Skip to Content
SustainabilitySustainability
  • Article
  • Open Access

19 November 2026

22 Pages

What Nonlinear Unit Root Tests Can and Cannot Tell Us: Carbon and Energy Convergence in Sixteen Major Economies

Department of Economics, Osmaniye Korkut Ata University, Osmaniye 80010, Türkiye
Sustainability2026, 18(22), 11767;https://doi.org/10.3390/su182211767 
(registering DOI)
This article belongs to the Section Economic and Business Aspects of Sustainability

Abstract

Smooth-transition unit root tests are the standard instrument in the literature on carbon and energy convergence. Applied work in that literature routinely treats the transition speed, the threshold and the transition date that these tests produce as economic quantities in their own right. This paper asks what those estimates are worth. Under the null, the transition parameters are not identified; the open question is how much that matters in samples of the length actually available. We settle it by simulating from processes whose nonlinearity is known by construction and running the identical estimation code applied to the data. The tests themselves survive. Their matched critical values deliver the nominal size, including under the cross-sectional dependence that the usual relative transform induces, and the unrestricted and threshold-restricted procedures clearly outperform the linearised version on which most applied work relies. What does not survive is the interpretation of the estimates. The flatness of the criterion function, the shape of the fitted transition and the linearity of the fitted trend take the same values whether or not a genuine transition is present, and the failure is not numerical: the classifications are perfectly stable across lag rules, grid ranges and search intensities. Applied to four convergence variables for sixteen major economies, the rejection counts prove hard to separate from chance once dependence across countries and across variables is allowed for. The one count that crosses the conventional threshold does so marginally, and it depends on which member of the test family is used and on which countries define the comparison group. The recommendation is narrow: use these procedures as unit root tests, price every dimension of the search into the critical values, treat the composition of the comparison group as a result, and do not interpret the nonlinear parameters.

Article Metrics

Citations

Article Access Statistics

Article metric data becomes available approximately 24 hours after publication online.