Assessing transboundary air pollution and joint prevention control policies: evidence from China

This study addresses the pressing issue of transboundary air pollution, an environmental concern characterised by the dispersion of pollutants across administrative boundaries. Prior research in this area has lacked an in-depth examination of the efficacy of cooperative environmental policies in managing this challenge. To address this gap, our study first evaluates transboundary sulphur dioxide pollution across 31 provinces in China from 2005 to 2020 utilising the Hybrid Single-Particle Lagrangian Integrated Trajectory model. Following this, we apply a staggered Difference-in-Difference model to gauge the impact of the joint air pollution prevention and control policies adopted by China for high-priority provinces. Our analysis revealed an imbalance in emissions and transboundary pollution levels across provinces. Provinces such as Shandong, Shanxi, and Tianjin face the most severe transboundary SO2 pollution, whereas the highest SO2 emissions were noted in Shandong, Inner Mongolia, and Hebei. The implementation of the joint pollution prevention and control policy resulted in a significant reduction in SO2 emissions and transboundary SO2 pollution by factors of 10.60 and 9.70, respectively, when compared to other provinces. These findings provide valuable insights for shaping environmental cooperation policies and identifying priority provinces for mitigating air pollution.


Introduction
Air pollution, particularly transboundary air pollution, is a global environmental issue influenced by a range of factors, such as meteorology, topography, and emission levels (Hekmatpour and Leslie 2022).Local efforts alone often fail to address this systemic problem, stressing the need for cross-regional cooperation and coordinated policy measures (Boucekkine et al 2022).However, transboundary pollution presents a political challenge as it requires governments to take responsibility for environmental issues beyond their jurisdictions (Spiteri and Brockdorff 2021).
One effective strategy for addressing air pollution is to establish regional cooperation mechanisms such as environmental cooperation agreements (Yao et al 2020).In China, the State Council implemented a Joint Prevention and Control of Air Pollution (JPCAP) Strategy in 2010.While some studies have evaluated the impact of air pollution control policies, there is a gap in the literature that specifically examines the effectiveness of such cooperation policies in mitigating transboundary air pollution.This mainly stems from the difficulty in obtaining data and the challenge in finding a scientifically sound method to measure the degree of transboundary pollution (Quah et al 2021).
Moreover, extant studies have primarily focused on specific regions or administrative units, leaving a relative dearth of research addressing transboundary pollution regulation within national boundaries (Du andSun 2022, Zeng andYi 2022).For instance, Zhang et al (2020) found that market integration contributes to emission reduction based on their study of panel data from the Yangtze River Delta region in China.Huang et al (2020) highlighted the amplifying effect of long-range transport and aerosol boundary layer feedback on transboundary air pollution, emphasizing the importance of coordinated cross-regional emission reduction.Yan et al (2021) discovered significant inter-city diversity within urban clusters, which is attributed to unique local emission sources.They emphasized that pollution at the city level can only be reduced through local mitigation actions, while addressing most of the pollution requires cooperation at the national or global level.Sulaymon et al (2021) investigated air quality issues in three major cities in Anhui Province, China, and identified neighboring cities as the main contributors to increased air pollutant concentrations.Chen et al (2021) indicated that transboundary air mass transport still significantly contributes to pollutant levels, suggesting the need for further regional control strategies.Wu et al (2022) conducted modeling research showing that transboundary transport of air pollutants from northern China and the Yangtze River Delta has a significant impact on air quality in the southern coastal areas of China.
This study addresses the prevalence of transboundary air pollution in China in a comprehensive manner.First, it deviates from the traditional focus on direct emissions and adopts a broader perspective by integrating atmospheric dispersion modelling at the provincial level, revealing a more evenly spread distribution of transboundary pollution.It employs combined methodologies (Hybrid Single-Particle Lagrangian Integrated Trajectory model; HYSPLIT and the Staggered Difference-in-Difference model; Staggered DID) to analyse the transboundary sulphur dioxide (SO 2 ) distribution and evaluate the effectiveness of the JPCAP strategy across 31 provinces from 2005 to 2020.Importantly, it extends the scope of examination beyond priority provinces implementing JPCAP to include non-implementing provinces, thus enriching our understanding of policy effectiveness in different contexts.Furthermore, this study underscores the necessity of establishing environmental cooperation mechanisms among high-emission and polluted regions, offering a fresh perspective on the persistent challenge of transboundary pollution.Overall, this study contributes an innovative approach to understanding and addressing transboundary air pollution, enhancing our knowledge of environmental cooperation policies and providing actionable policy suggestions for future research and combating air pollution.

