Sticky growth, sticky carbon: Bayesian joint modelling evidence on a growth-carbon lock-in in emerging economies

This study examines the long-run joint dynamics between carbon dioxide emissions per capita and GDP per capita in a balanced panel of 22 low- and middle-income countries over the period 1990–2021. Using a Bayesian bivariate structural model, the analysis allows for bidirectional dependence, persistence, nonlinear feedback, country heterogeneity, and time-varying residual dependence within a unified empirical framework. Pooled estimates indicate a strong positive long-run association from income to emissions, with no credible evidence of an Environmental Kuznets Curve turning point within the observed income range. At the same time, emissions are positively associated with long-run income, and the squared emissions term is positive and statistically credible on average, indicating that the marginal association between emissions and income does not weaken at higher pollution levels. Country-level results reveal substantial heterogeneity: income-to-emissions elasticities are large and positive in most countries, while credible moderation in emissions growth is confined to a small subset and does not imply in-sample decoupling. Energy efficiency per capita emerges as a central structural channel, exhibiting a robust negative association with emissions and a positive relationship with income across nearly all countries, though with varying magnitudes. Renewable energy adoption is associated with lower emissions in many cases, while energy imports and foreign direct investment play limited roles once domestic structure is accounted for. Time-varying residual correlations indicate that, in some specific years, countries experienced higher-than-expected income growth alongside lower-than-expected emissions after controlling for observed factors, but these episodes are confined to a small number of years.