How Does Straw Burning Affect Urban Air Quality in China? Shiqi (Steven) Guo The Graduate Institute of International and Development Studies, Geneva September 2017, UNU-WIDER
Effects of Air Pollution Health mortality rate in US (Chay and Greenstone, 2003), Indonesia (Jayachandran, 2009), China (Tanaka, 2015; He et al., 2016), India (Greenstone and Hanna, 2015), South Korean (Jia and Ku, 2016), Mexico (Arceo et al., 2016), Brazil (Rangel and Vogl, 2017) life expectancy in China (Chen et al., 2013) mental health in China (Zhang et al., 2017) Individual performance agricultural worker productivity in US (Graff Zivin and Neidell, 2013) cognitive performance in Israel (Ebenstein et al., 2016) investment performance in China (Huang et al., 2016) Labor market labor supply in Mexico (Hanna and Oliva, 2015) Consumption air purifiers in China (Ito and Zhang, 2016) particulate-filtering masks in China (Zhang and Mu, 2017)
Straw Burning in China fuels, forages, fertilizers changes in rural economy (energy structure, farm mechanization, rural labor) clear the fields in time for the next plantings
Straw Burning in China The day of burning straw, is the day when you will be in prison. 7 days detention and 1000 RMB fine for straw burning 15 days detention and 3000 RMB fine for straw burning Banning straw burning is patriotism.
Environmental Literature Research areas causal link, general effect Emission factors (Cao et al.,2008; Huang et al., 2012; Zhang et al.,2016) Co-movement of air pollution and straw burning (Li et al., 2008; Zha et al., 2013) Meteorological models (Yamaji et al., 2010; Cheng et al., 2014; Zhong et al., 2017) Microstructure of pollutants (Li et al., 2010) Case studies with severe pollution scenarios overestimate Mount Tai, June 2006 (Yamaji et al., 2010); Beijing, 12-30 June 2007 (Li et al., 2010); Shanting, 14-27 June 2010 (Zha et al., 2013); Chengdu, 18-21 May 2012 (Chen and Xie, 2014); Huai River Basin, October 2015 (Zhong et al., 2017)
Overview 1 Data 2 Main Effects temporal effect density effect spillover effect 3 Heterogeneous Effects main pollutants pollution levels 4 Robustness Check samples models randomly generated burning
Data Straw Burning Ministry of Environmental Protection (MEP) of China various satellites: 10:30, 13:30, 14:30 16:30 14,528 fire points in 26 October 2014 31 December 2016 Satellites Data Availability Urban Air Quality MEP: 1,496 ground monitoring stations Air Quality Index (AQI), PM2.5, PM10, SO2, NO2, CO, O3 142 cities at first, 284 cities in 2016 Weather tianqi.2345.com maximum temperature, minimum temperature, smog, rain, sun, cloud, overcast, wind Observations: 284 prefectural-level cities 538 days
Straw Burning
Air Quality
Straw Burning And Air Quality over Time
Summary Statistics Variable Mean Median St.d Min Max Description AQI 68.35 59.17 39.69 5 500 Air Quality Index PM2.5 44.38 35.4 37.47 2 1793 Fine particles 2.5µm in diameter in µg/m3 PM10 79.49 64 73.31 3 8775 in µg/m3 SO2 21.13 15.5 20.77 1 739.2 in µg/m3 CO 1 0.88 0.55 0 18.94 in mg/m3 NO2 28.71 25.17 16.26 1.8 461 in µg/m3 O3 107.4 101 47.02 2.25 863 in µg/m3 Fire 0.1 0 1.5 0 169 Number of straw burning fire points Fired 0.02 0 0.15 0 1 Straw burning dummy Htemp 22.44 25 9.63-27 43 Maximum temperature in degrees Celsius Ltemp 13.09 15 10.16-40 31 Minimum temperature in degrees Celsius Smog 0 0 0.06 0 1 Smoggy day dummy Rain 0.39 0 0.49 0 1 Rainy day dummy Sun 0.31 0 0.46 0 1 Sunny day dummy Cloud 0.5 1 0.5 0 1 Cloudy day dummy Overcast 0.16 0 0.37 0 1 Overcast day dummy Wind 0.38 0 0.49 0 1 Windy day dummy
Main Effects Temporal effect How does straw burning affect urban AQI in the following days? Density effect number of fire points in the city-date grids Spillover effect How does straw burning affect urban AQI of the surrounding cities?
