XTCOMBREAK: Stata module to estimate the common break date with a confidence interval (Bai 2010) and test whether the break really is common across units (Jiang and Kurozumi 2026)

xtcombreak implements two complementary papers on common breaks in panel data. xtcombreak estimate implements Bai (2010). In a single time series only the break fraction can be estimated consistently; the break date cannot. Bai's result is that a panel fixes this: as N grows, P(khat = k0) -> 1, so the exact date is recovered. This holds even when a regime contains a {bf:single observation}, which makes it useful for pinning down the onset of a new regime quickly. The command estimates the date by least squares, quasi-maximum likelihood or feasible GLS, reports Bai's parameter-free confidence interval, and runs his series-by-series Chow test to say which units actually broke. xtcombreak test implements Jiang and Kurozumi (2026). Every common-break method, including Bai's, assumes the break is at the same date for all units. This is a real restriction and can fail. The command tests that assumption with a self-normalised CUSUM statistic built from OLS residuals.