Beijing Blue: Impact of the 2008 Olympic Games and 2014 APEC Summit on Air Quality

Abstract

This paper examines the effects of interventions to reduce air pollution during two international events on air quality in Beijing and its neighbor cities. Air quality data were gathered from China’s Ministry of Environmental Protection, meteorological data from the China Meteorological Administration and economic data from the China Statistical Yearbook. The paper uses fixed-effect panel data models to empirically evaluate air quality improvement in Beijing and other affected cities before, during, and after the 2008 Olympic Games and the 2014 Asia-Pacific Economic Cooperation summit. Results show substantial improvement in air quality in Beijing and neighboring cities during the two events. However, some of the air quality improvement achieved reverted within a year after the games and within a week after the summit. Furthermore, the improvement achieved during the summit completely reverted and air quality deteriorated severely five days after the summit. It is also found that air quality in China, at least in the cities included in this study, gradually improved over the past 15 years or so. The findings suggest that sustainable interventions and incentive-based programs to reduce emissions from industry production and traffic are the key to maintaining the air pollution reduction achieved during the events.

Introduction

The Chinese government has been implementing environmental interventions to improve air quality through the Law of the People's Republic of China on the Prevention and Control of Atmospheric Pollution, established in 1989 and amended in 2000 and 2015. However, such policy instruments usually are not sufficient to allow China to hold major international events in Beijing without concerns being raised about its air quality. Therefore, the central and the local governments took specific air-quality improvement actions that temporarily target Beijing and the surrounding areas to prepare for major international and national events.

A series of measures were undertaken between 2000 and 2008 to improve the air quality for the 2008 Olympic Games in Beijing. These measures included plant relocations and closures, traffic controls, and reduction of industrial use of coal in and around Beijing and co-host cities. In November 2014 and August 2015, China took similar actions, albeit on a smaller scale to prepare for the 2014 Asia-Pacific Economic Cooperation (APEC) summit and the 2015 China National Day Parade in Beijing. Such actions resulted in consistent blue-sky days in Beijing during the APEC summit and Parade Day. The blue sky1 that appeared during these two events is called APEC blue, Beijing Blue or man-made blue.

These event-specific interventions improved the air quality in Beijing during the events. However, the impact was not long lasting since most of the measures were implemented only before and during the events. This raises questions about the effectiveness and efficiency of such measures on improvements in public health and overall environmental conditions. As many international and national events are held in countries with concerning air quality, understanding if these measures were worthwhile and/or sustainable becomes very important. The purpose of this paper is to examine the impact of such interventions during and after two specific international events, the 2008 Olympic Games and the 2014 APEC summit. While intervention measures were taken in both events, the magnitude, duration, and specifications differ. Thus, the study of both events allows for examination of differences in air quality improvement, if any, in response to the measures.

The 2008 Olympic Games in Beijing is a special event to study the effects of certain environmental policy interventions. Several studies have examined the effects of this event on air quality. Wang et al. (2009a) used self-collected particulate matter (PM)10 and PM2.52 data (see the United States Environmental Protection Agency (2022) for definition of PM measurement) at Beijing University between July 28 and October 7, 2008, and found that on average PM10 and PM2.5 concentrations were lower during the games than during a non-Olympic period. Using a set of on-road emission measures including black carbon (BC) concentration in 2007 and 2008, Wang et al. (2009b) showed that traffic control during the 2008 Olympic Games substantially reduced BC concentration and other vehicle-related emissions. Viard and Fu (2015) used officially published air pollution index (API) data from 2007 to 2009 to investigate the effect of traffic control on air quality. They also found decreased air pollution during traffic restriction days.

Chen et al. (2013) conducted a more comprehensive study of the impact of the 2008 Olympic Games on air quality. Given concerns raised about the quality and potential manipulation of the official air pollution data published by the government, they used both API and aerosol optical depth (AOD) data from The National Aeronautics and Space Administration (NASA) to measure air quality in Beijing. They found better air quality during the games, but such effects virtually disappeared about a year later. Their findings also suggest correlations between air quality and the timing and locations of restrictions on plant production and traffic. Instead of focusing on air quality improvement, He et al. (2016) examined the effects of better air quality during the games on mortality rate and found that an eight-percent decrease in mortality rate was associated with a ten-percent reduction in PM10. Shi et al. (2020) studied the effect of temporary environmental measures during Chinese President Xi Jinping’s visits to different cities between 2013 and 2017. The study reported decreases in PM2.5 and PM10 measures, but no significant changes in other pollutants during the visits.

