Estimating the Local Air Pollution Impacts of Maritime Traffic: A Principled Approach for Observational Data

TL;DR

This paper introduces a new method for estimating the causal impact of maritime traffic on local air pollution using observational data, applying matching algorithms and randomization inference to a case study in Marseille.

Contribution

It presents a novel approach combining matching algorithms and causal inference techniques to estimate maritime traffic's effects on air pollution without natural experiments.

Findings

Cruise vessel arrivals significantly increase nitrogen dioxide, particulate matter, and sulfur dioxide levels.

Road traffic has a larger daily impact on air pollution than cruise traffic.

The method transparently addresses challenges in high-frequency time series causal inference.

Abstract

We propose a new approach to estimate the causal effects of maritime traffic when natural or policy experiments are not available. We apply this method to the case of Marseille, a large Mediterranean port city, where air pollution emitted by cruise vessels is a growing concern. Using a recent matching algorithm designed for time series data, we create hypothetical randomized experiments to estimate the change in local air pollution caused by a short-term increase in cruise traffic. We then rely on randomization inference to compute nonparametric 95\% uncertainty intervals. We find that cruise vessels' arrivals have large impacts on city-level hourly concentrations of nitrogen dioxide, particulate matter, and sulfur dioxide. At the daily level, road traffic seems however to have a much larger impact than cruise traffic. Our procedure also helps assess in a transparent manner the identification challenges specific to this type of high-frequency time series data.