Temporal Effect AQI i,t = τ=15 τ=0 b τ Fired i,t τ + W i,t γ + u i + v t + w i,t Fired i,t : whether there exists straw burning in city i on day t W i,t : weather covariates u i, v t : city, date fixed effects s.e. clustered at city level
Temporal Effect Obs = 126,106; R-squared = 0.2889 AQI Helsinki: 22
Temporal Effect
Density Effect Linear: number of fire points detected in city i on day t AQI i,t = τ=10 τ=0 b τ Fire i,t τ + W i,t γ + u i + v t + w i,t Categorical: number of fire points in {1}, [2,4], [5,+ ) AQI i,t = τ=10 τ=0 τ=10 b τ FireD1 i,t τ + τ=0 Quadratic: linear and quadratic terms AQI i,t = τ=10 τ=0 τ=10 b τ Fire i,t τ + τ=0 τ=10 b τ FireD2 i,t τ + τ=0 b τ FireD3 i,t τ +W i,t γ + u i + v t + w i,t a τ Fire 2 i,t τ + W i,t γ + u i + v t + w i,t
Density Effect (1) (2) (3) (4) (5) (6) Models Linear Categorical Quadratic Average AQI 68.35 68.35 68.35 1 point 2-4 points 5 points linear terms quadratic terms Fire t 0.28** 0.17-2.18* -0.83 0.14-0.0001 Fire t 1 0.92*** 3.33*** 5.09*** 16.59*** 1.40*** -0.007*** Fire t 2 0.68*** 3.56*** 5.10*** 13.83*** 1.08*** -0.006** Fire t 3 0.17*** 3.64*** 4.43*** 3.25** 0.41*** -0.004*** Fire t 4-0.02 2.99*** 2.35* 6.81*** 0.47*** -0.008*** Fire t 5 0.19 3.24*** 2.46* 4.58** 0.52*** -0.006*** Fire t 6 0.16 1.60* 4.75*** 0.80 0.29-0.003 Fire t 7 0.34*** 2.90*** 4.10*** 10.69*** 0.91*** -0.009*** Fire t 8 0.05 1.87** 4.87*** 5.79*** 0.56*** -0.008*** Fire t 9 0.002 1.80* -0.36-0.78 0.14-0.003** Fire t 10 0.11 2.13** 1.15 4.23* 0.22-0.002 s.e. (0.05,0.18) (0.78,1.06) (1.10,1.67) (1.59,2.78) (0.15,0.26) (0.001,0.003) City, date FE Yes Yes Yes Weather Yes Yes Yes Observations 126,106 126,106 126,106 R-squared 0.3449 0.3465 0.3460 Number of cities 284 284 284
Spillover Effect AQI i,t = τ=10 τ=0 τ=10 b τ Fired i,t τ + τ=0 τ=10 + τ=0 τ=10 b τ FiredR1 i,t τ + τ=0 b τ FiredR2 i,t τ b τ FiredR3 i,t τ + W i,t γ + u i + v t + w i,t Fired i,t : whether exists straw burning in city i on day t FiredR1 i,t : whether exists straw burning in other cities within 200 km from city i on day t FiredR2 i,t : 200 km - 400 km FiredR3 i,t : 400 km - 600 km
Spillover Effect (1) (2) (3) (4) Distance 0 km 0-200 km 200-400 km 400-600 km (Helsinki) (Turku) (Stockholm) (Oulu) Number of other cities 0 7 18 25 Fired t -0.22-1.14*** 0.63** -0.10 Fired t 1 4.50*** 1.30*** 1.56*** 1.35*** Fired t 2 4.48*** 1.10*** 1.65*** 0.69** Fired t 3 3.60*** 1.18*** 0.62** 0.05 Fired t 4 2.81*** 1.77*** 0.53* -0.56** Fired t 5 3.47*** 0.42-0.54* -1.30*** Fired t 6 2.93*** 0.11-0.82*** -0.62** Fired t 7 3.82*** 1.45*** 0.43-0.54** Fired t 8 3.10*** 0.64 0.40-0.32 Fired t 9 1.33** -0.26-0.03-0.51* Fired t 10 2.35*** -0.51 0.22-0.07 s.e. (0.64, 1.00) (0.37, 0.43) (0.28, 0.37) (0.24, 0.32) City FE, date FE, weather Obs = 126,106; cities = 284; R-squared = 0.3470 Yes
