China’s economic development has brought about a number of environmental issues, and the Chinese government has been trying to cope with environmental pollution through multiple policy tools. In 1979, the central government of China implemented the Environmental Protection Law, and a pilot system for emission charging was also proposed in some places to control the emission of industrial pollutants. In 1982, the central government further implemented the Tentative Provisions on Pollution Charge, and clarified the standards and methods for the collection of emission charges. However, the policy did not achieve the expected targets of emission reduction due to the low standard of emission charges. Along with the continuous development of industry, increased pollution issues emerged. China’s pollution problems have therefore not been effectively addressed (Jiang et al., 2014). After that, in 2003, the central government published the Administrative Regulations on Levy and Use of Emission Discharge Fee, as a further step to reduce pollution levels across the country.

Emission charges have always been a means for the Chinese government in emission reduction (Wang and Jin, 2007). However, the effect of emission charges on environmental improvement is weak because of lenient enforcement (Jiang et al., 2014; Maung et al., 2016). In order to solve the increasingly serious environmental problems, the Chinese government implemented the Environmental Protection Tax Law in 2018, which involves taxes on air pollutants, water pollutants, solid wastes and industrial noise. In contrast with the emission charging system, which is only an administrative rule, the EPTL has a stronger effect on pollution behaviors (Maung et al., 2016).

The EPTL sets minimum tax rates nationally, and provinces are authorized to levy environmental protection taxes within ten times of the prescribed minimum value. The EPTL is developed based on the previous emission charging system, and a number of provinces still charge the same amount of cash to polluters as before. In other words, the amount of money paid by polluters in these provinces for discharging pollutants remains unchanged after the reform. In contrast, 14 provinces have raised the charges for emission after the reform, including Beijing, Tianjin, Hebei, Jiangsu, Shandong, Henan, Sichuan, Chongqing, Hunan, Hainan, Guizhou, Guangdong, Guangxi and Shanxi.

This paper is related to two strands of literature. The first is the effect of environmental taxes. Environmental taxes can increase tax burden for polluting firms, which in turn influences both the environment and the economy (Pearce, 1991; Bonnet et al., 2018; Li et al., 2022). A growing body of literature indicates that environmental taxes are among the most effective ways to reduce pollution and improve environmental quality (Baumol et al., 1988; Bovenberg and Mooij, 1997; Patuelli et al., 2005; Ekins et al., 2011; Ekins et al., 2012; Bonnet et al., 2018; Li et al., 2022). For instance, Agnolucci (2009) investigates environmental tax reform in Germany and Britain, and finds that environmental taxes can reduce both energy consumption and carbon dioxide emissions in a significant way. Also, Bruvoll and Larsen (2004), Guo et al. (2014), Wang et al. (2018), Zhai et al. (2021) find that carbon taxes can reduce emissions of carbon dioxide and air pollutants at the same time.

Environmental taxes improve environmental quality primarily by forcing producers to reduce emissions, rather than lowering consumers’ consumption (Pang, 2018; van der Ploeg et al., 2022). This is because for producers, the supply elasticity is relatively higher, and producers can adjust their output to achieve emission reduction targets. While for consumers, the demand elasticity is relatively lower, which makes it difficult for consumers to change their demand significantly in a short time (van der Ploeg et al., 2022). For producers, environmental taxes increase the marginal cost of energy consumption, and thus lead to lower output levels as well as reduced emissions (Gray and Shadbegian, 2003; Pal and Saha, 2015). In addition, if environmental taxes paid by firms are higher than employing clean technologies, firms are more willing to pay for cleaner technologies and emission-related technological innovations (Cremer and Gahvari, 2004; Brécard, 2011; Ekins, et al., 2011; Lanoie et al., 2011; Ebenstein et al., 2015; Aghion et al., 2016; Leslie, 2018; Chen and Ma, 2021; Cheng et al., 2022).

Environmental taxes are widely regarded as an effective means for environmental governance (Patuelli et al., 2005; Ekins et al., 2011; Ekins et al., 2012; Leslie, 2018; Li et al., 2022), and that factors including industrial structure, the use of environmental tax revenues and trade liberalization can reduce the effectiveness of environmental taxes for pollution abatement (Bosquet, 2000; Yates, 2012; Duan et al., 2021; He et al., 2021; Zhang et al., 2022). For example, Fredriksson (2001) finds that environmental lobbyists can influence the effectiveness of environmental taxes. This is because, when politicians are able to influence the provision of related subsidies to firms, lobbyists may finance campaigns in exchange for the subsidies, which are greater than their lobbying costs. As a result, emission costs for lobbyist-associated polluters are reduced, and their emissions will further increase (Fredriksson, 2001). In addition, environmental benefits of environmental taxes may decline due to inflation, entry of new polluters, and government’s implementation of other alternative policies (Clinch et al., 2006; Cao et al., 2021).

