This paper introduces PANIC, a new methodology for analyzing non-stationarity in panel data by decomposing it into common and idiosyncratic components. PANIC uses factor structure to detect whether non-stationarity is pervasive or specific to individual units. It tests components rather than observed series, improving inference when components have different integration orders. PANIC allows valid panel tests even when cross-section correlation invalidates pooling. The key is consistent estimation of components despite spurious regressions. Theoretical results show good size and power in Monte Carlo simulations. PANIC is applied to inflation data. The method uses principal components to estimate factors and idiosyncratic components, accommodating both I(1) and I(0) errors. It addresses three econometric problems: size issues in summing series with different integration orders, weak cross-section correlation in idiosyncratic components, and power of pooled tests. PANIC tests common factors and idiosyncratic components separately, allowing determination of non-stationarity sources. It also resolves the issue of pooled tests by using demeaned and detrended data. The method is robust to different stationarity assumptions and allows for pooled tests based on idiosyncratic components. Monte Carlo simulations show good performance, with PANIC tests having appropriate size and power. The results support the use of PANIC for analyzing non-stationarity in panel data.This paper introduces PANIC, a new methodology for analyzing non-stationarity in panel data by decomposing it into common and idiosyncratic components. PANIC uses factor structure to detect whether non-stationarity is pervasive or specific to individual units. It tests components rather than observed series, improving inference when components have different integration orders. PANIC allows valid panel tests even when cross-section correlation invalidates pooling. The key is consistent estimation of components despite spurious regressions. Theoretical results show good size and power in Monte Carlo simulations. PANIC is applied to inflation data. The method uses principal components to estimate factors and idiosyncratic components, accommodating both I(1) and I(0) errors. It addresses three econometric problems: size issues in summing series with different integration orders, weak cross-section correlation in idiosyncratic components, and power of pooled tests. PANIC tests common factors and idiosyncratic components separately, allowing determination of non-stationarity sources. It also resolves the issue of pooled tests by using demeaned and detrended data. The method is robust to different stationarity assumptions and allows for pooled tests based on idiosyncratic components. Monte Carlo simulations show good performance, with PANIC tests having appropriate size and power. The results support the use of PANIC for analyzing non-stationarity in panel data.