Air pollution transport model
This paper uses HYSPLIT (Draxier and Hess 1998) to determine the geographic distribution of transboundary air pollution.HYSPLIT was developed by the Air Resources Laboratory of the National Oceanic and Atmospheric Administration (Stein et al 2015, Rolph et al 2017).Over the past 30 years, HYSPLIT has been applied for atmospheric trajectory and dispersion calculations (Stein et al 2015).The dispersion of the pollutants was simulated by calculating the advection velocity at the source and superimposing a stochastic wind component.Vertical and horizontal turbulence was calculated by estimating the local stability from the wind and temperature fields.Therefore, forward-and backward-trajectory calculations were performed.
To capture the trajectory of emissions, forward trajectory simulations were initiated for each Chinese province, starting from their latitude and longitude.Trajectory simulations were conducted at 0:00, 06:00, 12:00, and 18:00 UTC every day in January, April, July, and October from 2005 to 2020.These 4 months represented the effects of different seasons on the trajectories.This can help reduce the running time of models of long periods (Dodla et al 2017).This study assumed that the gas particles have a lifetime of four days (96 h) and are dispersed at an altitude of 100 m (Du et al 2020b).
In this study, we first calculated the SO 2 emissions intensity for each province.Additionally, a raster with a resolution of 1°was created based on the geographical extent of the 31 provinces.The trajectory values weighted by the emission intensity passing through each unit of the raster were calculated.These values represent the emission intensities at the sources and dispersion simulations of the pollutants transported to downwind units.The essence of this process is the redistribution of pollutants after atmospheric motion, which is defined as transboundary air pollution in this study.The transboundary air pollution of each province is the weighted average of the grids in the square in which each province is located.Finally, we constructed provincial panel data on transboundary air pollution to evaluate changes over time.This study used the trajectory mode of HYSPLIT and the Jython-based Meteoinfo suite to describe the trajectory distributions (Wang 2014).A relatively simplified trajectory-based approach was used to estimate the spatial distribution of transboundary air pollution.Although this approach simplifies the complex transport of pollutants, it can help capture the essential elements of the source-receptor relationships involved and contribute to the further analysis of environmental cooperation policies.Previous studies have verified that different HYSPLIT model settings do not affect the results (Du et al 2020b, Kopas et al 2020).Appendix A includes more details on the use of HYSPLIT and additional analyses that show that the conclusions are robust regardless of the settings.For instance, utilising more detailed city or county data can provide a more nuanced understanding of transboundary pollution.However, such a refinement also implies a substantial increase in computational demand and extensive time for data processing.We have included a discussion in appendix A stating that shifting to city units did not compromise the robustness of our findings.

Policy evaluation model
This study examined whether transboundary SO 2 was significantly reduced in the focal provinces involved in JPCAP.DID considers new policy implementation a natural and quasi-experiment that is exogenous to environmental economics (Greenstone and Gayer 2009).On the one hand, JPCAP might lead to differences in priority provinces before and after policy implementation.On the other hand, JPCAP might differ between focal and non-focal provinces at the same point in time.A regression estimation based on these dual differences can control for the effects of other co-occurring policies and ex ante differences between focal and non-focal regions (Abadie et al 2010).Therefore, the net impact of JPCAP on transboundary SO 2 in each province can be identified.Considering that JPCAP is initiated in three batches and that there are exit scenarios, this study uses a Staggered DID for assessment (Goodman-Bacon 2021).
Based on the discussion above, the effect of JPCAP implementation of JPCAP is assessed at the provincial level.The model is set as: where y it is the explanatory variable, which is the transboundary SO 2 in province i in year t.coop is the core explanatory variable, which is whether province i implemented JPCAP in year t.It takes the value of 1 if province i implements the policy in year t and 0 if it does not.m i and d t denote the province fixed effect and year fixed effect, respectively.e it is the stochastic disturbance term affecting transboundary SO 2 .b is the estimator that measures the effect of JPCAP on the amount of transboundary SO 2 .Control it is a set of control variables that affect the amount of SO 2 pollution in province i in year t.
Economic characteristics may affect emissions, which in turn affect transboundary air pollution (Du et al 2020a).Therefore, this study considered five factors as control variables: (1) economic development, the logarithm of GDP per capita, represented by logGDPP; (2) population growth, with population expressed as POP; (3) industrialisation, expressed using the industrialisation index InI, which is the share of the secondary industry in GDP.In terms of industrial structure, decreasing the secondary industry share can significantly mitigate SO 2 (Yang et al 2017); (4) urbanisation, expressed as UI, which is the share of the urban population in the total population.The urbanization index may affect SO 2 positively (Ren and Matsumoto 2020); and (5) fiscal decentralisation, expressed using the fiscal decentralization index (FDI), which is the share of provincial fiscal revenue in central fiscal revenue.The relationship between the central and local governments is crucial for political decisions to tackle pollution issues in China (Zheng et al 2014).