In terms of economic effects, environmental taxes have impacts on economic growth, tax distortion, income distribution, and fiscal sustainability (Bosquet, 2000; Ekins et al., 2011; Karydas and Zhang, 2019; Costantini and Sforna, 2020; Spinesi, 2022; Zhang et al., 2022). Controversy remains over the economic impacts of environmental taxes. For instance, Ekins et al. (2011), Aubert and Chiroleu-Assouline (2019) argue that environmental taxes can reduce tax distortions through redistribution, and contribute to higher social welfare. While Bovenberg and De Mooij (1994), and Bonnet et al. (2018) propose that, when tax distortions already exist, environmental taxes will exacerbate distortions. This is because environmental taxes can contribute to increased production costs and lower wages for workers, and lead to higher tax distortions (Bovenberg and Goulder, 2002). Karydas and Zhang (2019) explores the influence of environmental taxes on economic growth and technological innovation and find that the impacts may vary due to related external factors. Oueslati (2014) also argues that the type of taxes vary the associated economic and welfare effects.

Previous studies focused on the environmental and economic effects of environmental taxes. However, the extent of environmental taxes on air pollution and the characteristics of the effects are relatively underexplored. In addition, most of the existing studies investigate the impact of environmental taxes through model derivation, no one has yet analyzed environmental taxes’ contribution to emission reduction using micro data from developing economies.

Another strand of literature is the influence of government related factors on pollution in three aspects. The first is environmental policies (e.g., Sadik-Zada and Sadik-Zada, 2020). For example, in the United States, the Clean Air Act has reduced pollution in hundreds of counties and has improved air quality in the long-term (Greenstone, 2002; Isen et al., 2017). And in China, the Environmental Protection Law has been proved to be effective in air pollution reduction (Li et al., 2016; Geng et al., 2021).

The second is local governments’ fiscal stress. Previous studies find that fiscal stress is closely associated with environmental quality (Wen and Zhang, 2022). Fiscal stress contributes to laxer environmental regulations, which is beneficial for local governments to attract investment, and contribute to environmental degradation (Chen 2017; Bai et al., 2019; Wen and Zhang, 2022). In addition, fiscal stress can distort the structure of government spending, governments may reduce expenditures on environmental conservation even though faced with severe pollution, environmental degradation can thus be accelerated (Hettige et al., 2000; López et al., 2011). For instance, Kong and Zhu (2022) find that the abolition of agricultural taxes caused increases in local fiscal stress, and led to a 4 percent increase in emissions.

The third is tax competition. Tax competition is the phenomenon that local governments compete with each other for tax sources. Cremer and Gahvari (2004) find that tax competition causes an increase in pollution emission. Tax competition can contribute to laxer environmental standards, and result in environmental degradation (Wilson, 1999; Bai et al., 2019). In general, local governments engage in tax competition by reducing tax rates and providing tax incentives, which lead to losses in tax revenues. Local governments are therefore motivated to provide fewer environment-related public goods. Environmental degradation can thus be accelerated. In conclusion, a large body of literature has focused on government-related factors that can affect pollution, little attention has been paid to the influence of the environmental tax reform on pollution in China.

The environmental tax reform increases the cost of emitting, and forces firms to reduce emissions by reducing production and applying cleaner technologies. Before the reform in 2018, China had been charging for emissions based on an administrative rule known as the “emission charging system”. Due to local governments’ various considerations of tax revenues, employment and economic development, the regulation had not been strictly enforced. For instance, a number of polluting firms with large contribution to government revenue are exempted from the charges, the effectiveness of the regulation on air pollution is therefore limited (Wu and Tal, 2018). After the EPTL reform, China taxes emissions according to the law. The emission charging system is only an administrative rule, while the EPTL is a law and is therefore of greater mandatory character (Wang et al., 2018). The EPTL contributes to higher regulatory pressures for firms and increased cost of emissions, and further forces polluting firms to participate in environmental management (Acemoglu et al., 2012).

In addition, on the basis of the emission charging system, 14 provinces have raised the charges according to the EPTL, which contribute to enhanced emission reduction effect of the EPTL in these provinces ¹ . Local governments are authorized by the EPTL to adjust tax rates upward, and 14 provinces have raised tax rates. For other provinces, charges for emission remains unchanged as it was in the emission charging system. