This paper introduces a new methodology called PANIC (Panel Analysis of Non-stationarity in Idiosyncratic and Common components) to understand non-stationarity in large-dimensional panel data. PANIC consists of univariate and panel tests that can detect whether non-stationarity is pervasive, variable-specific, or both. It tests the components of the data rather than the observed series, leading to more accurate inference when the components have different orders of integration. PANIC also allows for valid panel tests even when cross-section correlation invalidates pooling of statistics constructed using observed data. The key to PANIC is consistent estimation of the components, even when individual regressions are spurious. The paper provides a rigorous theory for estimation and inference, and Monte Carlo simulations show that the tests have good size and power. PANIC is applied to a panel of inflation series to illustrate its effectiveness.This paper introduces a new methodology called PANIC (Panel Analysis of Non-stationarity in Idiosyncratic and Common components) to understand non-stationarity in large-dimensional panel data. PANIC consists of univariate and panel tests that can detect whether non-stationarity is pervasive, variable-specific, or both. It tests the components of the data rather than the observed series, leading to more accurate inference when the components have different orders of integration. PANIC also allows for valid panel tests even when cross-section correlation invalidates pooling of statistics constructed using observed data. The key to PANIC is consistent estimation of the components, even when individual regressions are spurious. The paper provides a rigorous theory for estimation and inference, and Monte Carlo simulations show that the tests have good size and power. PANIC is applied to a panel of inflation series to illustrate its effectiveness.