The Causal Effect of Air Pollution on Anti-social Behaviour

How to cite

Lohmann, Paul; Gsottbauer, Elisabeth; You, Jing; Kontoleon, Andreas. 2019. "The Causal Effect of Air Pollution on Anti-social Behaviour." AEA RCT Registry. AEARCTR-0004856. https://doi.org/10.1257/rct.4856-1.4

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Paul Lohmann* , Elisabeth Gsottbauer , Jing You , Andreas Kontoleon

* Primary investigator

Registry ID
AEARCTR-0004856
DOI
https://doi.org/10.1257/rct.4856-1.4
First registered
Last updated
Status
Completed
Trial dates
2019-10-16 – 2020-03-14
Countries
China ()
Registry record
http://www.socialscienceregistry.org/trials/4856

Abstract

The aim of this study is to provide causal evidence on the link between pollution and antisocial behaviour. We plan to estimate the causal impact of air pollution on economic decision-making by exploiting exogenous variation in local air pollution in Beijing, China. The primary outcomes of interest are experimental measures of antisocial behaviour obtained using incentivized experimental games including the 'Take Game' (as a measure for crime in the lab), 'Joy of Destruction Game' (as a measure of anti-social behaviour) and a third-party punishment game (as a measure of enforcement of pro-social behaviour). We also collect a range of secondary outcomes, which constitute potential mechanisms for effects on the primary outcomes, including tests of cognitive capacity, revealed measures for risk and time preferences and self-reported levels of mood, stress, anxiety and self-control. Experiments will be conducted with a large sample of students from Universities in Beijing.

Intervention

In this RCT we exploit the naturally occurring exogenous variation in air pollution as our treatment intervention. Pollution episodes generally occur over a series of days, followed by wind-driven clean-air episodes. By exploiting this natural discontinuity in air pollution exposure, we are able to survey both treatment and control groups within a time frame of several days. Participants will be randomly assigned to one of five groups, a low pollution (control group) or one of four high pollution treatment groups. The four high pollution treatment groups differ with respect to whether a pollution alert was issued by the research team prior to the survey distribution. All participants will be notified via direct messages about the upcoming survey, 24-hours prior to survey distribution. Pollution and weather forecasts will be used to determine the exact distribution date of the experimental survey. We utilise the official Air Quality Index (AQI) classifications as a broad guideline to define objective pollution levels for each treatment group:

1) Participants in the low pollution control group will be invited to complete a set of incentivised tasks when air pollution levels are objectively low, with AQI values in the ‘Good’ to ‘Unhealthy for Sensitive Groups’ range. They will be notified 24-hours in advance about the upcoming survey.

2) Participants in the high pollution treatment group will be invited to complete the same set of incentivised tasks when air pollution levels are objectively high, with AQI values exceeding ‘Very Unhealthy’ levels of pollution. They will be notified 24-hours in advance about the upcoming survey.

3) Participants in the high_alert pollution treatment group will be invited to complete the same set of incentivised tasks at the same time as group (2). They will be notified 24-hours in advance about the upcoming survey and receive an additional warning message about the expected unhealthy levels of air pollution.

4) Participants in the high_2 and high_alert_2 pollution treatment groups follow the same procedure as (2) and (3).

Absolute levels of pollution exposure may vary and AQI classifications will serve as a guideline only. To verify that official pollution levels are precise, we will take additional air quality measurements using modern air pollution measuring equipment located on the Campus of Renmin University.

Experimental design

The proposed RCT involves approximately 700 student participants enrolled at universities in Beijing, China. Data will be collected via online survey experiments using the Chinese messenger app 'WeChat'. Students will be contacted via direct messages and invited to participate in the study by following a link to an external survey platform. Once participants have completed the online survey, they will be immediately compensated via direct payment to their 'WeChat Wallets'.

We will implement an online between-subject experimental design in which students will be randomly assigned to a low pollution control group or one of four high pollution treatment (high, high_alert, high_2, high_alert_2), using a stratified randomisation procedure. All five groups will be sampled in December 2019. Participants will be notified 24-hours in advance about the upcoming survey via direct message. Participants in the high treatment condition will be sampled at the peak of a pollution episode. Participants in the high_alert treatment condition will be sampled at the same time but will have received an additional pollution alert with the survey notification (24-hours prior to the survey). Participants in the low pollution treatment conditions will be invited to complete the survey when air pollution levels are objectively low, immediately after the pollution episode. To collect additional data for our high-pollution treatment conditions, high_2 and high_alert_2 are sampled at the peak of a second pollution episode at the end of December 2019. The exact sampling dates will be determined based on pollution and weather forecasts.

Design details

Randomization method
The stratified randomisation will be based on respondent characteristics and performed using a statistical software package. The participants will first be stratified by gender, university, year of study, Hukou status and health status. Within each stratum, every fourth student will be assigned to a given treatment or control group.
Randomization unit
The randomisation is performed at the individual level
Clusters
N/A
Observations
700 students (undergraduate & postgraduate)
Arms
The expected sample size for each high-pollution treatment arm is approximately 250 participants per treatment (combined high and high_alert treatment groups). The sample size for the low-pollution control group is 169.
Primary outcomes
As primary outcomes we collect two measures of anti-social behaviour and one measure of enforcement of pro-social behaviour. 1) An experimental measure of crime obtained from the Take Game (Schildberg-Hörisch and Strassmair, 2014). 2) A measure of anti-social behaviour obtained from the Joy of Destruction mini game (Abbink & Herrmann, 2011). 3) A measure of norm enforcement obtained from a modified dictator game (Fehr & Fischbacher, 2004).
Secondary outcomes
As secondary outcomes we collect a range of measures including risk preferences (decision over gambles); time preferences (convex time budgets); cognitive performance (Raven’s matrices and Numerical Stroop) and additional subjective well-being & mental health variables.

Papers from this trial

  1. Lohmann, P. M., Gsottbauer, E., You, J., & Kontoleon, A. (2023). Air pollution and anti-social behaviour: Evidence from a randomised lab-in-the-field experiment. Social Science & Medicine, 320, 115617. https://doi.org/10.1016/j.socscimed.2022.115617

Cited by

No citation data available for this registration yet.

Keywords