A panel data investigation of macroeconomic determinants of carbon emission and food waste

This paper investigates the macroeconomic determinants of CO 2 emissions in four emerging economies (India, Brazil, South Africa and Indonesia—IBSA-Indonesia) during 1995–2022. It evaluates the independent role of food waste (FW) in CO 2 emissions. All variables are found to be integrated of order one using panel unit-root tests, and there is significant cross-sectional dependence across the panel throughout, as indicated by Pesaran's CD test. All three of the four statistics from the Westerlund cointegration test show that the null hypothesis of no cointegration is rejected, which gives reasonable, but not overwhelming, support for a long-run equilibrium relationship. There is no consistent annual series for food waste, so it was constructed using the Chow and Lin [59] method, with population and food price inflation as indicators; dietary energy supply was chosen as a proxy because it was too highly correlated with GDP per capita (r ≈ 0.93) to be used as an independent indicator. The Cross-Sectionally Augmented Mean Group (CCE-MG) estimator was used to estimate long-run coefficients, which were compared with pooled fixed-effects. GDP, renewable energy, and urbanization carry signs consistent with the EKC-EMT literature, with stronger significance in the pooled specification. The effect of electricity consumption was unresolved, with the two specifications disagreeing on the direction. Food waste's coefficient changes by more than a factor of three and reverses in significance between the two CCE-MG specifications that differ only in a cross-sectional lag, a choice that should not matter for a genuine effect. This instability suggests food waste's independent contribution cannot be reliably isolated from a small panel using a constructed proxy, reflecting a trade-off between avoiding income-collinearity and retaining explanatory power. Given the small number of countries in this sample, we present these findings as indicative associations rather than definitive causal estimates, alongside a transparent account of the obstacles that future researchers will need to address, aligning with UN SDGs 12 and 13.