signaled a recession 16 times over a 12-month horizon and five times the next 12-18 months, but zero times over an 18-24 month horizon. Extending the history for those indicators only worsened results (see table 3 in the appendix).

Keep in mind that an inverted curve is a signal that markets expect the Fed will need to loosen monetary policy in the future, and the overarching reason is in response to fears that a slowdown in economic conditions could lead to recession. In other words, an inverted curve reflects market recession fears.

But when do we do pay attention, and when do we ignore?

Table 1

Reading The Yield Curve "T" Leaves

With daily signals often premature and wrong, we find that average monthly spreads are less noisy and thus give fewer false signals (see table 2). We watch for two consecutive months of inversion before a signal is confirmed.

For instance, negative spreads of the 10Y/2Y and 10Y/1 gave a false signal only once since 1982. No false signal was indicated by the 10Y/3M, which is why we favor this indicator in our own analysis.

Interestingly, the near-term forward rate, the Fed's favorite, indicated the highest number of false signals, at three. This trend is in line with the false signals from the daily average spread. But regardless of the preferred indicator, historically, through 2020, when any two of the four spreads invert for at least five months, a recession occurred within the next 12 months.

S&P Global Economics also recognizes that recession risks have increased as uncertainty about the outlook has worsened. Our qualitative assessment of recession risk over the next 12 months is now about 30%. Given the heightened uncertainty, we're setting the range at 25%-35%. Because of the lag in monetary policy impact on the economy, we see recession risks greater in 2023, closer to the top of this range, as cumulative interest rate hikes take hold.

Table 2

But even if we assume a given signal was correct, the time between the initial inversion and when the recession starts is wide, varying between 12 and 17 months, according to our analysis, and with an average duration around 14.6 months. However, markets can see an inversion start up to 36.5 months before a recession. (These very early inversions are not considered signals in our tables.) The 10Y/2Y signal, when accurate, historically does lead inversions of other curves. On average, a recession arrives 16.8 months after its initial inversion. So, assuming the signal proves to be accurate, that would imply a recession wouldn't arrive until later in 2023. This is largely in line with our qualitative assessment that the risk of recession increases in 2023.

Respect The Curve

While we respect the curve, it's hard to ignore that market conditions today have made the curve's crystal ball cloudier. Indeed, the details around the business cycle make understanding any economic signal tough.

First, the reliability of average daily/intraday inversions can be misleading because of technical issues rather than a true signal of a slowing economy. The recent April inversion is one example, where the two-year note yield surpassed the 10-year benchmark yield, with the 10-year benchmark yield held down as demand for long-term bonds received a boost from end-of-quarter rebalancing by institutions. A week later, yields on long-term bonds steepened sharply on the news that the Fed would commence balance sheet reduction at its May meeting.

While monthly data helps smooth daily technical issues, monthly readings may still face challenges. Current market expectations for significant rate hikes at the short end of the yield curve would cause spreads to narrow over the near term, which is likely in play now. The eventual balance sheet reduction efforts, which we expect to be announced in May, could raise rates at the long end in response to liquidity constraints, ultimately steepening yields as the unwind progresses. Monetary policy actions may later leave spreads in positive territory, further complicating any proper signal of a weakening economy.

Over the last 65 years, 10 episodes of recession have occurred--the earliest in 1957 and the latest in 2020 (see charts 2 and 3). Based on available data, these yield spreads' ability to signal recessions within 12-24 months has held up relatively well, though the spreads differ in their predictive abilities.

Additionally, the near-term forward spread tends to stay deeply inverted and for a longer period. The 10Y/1Y inversion has preceded every recession. The 10Y/2Y inverted for just three days (Aug. 27-29) prior to March 2020, missing the pandemic-driven recession, according to our measure. Although, it stayed in zero territory through Sept. 3, which some classified as a true signal.

Similarly, our March economic outlook expects an overshoot in monetary policy through 2023 will slow growth and inflation further than the Fed would like, ultimately leading it to loosen rates in 2024. With the Fed likely to turn up the dial earlier than initially expected, that would also mean removing accommodation sooner. The question is whether the Fed will be able to loosen policy before the U.S. economy sinks into recession, or after.

