XTHKRCOINT: Stata module to perform Hadri-Kurozumi-Rao panel cointegration test with the null of cointegration

xthkrcoint implements the panel cointegration tests of Hadri, Kurozumi and Rao (2015, Econometrics Journal). The distinctive features of the test are: 1. The null hypothesis is cointegration for every unit; the alternative is that at least one unit is not cointegrated. This is the natural null when the researcher wishes to establish a long-run relationship, and it complements the more common no-cointegration-null tests (xtcointtest). 2. The asymptotics hold with N fixed and T large, so the test suits typical macroeconomic and financial panels where N is small relative to T. No estimation of the number of common factors is required. 3. Arbitrary cross-section dependence (weak or strong, including common factors and cross-unit cointegration) is mopped up nonparametrically through the long-run variance of the pooled autocovariances, in the spirit of Driscoll-Kraay. The pooled statistic is therefore asymptotically standard normal and no bootstrap critical values are needed. 4. Serial correlation is handled by dynamic OLS (DOLS) — the cointegrating regression is augmented with leads and lags of Dx — and a bias correction removes the finite-sample negative bias of the autocovariance statistic (which otherwise makes the test conservative). The command reports the pooled statistic S{sub:K} (uncorrected) and S~{sub:K} (bias-corrected), each with a one-sided (upper-tail) p-value, a decision at the 5% level, and a per-unit breakdown showing which cross-sections drive a rejection. With graph it draws a two-panel diagnostics dashboard.