This study is broadly related to the literature on environmental policy and air pollution. There have been extensive studies of various environmental policies in advanced economies (Grey & Shimshack, 2011). In recent years, with more data in developing countries becoming available to the public, research on such countries has emerged and the results are mixed. For example, Kumar and Foster (2009) found that air pollution regulations in Delhi failed to improve air quality. In addition, such regulations adversely affected the air quality in the surrounding areas by crowding out pollution to less regulated areas. On the other hand, a study by Foster et al. (2009) found improved air quality due to voluntary pollution reduction by plants in Mexico. Albulescu et al. (2022) used the Organization for Economic Co-operation and Development (OECD) data to explore the impact of environmental policy of 32 countries between 1990 and 2015. The study showed the importance of targeted policies. Wolde-Rufael and Eyob (2021) examined seven emerging economies between 1994–2015 to test the effectiveness of policy measures in improving air quality. In their 2022 study, Lou et al. investigated the impact of coronavirus 2019 induced environmental restrictions in ten countries and found a reduction in pollutant concentration in all ten countries.

Policy Interventions

On July 13, 2001, the International Olympic Committee announced Beijing as the host of the 2008 Olympic Games. Between 2001 and the start of the games in 2008, a series of air cleaning actions were undertaken on a large scale in Beijing and the surrounding areas. Such actions include polluting plant relocation, closure and temporary shut downs, traffic control, and more stringent emission standards.

By 2008, about 140 polluting companies in Beijing and its suburban areas were relocated and several major polluting plants were permanently closed (Chen et al., 2013). Then during the Olympic Games and the Paralympic Games (8/8/2008-9/17/2008), major polluting companies were required to either temporarily shut down or reduce production (Beijing Municipal Environmental Protection Bureau, 2009).

Starting from March 2008, more than 10,000 heavy-emission cars, the so-called yellow-label vehicles, were retired and more than 2000 buses and taxies were renovated. Between July 20, 2008, and September 20, 2008, the odd-even rule was implemented such that vehicles with a registration number ending in even (odd) number were allowed on the road only on even (odd)-numbered calendar days. During the same period, yellow-label vehicles were banned from the roads.

Neighboring provinces, Hebei, Shanxi, Neimenggu and Shandong, and a municipality, Tianjin, were also required to take measures to reduce air pollution, given that they could have contributed to the air pollution in Beijing. Co-host cities, including Tianjin, Shenyang, Qingdao, and Shanghai, also took actions to improve air quality, such as temporary closures or production reductions at polluting plants.

In 2014, when Beijing was to hold the APEC summit from November 10 to November 12, it again implemented temporary measures to reduce air pollution in addition to its normal environmental policies. The measures taken were similar to those implemented during the Olympic Games, but at a much smaller scale. Beijing, Tianjin, and surrounding provinces were all involved. One new measure taken during this event was that employees of the government, semi-government organizations and institutions were given a six-day APEC holiday from November 7 to November 12. This substantially reduced the amount of traffic on the roads.

Other than the specific measures taken to prepare for the two major events, China, especially Beijing, implemented a series of long-term environmental protection programs throughout the whole study period. Beijing has undertaken 13 phases of comprehensive air pollution control programs since 1998 and implemented the 14^th phase in 2008. According to a review by the United Nations Environment Programme (UNEP), Beijing has greatly reduced its dependence on coal for its energy needs, reducing from a peak of 9 million tons in 2005 to 6.44 million tons in 2013. In addition, Beijing also reduced total vehicle emissions by more than 40% from 1998 to 2013 and promoted public transportations, providing 48% of total transportation in 2014, in comparison to 26% in 2000 (UNEP, 2016).

Data and Model Specification

Data

Daily API data for major cities are published by China’s Ministry of Environmental Protection (MEP). API is an aggregated measure of the maximum intensities among NO₂, SO₂, and PM 10. The daily average intensities of monitored pollutants are first transformed nonlinearly into individual scores. Then the highest score is reported as the daily API together with the dominant pollutants. Thus, a higher API indicates heavier pollution. Table 1 provides a detailed classification of the API scores, air pollution levels, and the corresponding health effects. According to the MEP, air quality with API scores below 100 is typically considered to be good or excellent and safe for daily outdoor activities. API scores between 101 and 150 indicate slightly polluted air, 151-200 lighted polluted, 201-300 moderately polluted and API above 300 heavily polluted. For API scores above 300, all individuals should remain indoors and avoid unnecessary outdoor activities.

