Robust tests for unit roots in heterogeneous panels

We propose a robust test for unit roots in dynamic heterogeneous panels based on the sum of sign-type test statistics across the groups. Our test has an exact binomial null distribution and is consistent for a wide class of partially stationary processes with heterogeneous heavy-tailed errors. Monte-Carlo simulation shows that the new test has a stable size and is more powerful than the tests based on the ordinary least squares estimators (OLSE) in heterogeneous panels with heavy-tailed errors.