Is there an environmental Kuznets curve in Saudi Arabia? cointegration and decoupling evidence

Introduction This study asks whether the environmental Kuznets curve (EKC) describes the relationship between income and carbon dioxide (CO 2 ) emissions in Saudi Arabia. Methods A log-quadratic EKC specification, augmented with primary energy use and the renewable share of primary energy, is estimated on annual data for 1990–2022 (n = 33; 31 effective observations in the dynamic models). The sample ends in 2022 because the harmonised real-income series ends in that year; data for 2023–2024 are used descriptively only. The autoregressive distributed lag (ARDL) bounds test provides the decisive evidence. For the specification selected by both the Akaike and Schwarz criteria, ARDL(2,0,0,0,1), the bounds F-statistic is 2.01, below the 10% lower (I(0)) bound of both the asymptotic (2.45) and finite-sample (2.75) critical values. It remains below the I(0) bound in all 162 lag configurations, in the best-fitting configuration free of residual serial correlation, after adding level shifts for 2014 and 2016, and after adding oil rents and trade openness. A Gregory–Hansen test allowing a regime shift also detects no cointegration. Unit-root tests indicate that per-capita emissions are stationary around a shifting mean, whereas income is integrated of order one. Long-run coefficients from the ARDL, nonlinear ARDL (NARDL) and dynamic ordinary least squares (DOLS) models are therefore reported only as exploratory quantities with measures of uncertainty. The income terms are jointly insignificant in every specification, and the Lind–Mehlum slope conditions for an inverted U are not met. The NARDL Wald test does not reject long-run income symmetry (p = 0.62), a result conditional on a long-run framework that the bounds test does not support. Results The data provide no evidence of an inverted-U EKC, but they also do not establish a statistically significant U-shape. Total emissions show relative but not absolute decoupling from gross domestic product (GDP): the Tapio elasticity is 0.34 for 2016–2022, and total emissions rose by 7.9% between 2016 and 2024. The decline in carbon intensity slowed after 2016. Discussion These patterns are described as associations and are not attributed causally to Vision 2030. Code and data that reproduce every table and figure are provided.