Pooled mean group estimation with asymmetric effects and Fourier approximation
This study examines the estimation of long-run relationships in dynamic heterogeneous panels and proposes two extensions of the pooled mean group framework. First, the asymmetric pooled mean group estimator allows long-run responses to differ between positive and negative variation in the regressors through a partial-sum decomposition while preserving the pooled long-run homogeneity restriction. Second, the Fourier-augmented estimator embeds low-frequency sine and cosine terms in the error-correction equation to approximate smooth structural variation of unknown form. The asymptotic distribution of each estimator is derived and the Fourier augmentation is shown to leave the consistency and convergence rate of the long-run coefficient estimator unchanged. A Monte Carlo study spanning three data-generating processes establishes that the half-panel jackknife yields the most reliable inference among the corrections considered. In applications to the oil-price effect on output, aggregate energy demand, and the real effective exchange rate effect on output, the bias-corrected asymmetric estimator rejects long-run symmetry in all three cases, and the decomposition corrects a counter-theoretical sign in the estimated long-run income elasticity of energy demand. A single Fourier frequency is selected in each application, with the long-run estimates materially sensitive to that selection. Many economic relationships, such as how output responds to oil prices or how energy demand responds to income and prices, are estimated from panels of countries observed over a few decades. The widely used pooled mean group method assumes that rises and falls in a variable have mirror-image long-run effects and that the relationship is stable over time. This study relaxes both assumptions. It develops an estimator in which increases and decreases can have different long-run effects, and a further version that absorbs gradual structural change of unknown timing, and establishes the statistical properties of both. Simulation evidence identifies the bias correction that gives the most reliable inference in samples of realistic size. Applied to OECD economies, the methods show that long-run effects are asymmetric for oil prices, energy prices and exchange rates alike, giving researchers and policymakers a sounder basis for assessing responses to price and currency shocks.
Authors
- Fumitaka Furuoka (ORCID: https://orcid.org/0000-0003-1863-387X)
- Muhammad Afnan Arif (ORCID: https://orcid.org/0009-0008-9717-5275)
Publication Details
- Journal
- Figshare
- Published
- 2026-11-21
- DOI
- https://doi.org/10.6084/m9.figshare.35081980
- Primary Topic
- Spatial and Panel Data Analysis
- Type
- article
- Field-Weighted Citation Impact
- 0.00