The Inverted Yield Curve: Does It Predict Recessions?

On 6 July 2022 the 10-year US Treasury yield fell below the 2-year yield and stayed below it for 537 daily readings in a row, through 26 August 2024. That is the longest such run in a record that starts in June 1976. As of 15 October 2026, the National Bureau of Economic Research (NBER) lists no US recession after the one that peaked in February 2020.

This guide sets every inversion since 1976 against the six recessions the NBER has dated. For two different yield spreads, it gives the date the inversion began, the NBER peak month that followed and the gap between them. The gaps run from 8 to 23 months. The 2022 inversion is the one with no recession dated after it.

A Treasury bond is a loan to the US government. Its yield is the annual return the lender earns, as a percentage of the bond's price. Prices and yields move in opposite directions, as we show in how bond prices and yields work.

The yield curve plots those yields for loans of different lengths, from a few months to 30 years. Long loans usually pay more than short ones. We cover its shapes in what the yield curve is and how its shape changes.

The curve is inverted when a loan that ends soon pays more than one that ends later. A spread is the gap between two yields. A spread below zero means inversion.

This guide uses two spreads from FRED, the Federal Reserve Bank of St. Louis's data service. T10Y2Y is the 10-year yield minus the 2-year yield, from 1 June 1976. T10Y3M is the 10-year yield minus the 3-month yield, from 4 January 1982.

Each gives one reading a day. Each is labelled in every figure below, because they are different measures.

The NBER is a private research group. Its Business Cycle Dating Committee keeps the chronology of US recessions.

Each recession runs from a peak month to a trough month. The peak is the last month of the expansion before it. We explain how a recession is defined in what a recession is and who dates it.

The NBER announces a peak months after it happens. It announced the March 2001 peak on 26 November 2001, the December 2007 peak on 1 December 2008 and the February 2020 peak on 8 June 2020.

This guide dates each recession by its NBER peak month.

The chart plots T10Y2Y on every day since 1 June 1976. Grey bands mark the six NBER recessions, from the month after each peak to the trough month. The line dipped below zero before each of them.

The dip before the February 2020 peak lasted three readings, from 27 to 29 August 2019, which is too narrow to see at this scale. The long stretch below zero at the right runs from 2022 to 2024. No grey band follows it.

One day below zero can be noise, so we count an inversion from a run. The clock starts on the first day of the first run of 20 or more daily readings in a row below zero, after the previous recession's trough. Twenty readings is about a month.

The lead time is the number of calendar months from that day to the NBER peak month.

The rule matters. T10Y2Y closed below zero for 18 readings in a row from 15 June 1998. The March 2001 peak came 33 months later. A threshold of 5 readings would count that dip as the start of the 2001 lead.

Dates move with the threshold. For the December 2007 peak, a minimum of 60 readings would move the T10Y2Y start from 31 January 2006 to 17 August 2006. The gap would fall from 23 months to 16.

Before the February 2020 peak, T10Y3M first closed below zero on 22 March 2019, for five readings. Its first run of 20 began on 23 May 2019.

The table gives the T10Y2Y result for each recession and for the 2022 inversion.

Source: FRED (T10Y2Y); NBER; YX Insights

T10Y2Y ran below zero for 20 or more readings before five of the six peaks. The gaps were 17, 10, 18, 13 and 23 months, a range of 10 to 23 with a median of 17.

T10Y3M starts in January 1982, after the January 1980 and July 1981 peaks, so it covers four recessions.

Source: FRED (T10Y3M); NBER; YX Insights

T10Y3M ran below zero for 20 or more readings before all four. The gaps were 14, 8, 17 and 9 months, a range of 8 to 17 with a median of 11.5.

The two spreads give different answers for the same recession. For the March 2001 peak, T10Y2Y gives 13 months and T10Y3M gives 8. For the February 2020 peak, T10Y3M gives 9 and T10Y2Y gives no qualifying run. A lead time belongs to the spread that produced it.

The chart sets the gaps from both tables side by side. Blue bars are T10Y2Y and orange bars are T10Y3M. The dashed bars in the last row run to the edge of the chart. They mark the 2022 inversion, where no peak has been dated after 51 and 48 months.

The T10Y2Y run began on 6 July 2022 and ended on 26 August 2024, 537 readings in a row. Its lowest reading was −1.08 percentage points on 3 July 2023. The T10Y3M run began on 25 October 2022 and ended on 12 December 2024, 534 readings in a row, with a low of −1.89 on 4 May 2023.

As of 15 October 2026, the NBER table lists no peak after February 2020. That is 51 months after the T10Y2Y run began and 48 months after the T10Y3M run began. Those two counts are elapsed time with no peak dated yet. They are not gaps, so they do not belong beside the completed gaps of 23 and 17 months in the tables above.

The NBER took 4 to 12 months to announce the three peaks above, so the table cannot yet exclude a peak in the past year. The record holds one inversion of this length.

The record can support a count. A run of 20 or more readings below zero came before five of six peaks on T10Y2Y and four of four on T10Y3M.

It cannot fix the timing. On the 20-reading rule, the gaps across the two spreads run from 8 to 23 months. For the same recession, the two spreads differ by up to 6 months.

It cannot rank the recessions by size. Before the December 2007 peak, T10Y2Y never fell below −0.19 percentage point. That recession lasted 18 months. Before the January 1980 peak it reached −1.69. That recession lasted 6 months.

It cannot show why a recession followed. The NBER's June 2020 announcement says the pandemic and the public health response had resulted in a downturn with "different characteristics and dynamics than prior recessions". Six recessions in 50 years is too few to separate an inversion's effect from everything else that happened in those months.

One long inversion has no recession dated after it as of 15 October 2026. Six recessions can describe that record. They cannot fix its timing.

This explainer is part of the YX Insights Academy. Each one takes a single idea and checks it against real data.

The same approach runs through everything else we publish:

Good places to start on the website:

An inverted yield curve is one where yields on US Treasuries that mature soon are higher than yields on those that mature later. It is usually measured by the 10-year yield minus the 2-year yield (T10Y2Y), or the 10-year minus the 3-month yield (T10Y3M), falling below zero. We cover the shapes in what the yield curve is.

Not in this record. A run of 20 or more readings below zero came before five of the six recessions since 1976 on T10Y2Y. It came before all four since 1982 on T10Y3M. The inversion that began in 2022 has no NBER-dated recession after it as of 15 October 2026.

In this record, the gap from the first run of 20 or more readings below zero to the NBER peak month was 10 to 23 months on T10Y2Y. On T10Y3M it was 8 to 17 months. The gap changed with each recession and with each spread.

This record cannot rank them. T10Y2Y covers six recessions and T10Y3M covers four. They give different start dates for the same recession. A lead time belongs to the spread that produced it. For the March 2001 peak, the gap was 13 months on T10Y2Y and 8 months on T10Y3M.

The NBER's Business Cycle Dating Committee dates US recessions. It announces a peak months afterwards: 8 months for March 2001, 12 for December 2007 and 4 for February 2020. We explain the process in what a recession is and who dates it.

DISCLAIMER: This article is strictly educational. Any information or analysis in this note is not an offer to sell or the solicitation of an offer to buy any securities. Nothing in this note is intended to be investment advice and nor should it be relied upon to make investment decisions. Any opinions, analyses, or probabilities expressed in this note are those of the author as of the note's date of publication and are subject to change without notice.