Causal impacts of transport interventions on air quality

Abstract

The transport sector is one of the main sources of air pollution emissions, particularly for carbon monoxide, nitrogen oxides, and particulate matter. Evaluating the effectiveness of transport interventions on improving air quality is essential to informing future policy. However, a comparison of air quality observations before and after an intervention can be biased by various factors, such as weather conditions and seasonality effects. Causal inference methods generally have advantages in intervention evaluation in terms of data requirements, model building, and the interpretation of effect estimates. Causality goes beyond statistical association in the sense that it seeks to measure the net effect of an intervention on an outcome through all possible pathways directing from the intervention to the outcome. Causal inference methods have been applied to address the same question, however, the important confounders (such as weather conditions) are commonly controlled for by including variables in the causal inference model and assuming a parametric relationship. The thesis focuses on understanding the causal impacts of transport interventions on air quality. A novel ex-post policy evaluation framework, combining meteorological normalisation, change point detection, and causal inferencing, is proposed to overcome the limitations of previous approaches, and it is applied to three distinct transport interventions: improving public transport supply (Jubilee Line Extension), tightening road traffic emission standards (London Ultra Low Emission Zone), and restricting both transport activities and supply (COVID-19 lockdown). The Jubilee Line extension led to only small (< 1%) or insignificant changes in air pollution on average in London. The Ultra Low Emission Zone showed an average reduction of less than 3% for NO2 concentrations and insignificant effects on O3 and PM2.5 concentrations. The lockdown reduced the NO2 concentrations in London by less than 12% on average, and it had an insignificant effect on O3, PM10, and PM2.5. Therefore, the empirical results of the thesis consistently highlight the necessity of a multi-faceted set of policies that aim to reduce emissions across sectors with coordination among local, regional, and national government in order to achieve long-term improvements in air quality in cities.