Canada’s 10-year-to-three-month term spread was positive in September, while a pre-pandemic model would have predicted a 2024 recession, a contrast Menzie Chinn said depends partly on recession dating.
In an Oct. 26 Econbrowser analysis, Chinn examined the indicator against several recession chronologies. He said the spread alone offers little evidence that a downturn is taking shape, despite model results that pointed to a 2024 recession under some settings.
Research by Chinn and Ferrara (2026), as summarized in Chinn’s analysis, found that the model built on pre-pandemic observations would have signaled a downturn in 2024 if assessed at a 40% threshold. A full-sample version would have predicted a recession using a 35% threshold. The pre-pandemic version also would have missed the recessions of 2008 and 2020.
Other specifications changed the result. A debt-service-augmented version also would have predicted a recession at a 40% threshold, while the term-premium specification did not. Chinn noted that the latter used a shorter sample. The analysis’s probability chart refers to the projected recession date, rather than the calculation date, and assumes no recession through April 2026.
Chinn’s comparison also changes with the historical record used for testing. He considered recession dates from CD Howe, ECRi and Philip Cross, whose chronology was updated by Hao and Ng. In the comparison using Cross’s dates, the 2020 recession was aligned with ECRI’s dates. Different chronologies can change which episodes count as misses or successful warnings.
On the longer historical question, Chinn said data from 1970 onward show no recession after a term spread reached a similarly high level. He said historical correlations also offered little support for a 2026 recession signal. Chinn cautioned that the indicator looks 12 months ahead and cannot incorporate unforeseen developments, including a major escalation in the trade conflict.
