ReadingResearchRadarInvestment framework
Sign in / Sign up
← Back to reading
NewsMaterial2 min readDisconfirmAI Editorial

Econbrowser Blogger Menzie Chinn: Term Spread Inversions Barely Help Predict Australian Recessions

Econbrowser blogger Menzie Chinn finds that the 10-year minus 3-month term spread has negligible predictive power for Australian recessions across ECRI, Melbourne Institute, and modified Sahm rule chronologies, with pseudo-R2 below 0.14 over 1986–2026 and a coefficient of the wrong sign. Even restricting to the pre-pandemic period, which raises pseudo-R2 to as high as 50%, the spread still retains the wrong sign.

2026-11-22 08:14 GMT+8

Econbrowser blogger Menzie Chinn has tested whether the 10-year minus 3-month government bond term spread predicts Australian recessions. His answer: not really, even using alternative recession chronologies.

The standard yield curve hypothesis holds that an inversion—when short-term rates exceed long-term rates—signals that markets expect economic contraction ahead. Chinn's question was whether this logic applies to Australia.

He tested the 10-year minus 3-month spread against three different recession definitions: the Economic Cycle Research Institute (ECRI) peak-to-trough dates, the Melbourne Institute's chronology, and a modified Sahm rule using a 0.75 percentage point unemployment threshold (the level the Reserve Bank of Australia has suggested). The unemployment-based Sahm rule is available only from February 1978 onward.

The results are stark. Over the 1986–2026 window, none of these recession chronologies are predicted with a pseudo-R² greater than 0.14, and the coefficient on the spread is opposite of what the standard story would require. Restricting the sample to the pre-pandemic period raises the pseudo-R² to as high as 50% for the ECRI or Melbourne Institute chronologies, but the spread still retains the wrong sign.

Chinn also notes a complication with the Sahm rule itself. Using the 75 basis point threshold, the rule would flag recessions in 2001 and 2009—years when neither total employment nor GDP declined on a sustained basis. This suggests that even the recession definition introduces noise into the Australian context.

Sign in / Sign up
ReadingResearchRadarInvestment framework

ECRI and the Melbourne Institute largely agree on the mid-1980s and 1990–91 recessions, and both agree there was no 2009 recession. But they disagree on the 1970s episodes.

Chinn is careful to distinguish this finding from the broader question of whether the term spread predicts GDP growth levels. He cites Poke and Wells (2009) to note that the lack of predictive power for discrete recession events is different from a lack of predictive power for the growth rate itself.

For investors, the practical implication is narrow but concrete: an inverted 10-year minus 3-month spread in Australian government bonds has not historically preceded recessions, and the sign of the relationship is inverted relative to what the standard model predicts. This does not mean the spread carries no information—it may still relate to growth levels—but it undermines its use as a binary recession signal in the Australian context.

Sources

  • econbrowser.com↗
Continue researching →