Table 1.

API Classifications

API	Air Pollution Level	Health Implications
0 - 50	Excellent	No threats to health.
51 -100	Good	No threats to health except for extremely vulnerable people.
101-150	Slightly Polluted	Minor allergy may occur; individuals with breathing or heart problems should reduce outdoor activities.
151-200	Lightly Polluted	Light allergy may occur; individuals with breathing or heart problems should reduce outdoor activities.
201-300	Moderately Polluted	Healthy people will be noticeably affected. Individuals with breathing or heart problems and the elders should remain indoors and restrict activities.
301+	Heavily Polluted	Healthy people will experience severe adverse effects including strong allergy and breathing difficulties that can trigger other symptoms. All people should remain indoors and avoid unnecessary outdoor activities.

Source: China’s Ministry of Environmental Protection (2008-2014) Data Center

Recognizing the limitation of the old API in reflecting the true air quality, China shifted from API to an air quality index (AQI) in 2013. The main difference between AQI and API is the pollutants included. In API, pollutants included are SO₂, NO₂, and PM10, while AQI takes into consideration SO₂, NO₂, PM10, PM2.5m, Q3, and CO. The calculation of AQI is based on more stringent standards, which are more similar to the standards used in the United States, and thus results in relatively higher index values than using the API standards. Although AQI is calculated differently, the classification of the index stays the same.

Since the change in the air quality measure occurred in between the two events considered in this study, it is not feasible to combine the API and AQI data in one sample period. Therefore, the data were segmented into two sample periods: the 2008 Olympic Games period, running from 1/1/2001 to 12/31/2012, and the APEC Summit period, running from 1/1/2014 to 12/31/2014. To avoid confusion, API will be used to refer to both API and AQI in this paper.

The selection of the study cities and the time frame is mainly constrained by the availability of the API data. Overall, 35 major cities in mainland China were included, as shown in Fig. 1. The cities were classified into four cohorts. Beijing is a cohort in and of itself since it was the host city of the two events and the target city of most of the policy interventions. The second cohort consists of the neighboring cities, including Huhehaote, Taiyuan, Shijiazhuang, and Tianjin. In addition, Shenyang, Tianjin, Qingdao, and Shanghai co-hosted the Olympic Games and thus are classified as the third cohort, co-host cities. The fourth or last cohort is the control group, including all other cities in the sample.3 In the empirical analysis, the impact of the interventions on air quality in Beijing, neighboring cities and co-host cities was examined by comparing air quality in these cities with air quality in cities in the control group while controlling for other factors that can affect the air quality.

Fig. 1.

Major cities in mainland China. Cities included in the study: Host city: Beijing; Neighboring cities: Huhehaote, Taiyuan, Shijiazhuang, and Tianjin. Co-host cities for the Olympic Games: Shenyang, Tianjin, Qingdao, Shanghai. Control Group cities: Changchun, Changsha, Chengdu, Chongqing, Dalian, Fuzhou, Guangzhou, Guiyang, Hangzhou, Haerbin, Haikou, Hangzhou, Hefei, Jinan, Kunming, Lanzhou, Nanchang, Nanjing, Nanning, Ningbo, Shenzhen, Wuhan, Wulumuqi, Xi’an, Xining, Yinchuan, and Zhengzhou

Source: PNGwing (2023), accessed March 20, 2023

Table 2 provides the summary statistics of the API for each cohort during various periods. Overall, the API in all cohorts behaves similarly across the sample periods. The neighboring cities have the highest average API (125.9), followed by Beijing with an average API of 112.4 during the baseline period (1/1/2001-12/31/2001). Then the API of all cities declined slightly during the preparation period and dropped substantially during the games. The API in Beijing decreased by 46%, while the API in the neighboring cities declined by 43% during the games in comparison to the preparation period. Within six months after the games, the API in the co-host and control cities almost reverted to the levels during the preparation period. The API in Beijing and neighboring cities remained much lower than the levels before the games. This could be the continued benefits of the interventions. During the next 6 months, the API in all cities continued to decline. However, in comparison with other cohorts, the decline in API in Beijing was relatively small. The API in Beijing declined from an average of 83.71 to 78.64, a decrease of about 5 points on average. During the same periods, the decrease in the API in the other three cohorts ranged from of 12 points to 17 points.

Table 2.