Chart 2

Chart 3

Dangerous Curves Ahead

The disconnect between the long and short ends of the yield curve is possible evidence of how monetary policy interventions are distorting the dynamics of the yield curve. Since January 2022, the 10Y/2Y, 10Y/1Y, 5Y/2Y, and 5Y/3Y spreads have been trending lower. Indeed, the probability of recession over the next 12 months, according to our augmented yield-curve model, stands at 20% and 9% for the 10Y/2Y and 10Y/1Y spreads, respectively, in April (see chart 4). On the other hand, our augmented yield-curve model in April using the 10Y/3M and near-term forward spreads predict recessions at just 6% and 3%, respectively (see chart 5).

Historically, the monthly average probability of recession for the near-term forward spread averaged 40% in 2007, with a peak of 58% that year. The lowest monthly average was 13% in 1969. In the last three recession episodes, peaks have been 50% in 2000 and 44% in 2019, while monthly averages have been 29% in 2000 and 36% in 2019.

For the 10Y/3M spread, the monthly average probability of recession--within 12 months--averaged 57% in 1979, with a 79% peak that year. The curve remained high, at 32%, throughout the back-to-back recessions between 1980 to 1982. The predictive strength has lessened over time but remained strong and comparable to the near-term forward spread without the volatility. In the last three recessions, it was 39% in 2001, 33% in 2007, and 38% in 2020.

The 10Y/1Y spread saw the highest monthly average probability of recession--within 12 months--at 60% in 1979, with a 84% peak that year. The curve remained high, at a monthly average of 37%, throughout the back-to-back recessions in 1980-1982. In the last three recessions, peaks have been 34.6% in 2000, 34.3% in 2007, and 28.2% in 2019, while monthly averages have been 26% in 2000, 23% in 2007, and 22% in 2019. The probability of recession for the 10Y/1Y reached 9% in April.

For the 10Y/2Y spread, the monthly average probability of recession--within 12 months--was 44% in 1979, with a 60% peak that year. The curve remained high, at a monthly average of 34%, throughout the back-to-back recessions in 1980-1982. The 10Y/2Y later posted a probability of recession lower than both the near-term forward and 10Y/3M spreads. The monthly average probability of recession has ranged from 19% (in 2007) to 28% (in 2001) over the last three recessions. The probability of recession for the 10Y/2Y in April 2022 reached 20%, after crossing the 10% level in February 2022.

Chart 4

Chart 5

The Fed Wild Card

As we expected, the Federal Reserve doubled its funds rate to 25 bps-50 bps in March and signaled many more increases are expected this year and next. We now expect at least two 50 bps rate hikes, back to back, at the May and June meetings. And we wouldn't be surprised if the Fed also plans for a trifecta, with a third 50 bps hike in July.

We wouldn't rule out a 75 bps hike this year. While currently only St. Louis Fed President James Bullard has said that the Fed "shouldn't rule out a 75 bps hike" to tame raging inflation, what would stop the Federal Open Market Committee from ripping off its monetary accommodation band-aid with a 75 or 100 bps move this year? Balance sheet reduction efforts, which will likely be announced in May, may also give the Fed a chance to slow price pressures.

While both interest rate hikes and balance sheet reduction efforts are aspects of the bank's monetary policy tightening efforts, they affect the yield curve differently. The short end of the yield curve has already steepening this year on expectations for increased rate hikes. However, the long end of the curve has not moderated even in the face of geopolitical concerns. Perhaps markets (like us) don't see the Russia-Ukraine conflict wearing on U.S economic activity. However, the start of the balance sheet reduction at the long end of the curve will likely increase long-term yields, steepening the curve and possibly distorting the curve's ability to signal weakening economic conditions.

Yield Curve Monitoring Continues

While the debate on whether the yield curve as a predictor for recession is ongoing, we recognize that both positions have merit. However, we still have a certain respect for the historical performance of the yield-curve indicator, and we hope that the Fed manages to walk the line.

We'll continue to monitor the curve and inform the markets of our expectations in our publications.

Appendix

Table 3

Table 4

Table 5

Extending the 10Y/3M curve:  Since the three-month constant maturity rate only dates to 1982, to extend the series, two methods are applied:

• Use secondary market three-month rate during prior years and convert the three-month discount rate to a bond-equivalent basis ("3Mo-Bond Equivalent") where bond-equivalent = 100*(365*discount/100)/(360-91*discount/100), where "discount" is the discount yield expressed in percentage points
• Use the federal funds rate before 1982 as a substitute for three-month constant maturity ("-3Mo-Federal Funds")

While both had similar predictive power in signaling recession, the federal fund rate is volatile around the three-month constant maturity series relative to the three-month bond equivalent (see chart 6). The three-month bond equivalent tracks the three-month constant maturity series more closely.

Chart 6