THRESHOLD IDENTIFICATION IN GDP ENERGY INTENSITY DYNAMICS: ECONOMETRIC EVIDENCE FROM UZBEKISTAN, 2011–2023

Abstract

This study investigates threshold nonlinearity in the relationship between GDP growth and energy intensity (EI) in Uzbekistan over 2011–2023. Three nested models are estimated and compared: a linear OLS baseline (Model A), a two-regime model with an exogenously imposed structural break at 2017 (Model B), and a two-regime threshold regression with the threshold identified endogenously by grid search over the GDP growth rate (Model C). The key finding is that Model B — the exogenous 2017 reform break — substantially outperforms both the linear baseline (R² from 0.870 to 0.990, MAPE from 4.49% to 0.34%) and the endogenous threshold model (R² = 0.928, MAPE = 2.33%). The Chow test for Model B yields F = 48.78 (p < 0.001), compared to F = 3.25 (p = 0.093) for the endogenous threshold at τ* = 21.2% annual GDP growth. This finding implies that the break in the GDP elasticity of energy intensity — from γ_pre = −0.612 to γ_post = −0.860 — is best explained by the institutional discontinuity of the 2017 reforms rather than by a mechanical threshold in the GDP growth rate. The policy implication is that energy intensity improvement in Uzbekistan is institutionally driven: the quality of structural reform is a more powerful determinant of the EI–GDP elasticity than the pace of economic growth alone.