Estimating the environmental Kuznets curve for Spain by considering fuel oil prices (1874–2011)

We perform a structural analysis on an environmental Kuznets curve (EKC) for Spain by exploiting long time series (1874–2011) and by using real oil prices as an indicator of variations in fuel energy consumption. This empirical strategy allows us to both, capture the effect of the most pollutant energy on carbon dioxide (CO₂) emissions and, at the same time, preclude potential endogeneity problems derived from the direct inclusion of fuel consumption in econometric specification. Knowing the extent to which oil prices affect CO₂ emissions has a straightforward application for environmental policy. The dynamics estimates of the long and short-term relationships among CO₂, economic growth and oil prices are built through an autoregressive distributed lag (ARDL) model. Our test results support the EKC hypothesis. Moreover, real oil prices are clearly revealed as a valuable indicator of pollutant energy consumption.     

Mischaracterising density dependence biases estimated effects of coloured covariates on population dynamics

Environmental effects on population growth are often quantified by coupling environmental covariates with population time series, using statistical models that make particular assumptions about the shape of density dependence. We hypothesized that faulty assumptions about the shape of density dependence can bias estimated effect sizes of temporally autocorrelated covariates. We investigated the presence of bias using Monte Carlo simulations based on three common per capita growth functions with distinct density dependent forms (θ-Ricker, Ricker and Gompertz), autocorrelated (coloured) ‘known’ environmental covariates and uncorrelated (white) ‘unknown’ noise. Faulty assumptions about the shape of density dependence, combined with overcompensatory intrinsic population dynamics, can lead to strongly biased estimated effects of coloured covariates, associated with lower confidence interval coverage. Effects of negatively autocorrelated (blue) environmental covariates are overestimated, while those of positively autocorrelated (red) covariates can be underestimated, generally to a lesser extent. Prewhitening the focal environmental covariate effectively reduces the bias, at the expense of the estimate precision. Fitting models with flexible shapes of density dependence can also reduce bias, but increases model complexity and potentially introduces other problems of parameter identifiability. Model selection is a good option if an appropriate model is included in the set of candidate models. Under the specific and identifiable circumstances with high risk of bias, we recommend prewhitening or careful modelling of the shape of density dependence.