Data
HYSPLIT data were obtained from the Global Data Assimilation System (GDAS) at a resolution of 1°.The GDAS data included reanalysis information once a day for every six hours and forecast information once a day for every three hours.The GDAS data were at 23 vertical levels (from 1000 to 20 hPa ) with 35 parameters, including temperature, humidity, wind direction, wind speed, and other meteorological data.Owing to data availability, this study selected the period from 2005 to 2020 as the study interval and 31 Chinese provinces as the study area.The SO 2 emissions and control variables in the DID of each province were collected from China Statistical Yearbook and Statistical Yearbook.To ensure comparability, using a deflator, the economic data were adjusted to constant prices using 2005 as the base year.A linear interpolation method was used to fill in missing data.The descriptive statistics of the variables are provided in appendix B. Replication data and codes can be found in the supplementary materials.

Policy context and study area
In May 2010, the Chinese State Council released Guidance on Promoting JPCAP and Improving Regional Air Quality, which first proposed a cooperative mechanism.The priority regions for JPCAP were the Beijing-Tianjin-Hebei region, the Yangtze River Delta, and the Pearl River Delta.Moreover, central Liaoning, Shandong Peninsula, Wuhan, and its surroundings, Changzhutan and Chengdu-Chongqing, should actively promote JPCAP.
In September 2013, the Chinese State Council publicly released Air Pollution Control Action Plan and proposed JPCAP measures for the next five years.This document highlighted the Beijing-Tianjin-Hebei region, Yangtze River Delta, and Pearl River Delta as priority regions that should combine forces to strengthen the supervision of emission sources, investigate and punish environmental violations, and improve air quality.The State Council and provincial governments in priority regions established a collaborative mechanism for JPCAP to coordinate the resolution of outstanding regional environmental problems.
In July 2018, the Chinese State Council released the Three-Year Action Plan to Win the Blue Sky Defense War, which proposed JPCAP measures for the next three years.Based on the emissions of each province, Beijing-Tianjin-Hebei and surrounding areas, the Yangtze River Delta, and the Fenwei Plain were classified as crucial areas for continuous JPCAP actions.Priority regions are encouraged to use economic, legal, technical, and administrative measures.In addition, they should strengthen regional cooperation.By vigorously adjusting and optimising the industrial, energy, transport, and land use structures, these key regions should achieve multiple wins for society, economy, and environment.
Because of the different economic and social development stages and environmental conditions in each province of China, the above three policies determined the critical areas with some differences.Nevertheless, the main goal is to strengthen regional cooperation to combat air pollution.The central government must coordinate with the provinces to reach environmental cooperation agreements.If key regions achieve their expected goals, this may indicate the effectiveness of JPCAP.Therefore, inter-provincial cooperation seems to contribute to pollution reduction.

