Smooth structural changes and common factors in nonstationary panel data: an analysis of healthcare expenditures†

Abstract

This article suggests new panel unit root tests that allow for multiple structural breaks and control for cross-correlations in the panel. Breaks are modeled with a Fourier function, which allows for smooth or gradual change rather than abrupt breaks. Cross-correlations are corrected by using the PANIC procedure. The simulations show that our tests have good size and power properties and perform reasonably well when the nature of breaks or the factor structure is unknown. The new panel unit root tests support fresh evidence on the persistence of healthcare expenditures in OECD countries.