Heterogeneous Effects Main pollutants PM2.5, PM10, SO2, CO, NO2, O3 Pollution levels quantile regression Regions Northeast, North, Central and South China Seasons
Main Pollutants
Main Pollutants Emission factors (Cao et al., 2008) Wheat straw Rice straw Corn stover Cotton stalk PM 8.8 6.3 5.3 4.5 NO 2 0.4 0.3 0.3 0.2 SO 2 0.04 0.2 0.04 0 CO 58 68 68 106 (in g/kg) O 3 (Yamaji et al., 2010; Zhong et al., 2017) PM 10 by 10-15 µg/m 3 from rice residue in Eastern Spain (Viana et al, 2008) PM 10 and O 3 from sugarcane in Brazil (Rangel and Vogl, 2017)
Pollution Levels
Robustness Check Different samples missing days, no-burn days, year 2016, early cities, no-burn cities Different models dynamic model (Difference GMM) random coefficient model Panel Vector Autoregressive (Panel VAR) model Randomly generated burning same number of straw burning grids in every month, all over China
Different Samples (1) (2) (3) (4) (5) Sample + missing days - no-burn days Year 2016 Early cities + no-burn cities Cities 284 284 284 142 367 Days 798 386 335 538 538 Fired t 0.28-1.28-0.42-1.29 2.20 Fired t 1 5.94*** 6.95*** 5.50*** 4.52*** 7.81*** Fired t 2 5.79*** 8.03*** 5.25*** 5.86*** 5.96*** Fired t 3 4.77*** 7.20*** 3.92*** 6.21*** 4.76*** Fired t 4 3.83*** 5.27*** 3.26*** 5.32*** 3.83*** Fired t 5 3.83*** 5.23*** 2.95*** 6.30*** 4.06*** Fired t 6 3.19*** 4.14*** 1.16*** 4.28*** 3.31*** Fired t 7 4.41*** 5.79*** 2.49*** 5.56*** 4.61*** Fired t 8 3.63*** 4.68*** 1.75*** 4.94** 3.76*** Fired t 9 1.27** 0.92*** -1.74*** 2.36*** 1.11*** Fired t 10 2.38*** 3.36*** 1.10 3.83*** 2.95*** s.e. (0.6,1.1) (0.9,1.4) (0.6,1) (0.8,1.5) (0.7,1.1) Weather Y Y Y Y City, Day FE Y Y Y Y Y Observations 200,233 40,118 84,996 64,748 153,397 R-squared 0.35 0.24 0.32 0.35 0.23
Panel Vector Autoregressive model Rain i,t π 11j π 12j π 13j π 14j π 15j π 16j Rain i,t j u 1i v 1t w 1i,t Sun i,t 15 π 21j π 22j π 23j π 24j π 25j π 26j Sun i,t j u 2i v 2t w 2i,t Cloud i,t Wind i,t = π 31j π 32j π 33j π 34j π 35j π 36j Cloud i,t j j=1 π 41j π 42j π 43j π 44j π 45j π 46j Wind i,t j + u 3i u 4i + v 3t v 4t + w 3i,t w 4i,t Fire i,t π 51j π 52j π 53j π 54j π 55j π 56j Fire i,t j u 5i v 5t w 5i,t AQI i,t π 61j π 62j π 63j π 64j π 65j π 66j AQI i,t j u 6i v 6t w 6i,t
Impulse Responses
Impulse Responses All responses
Random Generated Burning
Conclusion Straw burning increases the urban AQI by 6.8 on the first two days after burning. The effect decreases gradually and remains significant for eleven days. Each fire point increase urban AQI by 0.9 on the first day after burning. The effect is larger with denser burning. The marginal effect is decreasing. Cities 400 to 600 km away are also affected. Heterogeneous effects are found with different pollutants, pollution levels, regions and seasons. Effects are robust with different sub-samples and models.
