@ARTICLE{26543120_995815590_2024, 
	author = {Anton Skrobotov}, 
	keywords = {, panel unit root testing, panel stationarity testing, detrending, common factorscross-sectional correlation},
	title = {Panel Data Unit Root Testing: An Overview},
	journal = {HSE Economic Journal },
	year = {2024},
	volume = {28},
	number = {4},
	pages = {664-701},
	url = {https://ej.hse.ru/en/2024-28-4/995815590.html},
	publisher = {},
	abstract = {This review discusses methods of testing for a panel unit root. Modern approaches to te­sting in cross-sectionally correlated panels are discussed, preceding the analysis with an analysis of independent panels. In addition, the de-trending methods and corresponding asymptoticresults are discussed. To account for cross-sectional correlation, the methods based on de-facto­rization and bootstrap are considered. In conclusion, links to existing packages that allow implementing some of the described methods are provided.This review discusses methods for testing for unit roots in panel data. The investigation of several time series together instead of analyzing each one separately and the motivation fortesting for panel unit roots are discussed. The review begins with a consideration of the simplest panel unit root tests with independent errors and two types of alternative hypothesis: homogeneous and heterogeneous. For the simplest tests, their asymptotic behavior is described under different types of convergence of the number of objects and the time horizon. Then, the issue of including a deterministic component and changing the asymptotic results are considered, aswell as methods for accounting for a weak dependence of errors. The first section concludes with methods based on p-values.The next section is devoted to the important issue of accounting for cross-sectional correlation in panels and its impact on classical panel unit root tests. Cross-sectional correlation takes place according to some macroeconomic theories, which state that there are some common factors (e.g., technological shocks) that affect not one, but some set of variables. Modifications of classical tests based on factorization are described, when cross-sectional correlation is approximated by (possibly non-stationary) common factors based on the principal component method and based on factor approximation using cross-sectional means. Alternative methods based on resampling are considered. The section ends with a comparative Monte Carlo simulations analysis of various tests described in the review. The problem of imbalanced panel is discussed. In conclusion, references are given to existing packages that allow implementing some of the described methods.},
	annote = {This review discusses methods of testing for a panel unit root. Modern approaches to te­sting in cross-sectionally correlated panels are discussed, preceding the analysis with an analysis of independent panels. In addition, the de-trending methods and corresponding asymptoticresults are discussed. To account for cross-sectional correlation, the methods based on de-facto­rization and bootstrap are considered. In conclusion, links to existing packages that allow implementing some of the described methods are provided.This review discusses methods for testing for unit roots in panel data. The investigation of several time series together instead of analyzing each one separately and the motivation fortesting for panel unit roots are discussed. The review begins with a consideration of the simplest panel unit root tests with independent errors and two types of alternative hypothesis: homogeneous and heterogeneous. For the simplest tests, their asymptotic behavior is described under different types of convergence of the number of objects and the time horizon. Then, the issue of including a deterministic component and changing the asymptotic results are considered, aswell as methods for accounting for a weak dependence of errors. The first section concludes with methods based on p-values.The next section is devoted to the important issue of accounting for cross-sectional correlation in panels and its impact on classical panel unit root tests. Cross-sectional correlation takes place according to some macroeconomic theories, which state that there are some common factors (e.g., technological shocks) that affect not one, but some set of variables. Modifications of classical tests based on factorization are described, when cross-sectional correlation is approximated by (possibly non-stationary) common factors based on the principal component method and based on factor approximation using cross-sectional means. Alternative methods based on resampling are considered. The section ends with a comparative Monte Carlo simulations analysis of various tests described in the review. The problem of imbalanced panel is discussed. In conclusion, references are given to existing packages that allow implementing some of the described methods.}
}