Independent Verification of the European Climate Law's Effect on Nitrogen Dioxide Pollution: A Satellite-Derived, Difference-in-Differences Analysis

SAKSHI D. MASKE 

Satellite observations make it easy to compare air pollution in countries that adopted a policy with countries that did not. This paper examines what such a comparison identifies. It uses Sentinel-5P TROPOMI tropospheric NO₂ for 2019–2024 and a country-level difference-in-differences (DiD) model of the European Climate Law that compares the EU-27 with nine neighbouring non-EU countries. NO₂ fell by 1.34 × 10⁻⁶ mol/m² more in the EU-27 (4.3% of its pre-treatment level, p = 0.022). The estimate is stable under year-month fixed effects, group-specific seasonality, meteorological and COVID-19 controls, a wild cluster bootstrap and the removal of any single comparison country. It loses significance without the winter months (p = 0.11), with countries weighted by area (p = 0.11), with log GDP (p = 0.21), with real GDP per capita in place of nominal GDP (p = 0.072) and under randomization inference on the coefficient (p = 0.17), and it appears only from the third year after the law. Most of it reflects the composition of the two groups. When the treatment label is given to random sets of countries, the estimate tracks how much more polluted the labelled group was before treatment (r = −0.81), and the EU-27 estimate is close to what its pre-treatment gap predicts. On a 0.1° grid, comparing EU and comparison-group pixels that started at the same NO₂ level reduces the estimate to −0.21 × 10⁻⁶ (95% CI −0.67 to +0.24). The same comparison does detect a known 2.5% reduction added to EU-27 NO₂ after July 2021 (−0.80 × 10⁻⁶, p = 0.002). The satellite values agree with 1,543 ground stations of the European Environment Agency (monthly r = 0.68). The stations give a relative EU-27 decline of similar size that is not significant (−4.0%, p = 0.12), and they do not show the December 2023 drop that the satellite records in Denmark and Sweden. Descriptively, polluted areas lost about 15–25% of their NO₂ in north-western Europe, including the non-EU United Kingdom and Switzerland, but little in south-eastern Europe, including several EU members, in western Turkey and in Serbia. The data cannot separate EU membership from region, shared emission standards or winter retrieval conditions. Country-level DiD on satellite NO₂ should be checked against comparisons matched on baseline pollution before it is read as a policy effect.