Thank you! Email: shiqi.guo@graduateinstitute.ch Webpage: https://sites.google.com/site/stevenshiqiguo/shiqi-guo
Regions (1) (2) (3) (4) (5) (6) Regions Northeast North Central, South Cities 46 56 129 Average AQI 70.1 103.4 67.3 Average Fire 0.3 0.06 0.01 Straw burning Dummy Number Dummy Number Dummy Number Fire t 1.17 0.24** -1.63 0.4 2.41** 0.64 Fire t 1 7.74*** 0.81*** -0.22 0.72*** 5.9*** 2.21*** Fire t 2 4.95*** 0.47*** 2.59** 0.42*** 4.08*** 1.27** Fire t 3 5.54*** 0.07 3.13*** 0.27* 2.43** 0.34 Fire t 4 1.91-0.08 3.93*** 0.6*** 0.81-0.13 Fire t 5 1.51 0.11 4.41*** 0.78*** 0.84-0.06 Fire t 6 2.06-0.01 3.42*** 0.04 2.05* 0.67* Fire t 7 2.66** -0.01 2.92*** 0.48** 1.06 0.61** Fire t 8 3.21*** -0.21** 2.68** 0.42 0.33-0.37 Fire t 9 1.4 0.01 1.02 0.48*** -0.22-0.64** Fire t 10 2.07** 0.09 2.61** 0.78*** -1.13-0.9*** s.e (1,1.7) (0.06,0.19) (1,1.4) (0.12,0.3) (0.8,1.4) (0.2,0.6) Weather Y Y Y Y Y Y City, day FE Y Y Y Y Y Y Observations 32,267 32,267 40,482 40,482 91,389 91,389 R-squared 0.5042 0.5036 0.5965 0.5963 0.4562 0.4562 Northeast: Heilongjiang, Jilin, Liaoning, Neimenggu; North: Hebei, Henan, Shandong, Shanxi; Central and South: Hubei, Hunan, Sichuan, Chongqing, Yunnan, Jiangsu, Zhejiang, Anhui, Jiangxi, Fujian, Guangdong, Guangxi
Seasons
Seasons (1) (2) (3) (4) Months Mar-May Jun-Aug Sep-Nov Dec-Feb Average AQI 81.8 60.3 79.9 109.9 Average Fire 0.09 0.03 0.09 0.003 Fired t -0.76 1.88 0.03-17.46*** Fired t 1 2.97*** 3.13** 9.54*** -8.83** Fired t 2 4.07*** 1.41 9.34*** -1.16 Fired t 3 1.17 2.4*** 8.45*** 10.01*** Fired t 4 0.61 4.93*** 6.12*** -2.14 Fired t 5 1.65* 6.62*** 3.95*** -5.12 Fired t 6-0.23 4.26*** 5.89*** -5.73 Fired t 7-0.29 2.66*** 10.31*** -3.2 Fired t 8 0.86 4.02*** 5.77*** 1.06 Fired t 9-2.45*** 3.68*** 4.76*** -8.19* Fired t 10 0.82 3.8*** 5.01*** -14.59*** s.e. (0.7,1.1) (0.8,1.3) (1.1,1.6) (3.9,5.2) Weather Y Y Y Y City, Day FE Y Y Y Y Observations 51,497 50,523 52,567 45,788 R-squared 0.2192 0.1883 0.3202 0.2796
Random Coefficient Model
Dynamic Panel Model (1) (2) Models FE Arellano-Bond L.aqi 0.61*** 0.52*** (0.01) (0.009) L2.aqi -0.06*** -0.12*** (0.006) (0.005) Fire 0.21 1.52* l1fire 6.21*** 6.86*** l2fire 2.57*** 4.18*** l3fire 1.85** 3.91*** l4fire 1.88** 3.47*** l5fire 2.16*** 3.41*** l6fire 0.69 1.6** l7fire 1.73** 2.35*** l8fire 1* 1.29* l9fire -0.77-0.89 l10fire 2.1*** 1.22 s.e. (0.58,0.96) (0.68,1.08) Weather Y Y City, Month FE Y Y Cubic Trend Y Y Observations 199,345 198,690 R-squared 0.5024 -
All Impulse Responses Impulse Responses
Satellites Data Availability Data