Air Pollution and Media Slant: Evidence from Chinese Corporate News

ABSTRACT

This paper examines the impact of air pollution on the media slant of publicly listed firms in China. Using a large panel of air quality and media data at the city level, we find that lower air quality generally leads to a more negative media slant. When the air quality falls from lightly polluted to heavily polluted, the number of negative sentences in a news article increases by about 1%. Our subsample analysis shows that the effect of air pollution on media slant is similar for news articles covering large and small firms, SOE and non-SOE firms and for official and non-official newspaper articles. Furthermore, the effect of air pollution on media slant is stronger for firms in heavy polluting industries. These results suggest that air pollution affects media slant.

KEYWORDS:

• Air pollution
• media slant
• China

JEL:

• D91
• G41
• L82
• Q5

Acknowledgments

We are grateful for the valuable comments from Paresh Kumar Narayan (the editor), the subject editor, and two anonymous referees. We also appreciate the valuable comments from Rose Liao and Steven Xiao. Xinjie Wang gratefully acknowledges financial support from the National Natural Science Foundation of China (Project No. 72171107), the Department of Education of Guangdong Province (Project No. 2020KZDZX1189) and the Southern University of Science and Technology (Grant No. Y01246210, Y01246110).

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

1. Some more recent findings in Wang (2018, 2020) and DeVault, Sias, and Startks (2019) reveal that institutional investors, as experts, can be noise traders in stock markets.

2. The weather data can be downloaded from http://data.cma.cn.

3. We calculate the implied air pollution level that maximizes media sentiment by taking the first derivative of Equation (1)(1) $Textsentiment=\frac{\mathrm{\#}weightedpositivesentences-\mathrm{\#}weightednegativesentences}{\mathrm{\#}weightedpositivesentences+\mathrm{\#}weightednegativesentences+1}$(1) with respect to AQI and setting it equal to zero, which yields optimal sentiment $=-\widehat{{\beta }_1}/2\widehat{{\beta }_2}$.

Additional information

Funding

This work was supported by the Department of Education of Guangdong Province [2020KZDZX1189]; National Natural Science Foundation of China [72171107]; Southern University of Science and Technology [Y01246210, Y01246110].