Transboundary air pollution
Figure 1 shows the raster of trajectory-weighted SO 2 in the representative years.Using the HYSPLIT simulations described in section 2.1, 15252 trajectories were generated from 2005 to 2020, respectively.Based on the locations of the 31 provinces, the raster coverage ranged from 73°50′E to 135°09′E (longitude) and 6°32′N to 53°5 6′N (latitude).The raster contained 2867 units with a resolution of 1°.Each unit shows the weighting calculation of transboundary SO 2 by assigning the emission intensity to each trajectory.SO 2 diffusion decreased significantly from 2005 to 2020.This may be related to emission reduction and other efforts to improve air quality.In 2005, central and northern China presented large areas with SO 2 pollution intensities above 100, with intensity values between 100 and 220.However, in 2020, only Inner Mongolia had higher pollution intensity, with a maximum intensity of 31.18, which was substantially lower than that in 2005.Beijing, Tianjin, and Hebei (the Beijing-Tianjin-Hebei region) are typical JPCAP implementers.These were among the regions with the highest SO 2 pollution intensities in 2005, 2010, and 2015.However, in 2020, the Beijing-Tianjin-Hebei region showed significant improvements, which might be relevant to the effectiveness of JPCAP.In contrast, Tibet had the lowest SO 2 pollution intensity.This seems to be related to both lower emissions and topographical factors in Tibet.The high terrain and mountains of Tibet may have prevented this diffusion.Air pollution spread beyond national boundaries.However, the diffusion among countries is beyond the scope of this study.
China's atmospheric dynamics are strongly influenced by climatic diversity, which leads to differing atmospheric flow characteristics across provinces.For instance, maritime winds often affect coastal provinces in the east and south, establishing a distinctive land-sea circulation pattern that influences atmospheric diffusion and pollutant transmission.In contrast, the atmospheric flow in the northern provinces is predominantly controlled by the Siberian high-pressure system, often resulting in regional pollutant accumulation.More details about interprovincial transmission can be found in the appendix A.
In general, the characteristics of atmospheric flow between provinces in China are complex and are shaped by geographical location, terrain, climatic conditions, and human activities.For instance, the degree of economic development and industrialisation in each province in China affects the patterns of pollutant propagation.Understanding these characteristics is essential to develop effective air quality improvement strategies.
Figure 2 shows the transboundary SO 2 pollution in China.The total transboundary SO 2 pollution shows a similar downward trend to the SO 2 emissions in figure B.1 of appendix B. However, a slight rebound occurs between 2006 and 2011.This may be related to rapid economic development with increasing energy consumption.From 2016 to 2020, transboundary SO 2 pollution showed a significant decreasing trend.
Compared to the SO 2 emissions in figure B.1, the transboundary SO 2 pollution in figure 2 reflects the redistribution of pollution after considering atmospheric movement.This explains the actual local pollution.Although the trends in emissions and pollution were consistent, the share of transboundary SO 2 pollution varied significantly by province.Overall, transboundary pollution was more evenly distributed than emissions.Shandong (4.69%), Shanxi (4.53%), and Tianjin (4.17%) had the highest pollution shares.Inner Mongolia (6.59%) had the second-highest emissions in China (Figure B.1).However, Inner Mongolia bore a medium-high transboundary SO 2 pollution share (3.99%), as shown in figure 2. This imbalance between emissions and pollution indicates that SO 2 from Inner Mongolia spread to some extent to other places.Tibet (0.56%), Qinghai (1.15%), and Hainan (2.16%) were the bottom three provinces with the lowest transboundary SO 2 pollution percentages.
Based on figure 2 and B.1, emissions and transboundary pollution are categorised in figure 3.For example, Inner Mongolia was classified as a high-high province because its emissions and transboundary pollution were both above the median level.The first quadrant in figure 3 includes high-high provinces.These provinces are recommended to be considered priority regions for JPCAP because they must commit to reducing their emissions and improve their air quality.These provinces are inclined to set up cooperative mechanisms as they are anticipated to yield the most substantial environmental results.The second quadrant includes low-high provinces with low emissions and high pollution levels.These provinces may often be overlooked due to their relatively low emissions, despite bearing a substantial and often unseen burden of transboundary air pollution.However, they are also encouraged to cooperate with other provinces to solve air pollution problems.The third quadrant includes low-low provinces with low emissions and pollution.These provinces can be considered as regions where air pollution is not a severe problem.The fourth quadrant includes high-low provinces with high emissions and low pollution levels.These provinces are the source regions that export transboundary pollution.It is suggested that compensation mechanisms be implemented in these provinces and that they bear responsibility for the pollution in provinces in the second quadrant.To verify the validity of HYSPLIT estimations, this study compared transboundary SO 2 pollution with the annual average provincial SO 2 concentrations reported in the Statistical Yearbook.It is important to note that transboundary pollution is a redistribution of emissions and its value may be related to the observed surface SO 2 concentration.However, as mentioned earlier, the transnational dispersion and natural deposition of air pollution must be considered to determine the actual concentrations.The Pearson correlation coefficient between transboundary pollution and surface SO 2 concentration was 0.64, indicating a positive to moderate correlation.This indicates that the calculations in this study correctly reflected the transboundary SO 2 pollution in each province.