Average API by city groups and period

Period	Beijing	Neighbor	Co-host	Control
Baseline period (1/1-12/31/2001)	112.4	125.9	95.53	81.17
Preparation	103.1	92.71	81.55	75.92
During Olympic	54.88	52.44	59.19	58.96
1-6 months after Olympic	83.71	82.86	79.55	76.99
7-12 months after Olympic	78.64	65.48	67.96	62.79
Baseline Period (1/1-11/4/2014)	122.8	116.7		86.11
During APEC	72.83	96.83		90.97
1-5 days after APEC	73	81.80		99.98
6-10 days after APEC	192.4	203.9		121.2
11+ days after APEC	105.5	141		98.5
Number of cities	1	4	4	26

AQI was adopted in 2013 in place of API. Thus, the average API around the Olympic Games is not directly comparable with that around the APEC summit. Authors’ calculation based on data from China’s Ministry of Environmental Protection (2008-2014). Tianjin is both a co-host city and a neighbor city. It was treated as a co-host city in the Olympic Games analysis and as a neighbor-city in the APEC analysis

The air quality also showed certain improvements during the second event, the APEC summit. The API in Beijing and neighboring cities temporarily decreased during the summit and within five days after the event. However, the API increased dramatically within 6-10 days after the summit, reaching the highest levels of 192 and 203.9 for Beijing and the neighboring cities. In comparison, the API in the control cities continued to increase until about ten days after the APEC.4

Data on temperature, humidity, wind, and pressure were collected from the China Meteorological Administration (CMA) (2008-2011) and National Centers for Environmental Information (NCEI) (2014). Further, to control for socioeconomic factors, data on gross domestic product (GDP) growth, population density and industrial production by city and year were collected from the China Statistical Yearbook (National Bureau of Statistics of China, 2008-2014).5

Table 3 shows the summary statistics for the control variables discussed herein. The climate data for the two sample periods, although collected from different sources, are very similar when the units are converted.

Table 3.

Summary statistics of control variables used in regression analysis

Variable	Unit	All		Beijing		Neighbor cities		Control cities		Co-host cities
		Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.	Mean	Std. Dev.
2001-2012
API	-	76.49	39.04	96.03	55.33	87.58	45.94	74.1	37.49	78.96	35.34
Temperature	F	58.8	51.96	55.94	52.14	52.05	53.08	60.21	51.42	55.38	52.95
Humidity	Dew Point in F	47.04	57.56	38.81	23.51	37.39	92.70	49.00	53.09	43.72	24.00
Wind velocity	Miles/hour	4.92	2.55	5.14	2.17	3.67	1.88	4.89	2.53	6.10	2.75
Air pressure	Millibars	970.7	66.5	1,012.5	10.1	943.3	47.4	965.6	70.9	1,014.3	9.8
2014
API	-	76.31	39.37	120.52	75.77	119.94	73.44	88.21	50.37
Temperature	F	58.9	19.85	56.69	19.79	53.91	20.51	59.64	19.66
Humidity	Dew Point in F	45.79	22.91	35.62	24.95	34.77	23.08	47.6	22.31
Wind velocity	Miles/hour	6.12	4.14	5.96	2.97	5.3	2.55	6.22	3.23
Air pressure	Millibars	993.86	73.57	1,016.83	9.75	1,017.71	10.22	989.77	78.83
2001-2014
GDP growth	-	16.16	8.32	17.08	12.94	18.48	14.37	15.83	6.9
Industrial production	million RMB	17.68	19.81	33.54	18.28	16.87	20.12	17.26	19.65
Population density	person/sq. miles	16.12	10.35	18.84	1.21	13.26	6.54	16.41	10.86
Heating	-	0.23	0.42	0.33	0.47	0.33	0.47	0.21	0.41
Number of cities	-	34		1		4		26		4

Authors’ calculation based on data from China’s Ministry of Environmental Protection (2008-2014), China Meteorological Administration (2008-2011), National Centers for Environmental Information (2014), and National Bureau of Statistics of China (2008-2014). Tianjin is both a co-host city and a neighbor city. It was treated as a co-host city in the Olympic Games analysis and as a neighbor-city in the APEC analysis

In all cities, GDP grew at an average rate of 16.16%. The rate for Beijing was slightly higher, at 17.08% on average. The neighboring cities showed an even higher GDP growth rate (18.48%). In contrast, the industrial production in Beijing was almost twice the overall average. This is in line with the fact that the policy interventions focused mostly on reducing industrial production in Beijing and its surrounding areas. As suggested by the mean population density, Beijing is among the most populated cities in China, with a density of almost 19 people per square mile, while the overall population density for all cities is 16.12 per square mile.

