How can we use Weibo to
measure the impact of air pollution
on urban activities in China?


利用微博签到数据揭示空气污染对城市活动的影响

Air pollution in China has caused health consequences. Understanding the complex impact of air pollution is crucial to assess exposure risk and define public health policies. However, only a few empirical analyses have been conducted, showing that avoidance behavior may only exist under certain circumstances. Some environmental injustice studies also shed light on the potential association. In spite of the importance, evidence of the relationship between air pollution and urban activity is far from enough.

Using 50 million geotagged check-in records in 2015-2016 on Weibo, a Tweeter-like platform, we correlated the air pollution level in 251 Chinese prefecture cities with the amount of check-in activity and found strong evidence of avoidance behavior. People do change their activity patterns when severe air pollution happens. After every possible facet of this effect was measured in this paper, we provided clear answers to many hypotheses and extended the environmental justice discussions in China to more sophisticated context.

在当时,中国的空气污染已引发健康问题。理解空气污染的复杂影响对于测度居民暴露风险、评估环境公平、制定公共政策至关重要。但至今为止,关于空气污染如何影响城市活动的研究还十分有限,仅有少数研究提供了证据碎片。这些研究表明,城市活动受空气污染影响而采取的“躲避行为”仅在一定条件下存在。那么在中国,空气污染对城市活动的影响是否广泛存在?这一影响到底有多大、有什么特征、背后又存在哪些因素和机制?

为了填补这一空白,我们采用2015-2016年的微博签到数据作为城市活动的一种表征,基于251个地级市的面板数据,全面分析了空气污染对城市活动的影响。我们发现,在控制了天气、节假日和城市固定效用等因素之后,空气污染对城市活动存在着显著的负面影响,人们确实会在污染加重时调整活动模式。我们又系统分析了这一影响背后的因素和机制,为许多研究假设提供了明确的实证证据,并深入讨论了其对中国的污染暴露风险评估、环境公平、经济发展的政策含义。


Results

Pollutants

The result shows that avoidance behavior generally exists with respect to many pollutants. SO2 has the largest impact on urban activity, followed by PM2.5, NO2, and PM10, while CO and O3 seem to have little impact. An increase of 100 μg/m3 of the four pollutants will lead to an average decrease of 8.4-27.6% in the total local activities. Taking PM2.5 as an example, in the study period the total accumulative loss of the local activities across the country is estimated to be as large as 6.7%.

我们的研究表明,空气污染对城市活动确实存在着显著影响,且随不同污染物而所有差异。从回归系数来看,SO2的影响最大,随后是PM2.5、NO2和PM10,而CO和O3的影响则不太显著。前四种污染物浓度每增加100 μg/m3平均会导致本地居民活动量下降8.4-27.6%。以PM2.5为例作累积计算,在研究时段内全国因空气污染而减少的城市活动高达6.7%,如果(武断的)换算成城镇社会消费额相当于1.7万亿人民币/年。

People/Activity types

The impact differs dramatically with groups of people and activty types. Every increments in the concentration of the four pollutants (PM2.5, PM10, NO2 and SO2) causes a decrease in local citizens’ activities at least 4 times of visitors’ activity. The activity with higher spatial-temporal flexibility can be more easily re-arranged or canceled to reduce exposure risk, which leaves the more fixed activities, such as work and public transportation, to be more affected by air pollution.

空气污染对城市活动的影响还随人群和活动类型而不同。外来访客对空气污染敏感度更低,平均来看前四种污染物的单位变化导致的本地居民活动下降是外来访客活动下降的四倍以上。活动类型上看,当空气污染发生时,时空约束越低的活动更容易被重新安排或取消来降低暴露风险,这就使得更“固定”的活动(如上班、乘坐公共交通)成为暴露风险的主要来源。

Heterogeneity

Increasing income and deteriorative air quality could lead to higher sensitivity to pollution, while national Holidays tend to weaken sensitivity. People who live in a more polluted city with 100 unit increments in AvgAQI show on average the same sensitivity as those who live in a less polluted city and earn about 20.3 thousand yuan more in average income. The spatial heterogeneity also shows clear patterns: (1) the most developed area with the highest sensitivity; (2) the North, Northeast China and Middle China are next; (3) and the non-capital cities in the broader South and Northwest China have an insignificant correlation or significant but small correlation.

收入水平和平均空气质量是这一影响背后最重要的两个因素,法定节假日是稍弱的因素。收入水平越高、平均空气质量越差,对空气污染越敏感。一个住在平均空气质量更差(平均AQI每增加100)的城市的居民与一个住在空气更干净、但收入更高(职工平均工资高2.03万元)的城市的居民有着差不多的敏感度。在这两个因素综合作用下,空间上也呈现几大特征:(1) 最发达的长三角地区和大部分省会城市对空气污染最敏感;(2) 华北、东北和华中地区次之;(3) 华南地区非省会城市最不敏感。

Injustice

Environmental injustice in terms of air pollution, whether in China or in other countries, is not just about the difference in exposure risk measured by pollutant concentration and population distribution, rather it should take urban activity into consideration. Ignoring air pollution's imapct on activity may lead to an 4% underestimation of its consequences in underdeveloped area and an 11% overestimation in developed cities. The worst scenario would be that an economic slowdown occurs together with the absence of public health policy and the increase of air pollution in the Middle and broader South China.

我们认为,无论是在中国还是在其它国家,空气污染有关的环境公平不应只根据人口分布计算的暴露风险,还应该考虑人们的活动。忽视空气污染对城市活动的影响会导致暴露风险评估结果不准确。我们的研究表明,在发达地区有约11%的过高估计、在欠发达地区则有约4%的低估。最坏的情况将是,如果欠发达地区(如华中和华南的大多数城市)持续接收低端产业转移导致空气质量恶化,而收入又没有快速提高到使其意识到污染危害的程度,再加上公共健康政策欠缺,则会对这些地区的居民产生比一般估计的大得多的健康影响。

Live in a city


以下散点图示意各因素综合影响下的空气污染副作用

Link: Learn more via MIT Senseable City Lab.

Ref: YAN Longxu, DUARTE Fábio*, WANG De, ZHENG Siqi, RATTI Carlo. Exploring the effect of air pollution on social activity in China using geotagged social media check-in data[J]. Cities, 2019,91: 116-125.