Basic regression model
The evaluation of JPCAP was based on equation (1).Researchers should first use a two-way fixed-effects model as a comparative benchmark (Callaway and Sant'anna 2021).Therefore, policy evaluation begins with a traditional DID regression analysis that controls for province-and year-fixed effects.The regression results are presented in table 1. Column (1) shows the direct effect of JPCAP on transboundary SO 2 pollution.The estimated coefficient of the dummy variable coop is −11.362 and passes the significance test at the 1% level.This indicates that JPCAP significantly reduced transboundary SO 2 pollution in the priority provinces.In Column (2), the coefficient of coop remains negative after controlling for economic and demographic characteristics.The coefficient is −9.704 and passes the significance test at the 1% level.This indicates that with full consideration of other factors, transboundary SO 2 pollution reduction in JPCAP provinces was 9.70 times larger than that in other provinces.
Similarly, JPCAP significantly reduced SO 2 emissions in priority provinces by 10.60 times that of other provinces.However, the relative number of coefficients for coop in Column (4) is slightly larger than that in Table 1.Regression results of the effect of joint prevention and control of air pollution.Values in parentheses are robust standard errors with the province as the clustering variable for the regression coefficients.* , ** , and *** indicate significant at the 10%, 5%, and 1% levels, respectively.The results were calculated by the authors using Stata17. (1) ( 2) Column (2).This might be related to the fact that JPCAP provinces need to make tremendous efforts to reduce emissions through environmental cooperation mechanisms.As a result of atmospheric dispersion, JPCAP provinces were more effective in reducing emissions than transboundary pollution.These provinces have made significant efforts to reduce emissions, yet the tangible impact of their efforts does not equate to an equivalent reduction in the transboundary pollution they endure.
Our empirical findings offer several important policy recommendations.First, they provide robust evidence supporting the adoption and continuation of cooperative environmental policies, such as JPCAP, as these policies have been demonstrated to be effective in reducing trans-boundary air pollution.Moreover, considering the substantial effort required by priority provinces to reduce emissions, it would be beneficial to provide these provinces with additional resources and support to continue their mitigation efforts.Finally, the uneven distributions of contributions and benefits suggest the need for more equitable allocation of responsibilities and rewards within environmental cooperation agreements.This could foster a more sustainable and inclusive approach for combating air pollution.
The regression results examined the impact of JPCAP on transboundary SO 2 pollution.Because the control variables were not the focus of this study, the regression results of control variables are presented in appendix B.

Dynamic estimation of staggered DID
The traditional DID is appropriate for estimating the average effect within a single treatment period (Baker et al 2022).As JPCAP was implemented over three periods, with emphasis on different provinces, the above basic regression results might be biased.A staggered DID is recommended for multi-period estimations (Athey and Imbens 2022).The earliest provinces that participated in JPCAP might have consolidated pollution reduction because of the early establishment of the cooperation mechanism and higher degree of support.Therefore, the dynamic effect was examined using the estimators proposed by de Chaisemartin and D'haultfoeuille (2020).Considering data availability, the first seven years before and last seven years after JPCAP implementation were evaluated.
The estimation results are shown in figure 4. Prior to the implementation of JPCAP, the estimated coefficients were not significant for each year, suggesting that there may not have been significant differences among provinces at this time.We used absolute time in our analysis, with the timing of policy implementation varying across provinces, which might further contribute to the observed pre-policy trends.
Despite these limitations, after the implementation of JPCAP, transboundary SO 2 pollution was significantly reduced.The estimated coefficients of JPCAP were not significant in the first two years.This may be due to the initial stages of policy implementation, which often involve various preparatory measures, such as holding meetings to discuss research measures and strategies and monitoring emission reductions achieved in the first year to adjust reduction measures accordingly.Given these initial activities, it is understandable that the impact of the policy may not be immediately apparent in the first two years.Starting in the third year, the significant effects of the policy became more evident, reflecting the time needed for policy measures to take effect and demonstrate substantial results.After implementing JPCAP for three to five years, the estimated coefficients ranged from −6.873 to −11.260 and were significant at the 5% level.During the final two years, the estimated coefficients ranged from −18.163 to −20.552, and were significant at the 5% level.Although the estimated effects of JPCAP differed from the results of the basic regression, both models illustrate its effectiveness.