Empirical Specifications

Traditional fixed-effects panel data models were estimated to examine the treatment effects. The two events were estimated separately using the two sample periods, 1/1/2001-12/31/2012 and 1/1/2014-12/31/2014. There are two reasons for this decision. First, MEP shifted from API to AQI in 2013. These two measures are not consistent and thus the data cannot be combined. Second, given the long timespan between the two events, separate estimation can help reduce extra noise introduced by the observations between 2010 and 2012.

The fixed effect model takes the following format:

$${\mathit{API}}_{i,d}={\alpha }_i+\beta t+\gamma t^2++\vartheta C_{i}t+\tau MD+\sum _k{\delta }_{B,k}{I}_B{P}_k+\sum _k{\rho }_{NC,k}{I}_{\mathit{NC}}{P}_k+\vartheta S_{i,d}+{\phi M}_{i,d}+\omega H_{i,d}+{ϵ}_{i,d}.$$

Here ${\alpha }_i$ denotes the city-fixed effects; t measures the number of days since 1/1/2001 for the games and 1/1/ 2014 for APEC; $C_i$ is a city dummy and thus $C_{i}t$ denotes city-specific trends; MD denotes month dummies; $I_B$ takes a value of 1 if the city is Beijing and 0 otherwise; similarly $I_{\mathit{NC}}$ is the dummy variable for neighboring cities (as well as co-host cities for the event of the Olympic Games); $P_k$ denotes the various intervention periods with the following specifications.

2008 Olympic Games6

• P1: preparation before Olympic Games, 8/7/2008-1/1/2012

• P2: during the games, 8/8/2008-9/17/2008

• P3: up to six months after the games, 9/18/2008-3/17/2009

• P4: seven to twelve months after the games, 3/18/2009-9/17/2009

• P5: the rest of the period, 9/18/2009-12/31/2012

• Baseline period: 1/1/2001-12/31/2001

2014 APEC Summit7

• P1: during the summit, 11/7/2014-11/12 /2014 (selected based on the APEC holiday)

• P2: up to five days after the summit, 11/13/2014-11/17/2014

• P3: six to ten days after the summit, 11/18/2014-11/22/2014

• P4: the rest of the period, 11/23/2014-12/31/2014

• Baseline period: 1/1/2014-11/4/2014

Socioeconomic conditions are included in $S_{c,d}$; climate factors are denoted by $M_{c,d}$. Lastly, $H_{c,d}$ is the heating dummy, which takes a value of one if the date is between 11/15 and 3/15 and if the city supplies heating during winter. The fixed-effects models were estimated using the least squares method.

Results and Discussion

Results of the fixed-effects panel data regression for the Olympic sample period are summarized in Table 4. The specification in column 1 excludes socioeconomic factors and does not separate neighboring and co-host cities from the rest of the control cities. Column 2 includes socioeconomic factors in addition to what is included in column 1. Column 3 further separates neighboring and co-host cities from other control cities.

Table 4.

Regression results for the Olympic Games model

Variable	(1)	(2)	(3)
Beijing
Preparation	-4.562***	-2.151**	-2.644**
	(0.236)	(1.055)	(1.018)
During Olympic	-36.731***	-34.346***	-35.205***
	(1.138)	(1.939)	(1.824)
1-6 Months after Olympic	-26.132***	-23.720***	-24.332***
	(2.080)	(1.937)	(1.831)
7-12 Months after Olympic	-17.899***	-16.141***	-16.551***
	(1.059)	(1.772)	(1.702)
Rest period	-16.263***	-16.691***	-16.313***
	(0.948)	(1.209)	(1.129)
Neighbor cities
Preparation			-13.258***
			(3.684)
During Olympic			-22.088***
			(3.005)
1-6 Months after Olympic			-13.239***
			(4.632)
7-12 Months after Olympic			-12.584***
			(3.742)
Rest of the period			-6.323
			(4.517)
Co-host cities
Preparation			1.335
			(4.071)
During Olympic			1.127
			(6.891)
1-6 Months after Olympic			1.464
			(1.653)
7-12 Months after Olympic			3.530
			(4.338)
Rest of the period			5.694***
			(1.933)
Other Control Variables
t	-0.003***	-0.011***	-0.010***
	(0.000)	(0.001)	(0.001)
t squared		0.000**	0.000*
		(0.000)	(0.000)
Humidity	-0.001	-0.001	-0.001
	(0.001)	(0.001)	(0.001)
Air pressure	-1.291*	-1.242*	-1.246*
	(0.760)	(0.727)	(0.729)
Temperature	0.075**	0.078**	0.078**
	(0.031)	(0.030)	(0.030)
Wind velocity	-0.395***	-0.396***	-0.396***
	(0.062)	(0.063)	(0.063)
Heating supply	8.504***	8.484***	8.489***
	(2.438)	(2.413)	(2.420)
Population density		0.002	-0.006
		(0.016)	(0.015)
GDP growth		0.063	0.023
		(0.071)	(0.037)
Industrial production		-0.013	-0.014
		(0.020)	(0.019)
R-squared	0.161	0.163	0.164
Number of ids	35	35	35

