Energy Poverty, Human Capital, and the Long-Run Gdp-per-Capita Relationship in Nigeria: An ARDL Analysis
Year: 2026
Type: article
Abstract: This study examines the long-run association between energy poverty, human capital and the level of real GDP per capita in Nigeria using 40 annual observations (1985-2024). Energy poverty is an equal-weighted composite deprivation index (electricity access, clean-cooking access, electricity consumption per capita), normalised by min-max and entered in logs (lnEP; higher values mean greater deprivation). Unit-root tests classify lnGDPPC, lnK, lnTO and inflation as I(1) or I(0)/I(1) but leave lnEP and lnHC unresolved (first differences not clearly stationary), so all bounds-test inference is conditional on I(1) behaviour. An ARDL(1,1,1,1,1) error-correction model without human capital (Case III, k = 4) gives a bounds F-statistic of 8.82, above the Pesaran-Shin-Smith upper 1% bound (5.06) and a simulated finite-sample upper 1% bound (6.30); the long-run lnEP coefficient is −0.80 (95% CI −1.12 to −0.49), the total association. With lnHC added (k = 5), F = 3.80 is inconclusive against the finite-sample 5% upper bound (4.48) and the lnEP multiplier is positive and imprecise (0.85; 95% CI −1.45 to 3.16). A re-specified dynamic human-capital equation gives a long-run lnEP coefficient of −0.21 (p = 0.044). The product-of-coefficients indirect effect is −0.45 (Sobel 95% interval −1.20 to 0.30); the block-bootstrap intervals for the indirect and total effects are far wider and include zero, so a large indirect share is not supported and H4 is not accepted. Adding a 1996 intercept shift removes the long-run relationship, and results depend on the sample and estimator. The evidence is therefore exploratory and associational rather than causal. Every estimate is conditional on the equal-weighted composite index, and its robustness to index construction (alternative weights, leave-one-component-out indices and separate component indicators) has not been established. No mediated share is reliably estimated, because the indirect-effect intervals include zero.
DOI: 10.64229/ceqh9m04
Institution: Federal University Lafia
Language: English
Subfield: Economics and Econometrics
Domain: Social Sciences
Open Access status: hybrid
OpenAlex ID: W7236617889
