Official Chinese air pollution data has previously shown evidence of manipulation when compared with data from US embassies in the same cities. The Chinese government has already taken action against the local officials involved, but now an independent statistical analysis shows the extent of the manipulation.

Jesse Turiel at Harvard University and Robert Kaufmann at Boston University looked at data from official Chinese monitoring stations as well as readings collected by US embassies in five cities: Beijing, Shenyang, Shanghai, Guangzhou and Chengdu. They found that there were regular divergences in the amount of PM2.5, a size of particulate with proven links to lung cancer, asthma and heart disease, recorded by Chinese and US stations.

The researchers looked at data from between 2015 and 2017, at which point the US stopped collecting data. They noticed a statistically unlikely amount of days on which pollution levels were just below the limit imposed by China’s “blue sky” policy, which created an index for each city where results at 100 or above were deemed too high and results 99 or below were acceptable.

“What that encouraged therefore, was any days that were close to 100 you’d just report 99, 98, 97,” says Turiel. “You could see this in the data. There was a very obvious bubble right below 100 and a very low proportion right at 100. People will use creative methods if they can get away with it.”

The divergences were 40 per cent more frequent than would be expected by chance, and 63 per cent of the discrepancies saw the Chinese data lower than the US readings. It was also more common to see misreporting on the days with the worst pollution, which is when the worst associated health effects are found.

Although these data discrepancies have been noted before, the pair’s work is the first time a robust statistical analysis has ruled out the possibility of it happening by chance.

The Chinese environment ministry announced in 2017 that 1140 officials were “held to account” for violating pollution rules after inspections the prior year. In early 2018 it said that it had caught officials from seven cities manipulating data.

Turiel doesn’t know whether crackdowns have stopped the problem, and US embassy data is no longer available to check. There is, however, evidence to suggest that air quality in Chinese cities improved during the period of the study.  The US data shows that annual concentrations of PM2.5 fell by more than 25 per cent between 2013 and 2017.

Turiel and Kaufmann believe that their statistical approach could be used by governments to spot fraud within local government and guide enforcement.

Chinese air pollution data was altered, statistical analysis suggests | New Scientist