Robust standard errors in parentheses. Month dummies and city trend were included in all three specifications. Regression results based on data from China’s Ministry of Environmental Protection (2008-2014), China Meteorological Administration (2008-2011, National Centers for Environmental Information (2014), and National Bureau of Statistics of China (2008-2014)

^*p<0.01; **p<0.05; ***p<0.1

Overall, interventions for the Olympic Games imposed significant (p-values<0.01) and positive effects on air quality in Beijing during the games. The overwhelming negative and significant coefficients (p-values<0.01) for During Olympic across the three specifications suggest that the air cleaning actions substantially reduced the API in Beijing. In addition, the magnitudes of the estimates are relatively consistent across the three model specifications. Taking the third specification (Column 3) as an example, the API in Beijing is about 35.2 points lower during the Olympic Games, equivalent to a 31.4% decline from the baseline period. The coefficients for the post-game dummies, 1-6 months after Olympic, 7-12 months after Olympic, and Rest period are also negative and highly significant (p-values<0.01), suggesting that the intervention effects on air quality lasted beyond the game period. However, the API was only 24 points lower within six months after the games and 16 points lower ever since in comparison with the baseline period levels. This suggests that although the intervention effects continued for more than a year after the games, part of the improvement in air quality achieved during the games reverted after the games. This is consistent with the fact that many of the intervention measures were temporary and only targeted the games’ period.

The API in the neighboring cities showed a relatively similar pattern, although with a much lower magnitude. The highest improvement occurred during the games, when the API declined by about 22 points. Then the decrease dropped to around 13 points within a year after the games. The improvement in neighboring cities did not last longer than a year. The coefficient of the rest of the period was negative but not significant (p-value= 0.17), which means 12 months after the games, the API in these cities reverted to the baseline level.

Co-host cities, unlike Beijing and the neighboring cities, showed little improvement in air quality during or after the event. The coefficients for the co-host cities are positive but not significant (p-values=0.74, 0.87, 0.38, 0.42) before, during and within one year after the games. Several factors could have contributed to the results found in these cities. First, the limited number of games held in the co-host cities may have given the city governments much less incentive to take intensive interventions. Second, only one dummy variable was created for all the co-host cities, which captures the average treatment effects of these co-host cities. There could be differential effects across the cities, but the overall effects are averaged out when summarized and captured by the single dummy variable.

The regression results for the APEC sample period are shown in Table 5. The three model specifications are the same as those in Table 4. The coefficients of During APEC and 1-5 days after APEC for Beijing are negative and highly significant (p-values< 0.01). According to Column (3), the API decreased by about 40.8 points during the summit and about 49 points within five days after the event, corresponding to a 36% and a 43% decrease based on the pre-event API level. The highest decline in API appeared within five days after the summit, which implies that it took time for the interventions taken during the event to reach peak effectiveness and be reflected in the air quality measures. However, the air quality completely reversed a week after the summit and further deteriorated. The API in Beijing was 50-54 points higher than that of the baseline period. This could be the result of increased traffic when people returned to Beijing after the APEC holidays and factories started to catch up after the production ban. The findings of these analyses suggest that the impact of the interventions is temporary. Furthermore, such sharp and temporary improvement actually promotes unexpected severe air pollution afterwards, which can be more harmful to human health. After reaching its worst status a week after the summit, the air quality in Beijing regained some improvement, resulting in a decrease of about 23-29 points for the rest of period.

Table 5.

