Finite sample inference methods for dynamic energy demand models

This paper considers finite sample motivated inference methods in dynamic energy demand models, in which case commonly used econometric methods remain asymptotic. We focus on structural stability, and on exact confidence set estimation of elasticities. We account for intractable and nuisance parameter dependant distributions through Monte Carlo test procedures. For long-run elasticities which depend on parameter ratios, we assess available asymptotic and exact methods with Fieller based alternatives. Fieller based and exact methods invert approximate and exact relevant test criteria (respectively) and may lead to unbounded set estimates. Our empirical results underscore the importance of using identification-robust inference methods.