Pollution is driving the young and well-educated out of China's cities
Concern about air pollution is rising in China. In 2013, the New York Times quoted a worried resident of Beijing. Her child was suffering from persistent respiratory ailments. "I hope in the future we’ll move to a foreign country," she said, "Otherwise we’ll choke to death” (Wong 2013).
Rapid growth in China has generated high levels of airborne very fine particulate matter (PM2.5). This is a pollutant that passes through skin and tissue and collects in the body's vital organs. PM2.5 causes death by both cardiovascular and respiratory disease, as well as lung cancer (WHO 2013). Between 1980 and 2015 the average concentration of PM2.5 rose from 30.98 μg/m3 to 66.90 μg/m3. The current level is almost six times the National Ambient Air Quality Standard that the US Environmental Protection Agency would mandate. Concentrations of sulphur dioxide (SO2), another harmful pollutant, have followed a similar pattern.
Government officials fret over reports that highly educated people are fleeing from China's polluted cities. But, until now, there have been no objective statistics on whether these are isolated anecdotes, or evidence of a widespread trend. That's why, in our recent work (Chen et al. 2017), we have tried to discover whether air pollution in China causes people to migrate out of the city.
Migration in China
Economists know that in the US, house prices respond to permanent reductions in air pollution, and also to people with high income who seek better air quality (Chay and Greenstone 2005, Close and Phaneuf 2017). Researchers have also found that sales of air purifiers and face masks in China go up when air pollution is higher (Ito and Zhang 2016). Internal migration in China, however, is highly constrained by a strict programme of residential registration, and so it is expensive and hard to move house. Therefore it is not clear whether individuals feel strongly enough to use migration as a way to escape air pollution.
We found that, despite the high cost of migration and government restrictions on it, large numbers of people in China have been moving away when air pollution increases. We found that a county that experienced an increase in pollution of 10% in the last year (everything else constant) would also experience a reduction in its population of about 2.7%. Keep in mind that, during the last 20 years, the average county in China has experienced a 60% increase in pollution.
The experiment: Measuring migration due to pollution in China
It is complicated to determine how population flows respond to changes in pollution, because there are large economic benefits for living highly polluted areas. Over five years, we found that Chinese counties with larger air pollution increases also had larger average inflows of population. Clearly this does not mean that the promise of air pollution has attracted people to a location. It does not even mean that people have ignored air pollution when they decided where to live. It just means that there are many benefits associated with economic prosperity that would make a county more attractive, which can be traded off against the downside of pollution.
So to understand how pollution is shaping migration, we needed to rely on a change in air pollution that was independent of economic fluctuations and other factors that might simultaneously be changing the population of a county. Our approach was similar to a drug trial that randomly administers the treatment and a control (placebo) to participants. Obviously we could not randomly expose people to high levels of air pollution to see if they migrated afterward. But we could take advantage of nature, which provided a natural experiment that is as rigorous.
These is a meteorological phenomenon called thermal inversion that prevents vertical ventilation of emissions. Under normal conditions, atmospheric temperature decreases as altitude increases. In this case, air moves from hot to cool regions, so pollutants are carried upwards, decreasing air pollution near the ground. But, for example, when hot air from a valley is carried to the top of a mountain range, or radiation from the earth warms mid-level layers of air during the night, the temperature becomes warmer at higher altitudes. Pollutants now remain near the surface. Because these thermal inversions can be caused by a large number of factors, they occur at a frequency that is almost as good as random.
In China, thermal inversions provided us with a natural experiment. We could measure migration in periods of five years in counties with a high or low number of inversions. Because of their randomness, treatment and control periods had similar average economic growth and other determinants of migration. Hence, we could safely claim that the differences in migration flows from differences in thermal inversion frequency were caused by air pollution.1
Who moved, and what are the implications?
Economic opportunity and air pollution probably both influence the choice to move. But we found that pollution definitely contributed to migration flows in China, even though migration would have come at a high cost.
We calculated net outflows of population using census information, and so we also knew the individual characteristics of the people who migrated. We found, for example, that college graduates were twice as likely to migrate as junior high-school graduates. Women were twice as likely to migrate as men, and that younger cohorts were also more likely to move. In a nutshell, we find that these inflows are primarily driven by well educated people at the beginning of their professional careers. These patterns, along with the size of the effect, mean that air pollution has already helped to shape the demographic characteristics of China's regions.
Often, the environment is considered a secondary factor when discussing economic growth. It is still a popular belief that the problem of pollution is one to be dealt with by future generations. The findings we present show that, today, the costs of pollution outweigh the significantly positive economic gains of living in some big cities. Pollution not only has environmental, but economic consequences, and is driving talent out of big cities in China. We hope this serves as a wake-up call to the fact that pollution is shaping the demography and economy of our cities, and the environmental impacts of growth must be factored into our policy plans.
Arceo, E, R Hanna, and P Oliva (2016), "Does the Effect of Pollution on Infant Mortality Differ Between Developing and Developed Countries? Evidence from Mexico City", The Economic Journal 126: 257–280.
Chay, K and M Greenstone (2005), “Does air quality matter? Evidence from the housing market”, Journal of Political Economy 113(2): 376–424.
Chen, S, P Oliva and P Zhang (2017), “The Effect of Air Pollution on Migration: Evidence from China”, NBER Working Paper 24036.
Close, B and D Phaneuf (2017), “Valuation of Local Public Goods: Migration as Revealed Preference for Place”, mimeo.
Fu, S and V B Viard, and P Zhang (2017), "Air Pollution and Manufacturing Firm Productivity: Nationwide Estimates for China", mimeo.
Hicks, D L, P Marsh and P Oliva (2016), "Air Pollution and Procyclical Mortality: Causal Evidence from Thermal Inversions", mimeo.
Ito, K and S Zhang (2016), “Willingness to Pay for Clean Air: Evidence from Air Purifier Markets in China”, NBER Working Paper 22367.
Jans, J, P Johansson, Pand P Nilsson (2014), "Economic Status, Air Quality, and Child Health: Evidence from Inversion Episodes", IZA Discussion Paper 7929.
Wong, E (2013), “In China, Breathing Becomes a Childhood Risk”, New York Times, 23 April 23.
World Health Organization (2013), Health effects of particulate matter, WHO.
 Thermal inversion was first used as an instrumental variable for air pollution by Arceo et al. (2016). They used it to estimate the effect of air pollution on infant mortality in Mexico City. Jans et al. (2014) took advantage of the phenomenon to investigate the short-run effects of air pollution on children’s health in Sweden, and Hicks et al (2016) have used it to estimate the effect on adult mortality in the US. In China, Fu et al. (2017) used thermal inversions to investigate the effect of pollution on manufacturing labour productivity. This is the first study, however, in which they have been used as a source of variation for pollution over a longer period.