A changepoint detector on 41 years of Australian travel: Bali and SARS are not found

Monthly Australian resident departures 1976-2017: no seasonal mean model cuts at the Bali bombing or SARS, and the trend model's nearest cut is a month early.

We ran the changepoint-detector on 498 months of short-term resident departures from Australia (fpp2::departures). Two shocks are documented: the Bali bombing in October 2002 and SARS in April to June 2003.

With a 12-month season, the seasonal mean model cuts 42 times and the AR(1) model 28. Neither has a cut between 2002-09 and 2003-07. The trend model cuts 21 times, one of them at 2003-03, a month before SARS began, which we read as a nearby trend break rather than a detection. The series grows about eightfold with a widening season, so most cuts are the staircase-on-drift over-fit the tool already warns about.

So a short dip inside a growing, widening season sits below what the penalty keeps. The test (au-departures.test.mjs) pins these counts. Still wanted: a documented real step inside a cycle.

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