Are the Effects of Finland's Carbon Tax on Transport Emissions Robust to Donor Pool Composition?

This paper replicates Mideksa (2024), which applies the synthetic control method to estimate the effect of Finland's 1990 carbon tax introduction on per capita transport sector carbon emissions. I successfully replicate the main findings: the synthetic counterfactual closely tracks Finnish emissions in the pre-treatment period, with a substantial post-1990 divergence consistent with the original study. However, I show that this result is entirely driven by the inclusion of a single donor unit, Luxembourg, which is a structural outlier yet receives a positive weight in the optimization. Including Luxembourg is unusual in the literature: Luxembourg is a well-documented structural outlier, as its comparatively low fuel prices attract fuel tourism from neighboring countries, inflating its recorded transport emissions relative to actual domestic transport activity. Excluding Luxembourg deteriorates the pre-treatment fit and attenuates the estimated treatment effect, which falls to roughly a third of the original estimate and becomes indistinguishable from the placebo distribution. A donor pool resampling exercise confirms that the large treatment effect reported in the original study is systematically concentrated in subsamples containing Luxembourg. These findings highlight the sensitivity of synthetic control estimates to donor pool composition and provide an empirical illustration of interpolation bias in action.