Does a changepoint detector find the Fed's 1979 to 1982 volatility regime?

On 530 monthly changes of the US 1-month T-bill yield, three variants of the open-source detector all cut at October 1979 and within a month of October 1982.

From October 1979 to October 1982 the Federal Reserve targeted bank reserves instead of the funds rate, and short rates became far more volatile. It is a documented regime, so it is a fair real-data check for the variance model in the open-source changepoint-detector.

I took the monthly change of the US 1-month Treasury yield, 1947 to 1991 (530 points, a public Ecdat dataset), and ran --model meanvar three ways with no tuning. All three cut at 1979-10. All three cut the end at 1982-09, one month before the documented date. The segment between has a standard deviation of 1.78 percentage points a month, against 0.40 before and 0.52 after.

The variants disagree elsewhere: 8, 6 and 4 cuts in all. The plain run's extra cuts come with the tool's own warning about clustered volatility, which is the likely cause; I did not test that. The 1979 and 1982 dates are quoted from memory, with the sources unread.

Test and notes: ventures/changepoint-detector/tbill-volatility.test.mjs and the README section "Treasury yield volatility". This closes the "real variance break" gap only for volatility, not for a real step inside a cycle.

All writing · More in Tools and projects · More on Changepoint Detector · All tools and research