Regression results for APEC model

Variables	(1)	(2)	(3)
Beijing
During APEC	-42.459***	-42.381***	-40.756***
	(7.289)	(7.275)	(8.312)
1-5 days after APEC	-52.601***	-52.826***	-49.367***
	(6.420)	(6.389)	(7.953)
6-10 days after APEC	50.400***	49.872***	54.519***
	(6.796)	(6.830)	(7.572)
Rest of the period	-29.343***	-28.587***	-23.378***
	(6.439)	(6.605)	(8.274)
Neighbor cities
During APEC			-3.208
			(11.301)
1-5 days after APEC			-22.724***
			(7.423)
6-10 days after APEC			85.417**
			(32.103)
Rest of the period			30.206**
			(12.810)
Other Control Variables
t	-0.548***	-0.479**	-0.482**
	(0.170)	(0.178)	(0.177)
t squared	0.001***	0.001***	0.001***
	(0.000)	(0.000)	(0.000)
Humidity	0.562	0.552	0.584
	(0.337)	(0.336)	(0.350)
Air pressure	-0.432	-0.445	-0.415
	(0.291)	(0.292)	(0.280)
Temperature	0.417	0.402	0.401
	(0.289)	(0.292)	(0.298)
Wind velocity	-1.928***	-1.919***	-1.909***
	(0.349)	(0.347)	(0.343)
Heating supply	14.747**	14.590**	11.773**
	(6.167)	(6.149)	(5.599)
Population density		-1.745	-2.059
		(12.649)	(12.430)
GDP growth		-1.000**	-0.967**
		(0.440)	(0.427)
Industrial production		4.954	5.249
		(9.733)	(9.525)
R-squared	0.225	0.226	0.233
Number of ids	34	34	34

Robust standard errors in parentheses. Month dummies and city trend were included in all three specifications. Shanxi was dropped from the estimation due to missing values for air pressure. Regression results based on data from China’s Ministry of Environmental Protection (2008-2014), China Meteorological Administration (2008-2011), National Centers for Environmental Information (2014), and National Bureau of Statistics of China (2008-2014)

^* p<0.01; ** p<0.05; *** p<0.1

Interventions taken during APEC had no significant effect (p-value=0.78) on reducing the API in neighboring cities during the event, but showed negative and significant effects (p-value=0.004) after the event. Two factors could have contributed to this result. First, given the inconvenience in commuting in Beijing during the summit, people who could enjoy the holiday might have chosen to leave Beijing temporarily. Furthermore, the duration of the holiday was not long. Thus, visiting neighboring cities may have been more attractive than traveling further away. This may have caused increased traffic and activities in the neighboring cities during the summit, which may have contributed to the insignificance of the effects during the event. After the summit, with the individuals returning to their normal daily activities and thus reducing the sources of pollution, the effects of the interventions become significant (p-value=0.004). Second, it might have taken time for the interventions to have an impact on air quality. Similar to the case of Beijing, improvement achieved in the neighboring cities completely reverted five days after the summit.

In both the Olympic and the APEC regressions, variables related to time trend, including t and t-squared, showed significant effects on API (p-values ranged from 0.01 to 0.05). In particular, the coefficient of t is negative and significant (p-values=3.144e-32, 0.0001, and 0.0006, respectively, for Olympic and 0.008, 0.025 and 0.023, respectively, for APEC), while the coefficient of t-squared is positive and slightly significant (p-values=0.012 and 0.053 for Olympic and 0.003 and 0.004 for APEC). This suggests that the overall air quality was improving over the whole sample period at a slightly increasing rate. Such results reflect the effectiveness of general air pollution policies and interventions. Among the measures of weather conditions and socioeconomics, wind velocity shows negative and significant effects (p-values<0.01) in both regressions. As expected, API tends to be higher during heating seasons for cities with regular heating supply during the winter.

The contrast of better air quality during the events and worsened air quality after the events is worth a further look to better understand the aspects of air quality improvement strategies and the effectiveness of the air pollution interventions that were taken. While various measures were adopted before and during the two events, certain measures were temporary. Such measures were not sustainable and thus were dropped after the events due to economic and political reasons. During the Olympic Games, major polluting companies were required to either temporarily shut down or reduce production, including Beijing East Chemical Plant, Capital Steel Company, Beijing Yanshan Petrochemical Company, 27 cement production companies, and 18 construction material and metal production companies (Beijing Environmental Statement, 2009). The decisions to shut down or reduce production were not based on incentive programs. Instead, they were based on the magnitude of production and pollution regardless of the potential economic costs. It is not economically sustainable as companies will eventually catch up with production after the events, which may have contributed to the reverted air pollution after the events. One of the specific measures taken during the APEC was the six-day holiday for government organization and institutions. While this greatly reduced the traffic in Beijing and thus improved air quality, there are several associated issues. First, the holiday added traffic and air pollution to the surrounding areas. Second, the holiday was only implemented as a temporary measure. A more sustainable alternative would be to consider relocating certain organizations and institutions to surrounding areas, which would achieve traffic reduction in the long run.

