Short-Term Exposure to Air Pollution and Cognitive Performance: New Evidence from China’s College English Test

Abstract

This paper investigates the impact of air pollution on students’ cognitive performance in a high-stakes exam: China’s College English Test (CET). We match exams taken from 2013 to 2017 at 22 universities across China with hourly air pollution measures from the nearest monitoring stations. Identification leverages a student fixed effects model, which alleviates the concern of omitted variables, such as students’ ability. Our estimates indicate a statistically significant negative impact of fine particulate matters (PM$_{2.5}$) exposure during exam windows on cognitive performance. By focusing on a single language exam, instead of comparing performance across different test subjects, we are able to paint a more accurate picture of the cognitive impact of air pollution. We highlight the importance of short-term air pollution exposure for high-stakes cognitive performance. Our results suggest that temporary defensive measures could be important in mitigating the negative consequences of air pollution.

Appendix

A. Details of the IV Approach

Following previous studies such as Chen et al. (2018b), Fu et al. (2021), Chen et al. (2018a), Sager (2019), Yi et al. (2020) and Rivera (2021), we instrument air pollution using thermal inversion intensity. Thermal inversion occurs when the air temperature in the upper level of the atmosphere is higher than that of the lower level. This will trap local pollution near the ground, causing a higher level of human exposure to air pollution. Thermal inversion intensity is a continuous variable and larger values of it indicate stronger thermal inversion. Following He and Ji (2021), we constructed thermal inversion intensity based on temperature collected from NASA’s MERRA-2 Project.²³ We use thermal inversion intensity from counties where the universities are located as an instrument variable.

We use the following 2SLS model to estimate the effect of air pollution on students’ cognitive performance. The main assumption is that thermal inversion intensity is not directly affect students’ scores during the exam window, except through air pollution. In the first stage, we estimate the effects of thermal inversion intensity on PM$_{2.5}$ concentrations during the exam window.

$$\begin{aligned} \begin{aligned}&PM_{ut}=\varvec{\beta }_{tii} tii_{ut}+\varvec{\beta }_w {\textbf{W}}_{ut}+ \varvec{\beta }_x {\textbf{X}}_{iut} +\theta _t + \lambda _{ut} + \alpha _i + \mu _{iut}.\\ \end{aligned} \end{aligned}$$

In the second stage, we use predicted PM$_{2.5}$ to assess the impact on CET6 total scores.

$$\begin{aligned} \begin{aligned}&CET_{iut}=\varvec{\beta }_1 \widehat{PM_{ut}}+\varvec{\beta }_2 {\textbf{W}}_{ut}+ \varvec{\beta }_3 {\textbf{X}}_{iut} +\theta _t + \lambda _{ut} + \alpha _i + \epsilon _{iut}, \end{aligned} \end{aligned}$$

where $tii_{ut}$ is the thermal inversion intensity at university u at time t. As is described Sect. 4, ${\textbf{W}}_{ut}$ is a vector of weather measures, ${\textbf{X}}_{iut}$ is a vector of student-level indicators, $\theta _t$ denotes exam FEs, $\lambda _{ut}$ denotes university-by-month FEs, and $\alpha _i$ denotes student FEs.

Table 11 reports the first-stage estimates and Table 6 reports the 2SLS estimation results of impacts of PM$_{2.5}$ on CET6 total scores.

Table 11 First-Stage Results of the IV Approach

B. Additional Tables

Table 12 Distributions of the Number of Attempts and the Number of Students in the Full Sample

Table 13 Falsification Results: University FE Model

Table 14 Falsification Results: College/Department FE Model