Event study
The traditional DID estimation should also be accompanied by event study estimates that address the biases arising from the bad comparison problem (Baker et al 2022).Therefore, this study used a non-participatory event study to re-estimate the effect by retaining only JPCAP provinces.To better handle provinces with exit scenarios over multiple periods, the time variable was adjusted for the year of JPCAP implementation.Considering the small sample size, the first five years before and last five years after JPCAP implementation were evaluated.The results are shown in figure 5.The estimated coefficients were not significant at the beginning of JPCAP.This may be related to the lag effect of the policy, as discussed previously.The estimated coefficients during the final three years were significantly negative, ranging from −7.216 to −11.329.This indicates the effectiveness of JPCAP.
To further perform robustness tests on JPCAP, this study took the SO 2 concentrations monitored in each province as the explanatory variable in equation (1).The coefficient of coop was -3.09 (significant at the 1% level).The fact that the monitored SO 2 concentrations were partially relevant to transboundary SO 2 pollution illustrates the validity of JPCAP.Further discussion on the robustness tests can be found in appendix B, including the regression results of transboundary SO 2 pollution obtained from different HYSPLIT settings and the regression results after employing the Propensity Score Matching method.All tests affirm the effectiveness of JPCAP.

Conclusion
This study provided a comprehensive and systematic understanding of transboundary air pollution in China by combining SO 2 emissions with atmospheric dispersion modelling at the provincial level.The results showed that transboundary pollution was more evenly distributed than emissions owing to dispersion.Regions with the highest emissions were not necessarily the most severely polluted.This discrepancy largely resulted from the atmospheric dispersion patterns and regional meteorological conditions.For instance, Shandong, Inner Mongolia, and Hebei had high emissions due to intensive industrial activities.However, Shandong, Shanxi, and Tianjin experienced severe transboundary pollution due to factors such as wind direction and geographical location.The effectiveness of the joint pollution prevention and control policies implemented in China was also analysed.The results showed that JPCAP provinces significantly reduced SO 2 emissions by 10.60 times and transboundary SO 2 pollution by 9.70 times.
However, this study had some limitations.It simplifies transboundary air pollution modelling, an area that warrants further investigation.While offering a provincial-level framework for addressing this issue, future studies could benefit from a more granular approach, such as a city or county-level analysis, to yield more nuanced insights.This study emphasizes the necessity of region-specific environmental cooperation mechanisms.For highemission provinces, implementing stringent emissions controls such as a cap-and-trade system could effectively allocate emissions reductions.The central government should incentivize cleaner technologies via subsidies and tax benefits to reduce reliance on pollution-intensive industries.For regions severely affected by transboundary pollution, policies could bolster resilience through improved environmental cooperation regulations.It's recommended that the central government take a leading role in fostering regional environmental cooperation agreements, particularly aligning these with broader economic and developmental goals.Such agreements could include measures for joint monitoring and management of air pollution, shared responsibility for emissions reduction, and mutual support in the event of severe pollution incidents.In the long run, building such cooperative mechanisms would not only help reduce transboundary pollution but also promote sustainable economic development across provinces.
Overall, this study contributes to the understanding of transboundary air pollution and provides insights into environmental cooperation policies to reduce air pollution.

Figure 2 .
Figure 2. Transboundary SO 2 pollution in China by province from 2005 to 2020 (million tons).

Figure 3 .
Figure 3. Categorization of SO 2 emissions and transboundary pollution in China by province.The horizontal coordinate is the share of emissions.The vertical coordinate is the share of transboundary pollution.And the coordinate line is the median of the share.

Figure 4 .
Figure 4. Dynamic estimation of the effect of joint prevention and control of air pollution.The solid points represent the estimated coefficients of the policy effects.The short vertical lines are the 95% confidence intervals corresponding to the robust standard errors by the clustered province.

Figure 5 .
Figure 5. Event study estimation of the effect of joint prevention and control of air pollution.The dotted line is the previous year before policy treatment.The solid points represent the estimated coefficients of the policy effects.The short vertical lines are the 95% confidence intervals corresponding to the robust standard errors by the clustered province.