On the other hand, other measures taken can have a long-term impact on air quality. For example, about 140 polluting companies were relocated and several major polluting plants were permanently closed for the Olympic Games, including Beijing Dyeing Plant in June 2003, Beijing Coking Plant in July 2006, and the Second Beijing Chemical Plant by the end of 2007 (Chen et al., 2013). While relocation of plants may affect air quality in the relocated areas, it can certainly reduce air quality in Beijing, where air quality is usually below average in China and as the capital city, receives more attention and scrutiny. Another permanent measure taken is that Beijing adopted the new National IV emission standard in March 2008, way ahead of the scheduled date. This resulted in the retirement of heavy-emission cars (the so-called yellow-label vehicles) and renovation of more than 2000 buses and taxies. These measures contributed to the long-term air quality improvement as shown by the time trend variables in the model. One of the few measures that was extended to date is the odd-even rule for vehicles. That is, vehicles with odd/even registration numbers are allowed on the road on odd/even days. With heavy polluting plants relocated, vehicle emissions became one of the major contributors to air pollution in Beijing. Thus, restricting the number of vehicles on the road could greatly reduce air pollution and contribute to the long-term trend of improved air quality in Beijing.

Public awareness of air pollution after the two events has also contributed to the increased actions taken by the Chinese government to improve air quality. Citizens in China were increasingly concerned about air pollution and the associated health risks, especially after the severe air pollution in northern China in 2012 and 2013. According to a case study by Wang et al. (2016), 60% of the respondents surveyed expressed their willingness to support actions taken to fight heavy smog, one of the major air pollution sources in China. Increased public concern can put pressure on the government to implement stricter regulations on air pollution, which could contribute to long-run air quality improvement.

The duration of the intervention for the two events differs substantially and thus the effects differ. The effects lasted for more than a year in Beijing for the case of the Olympic Games. The effects were only significant (p-value<0.01) within five days after the APEC summit. Such contrast is not surprising though. Interventions were taken for more than six years on a much larger scale before the Olympic Games. In comparison, interventions for APEC were more temporary and taken most extensively during the event. Unfortunately, these efforts come at a high cost. Short-term fixes for the political blue sky made the air quality worse. Comparison of the two events also reveals that intensive interventions taken over a short period may bring more harm than good. As shown by Shi et al. (2020) and Fu et al. (2021), the political blue sky resulted in retaliatory pollution after major national and international events. These studies found that the increase in production due to the economic losses during the restriction periods caused the ramp up in pollutants.

Conclusions

This paper examined how air cleaning interventions affected air quality in Beijing and neighboring cities during two major international events held in Beijing, the 2008 Olympic Game and the APEC summit. Using panel data on 35 major cities in China, fixed-effect panel data models were utilized to empirically evaluate air quality improvement in Beijing, its neighboring cities and co-host cities before, during and after the events. Results showed significant (p-values<0.01) improvement in air quality in Beijing during the two events. The impacts lasted more than one year for the case of the Olympic Games and only about five days after the summit. In both cases, the improvement achieved during the events reverted. More importantly, within one week after the summit, the air quality severely deteriorated for five days, resulting in the worst air quality in the entire sample period. This can impose unnecessary harm to human health.

Examining the effectiveness of interventions has important policy implications and can help policymakers to better regulate daily industrial and public transportation in the fight for better air quality. Given that the interventions can be temporary or permanent, it is critical to examine whether they are sustainable. The interventions for the Olympic Games were implemented for almost seven years from 2002-2008 on a large scale, and the resulting impact of the Olympics lasted longer. On the other hand, interventions for the summit were more temporary and intensive, implemented mostly during the event on a smaller scale. The effects did not last long, but resulted in strong effects on API in the short term. These comparisons have several implications. First, the length of the impacts of the interventions relies on the duration of the policy implementations. As one would expect, the longer and stronger the policy interventions are, the greater the improvement in air quality. Second, air quality improvement can still be achieved even with small-scale interventions if such measures are taken on a continuous basis. Third, although temporary policies can improve air quality during the interventions, they may also create negative effects as the improvement reverts. As shown in the APEC case, the air quality further deteriorated after the interventions, and the high level of API imposed serious damage to human health. Thus, to really benefit from such interventions, continuous implementation of the policy is critical. Considering possible interventions may be adopted during such meetings by future hosts, caution should be taken if temporary policy interventions are to be considered.