Statistical corruption in Beijing’s air quality data has likely ended in 2012


Abstract

This research documents changes in likely misreporting in official air quality data from Beijing for the years 2008–2013. It is shown that, consistent with prior research, the official Chinese data report suspiciously few observations that exceed the politically important Blue Sky Day threshold, a particular air pollution level used to evaluate local officials, and an excess of observations just below that threshold. Similar data, measured by the US Embassy in Beijing, do not show this irregularity. To document likely misreporting, this analysis proposes a new way of comparing air quality data via Benford’s Law, a statistical regularity known to fit air pollution data. Using this method to compare the official data to the US Embassy data for the first time, I find that the Chinese data fit Benford’s Law poorly until a change in air quality measurements at the end of 2012. From 2013 onwards, the Chinese data fit Benford’s Law closely. The US Embassy data, by contrast, exhibit no variation over time in the fit with Benford’s Law, implying that the underlying pollution processes remain unchanged. These findings suggest that misreporting of air quality data for Beijing has likely ended in 2012. Additionally, I use aerosol optical density data to show the general applicability of this method of detecting likely misreporting in air pollution data.
In: Atmospheric Environment, Volume 127, February 2016, Pages 365–371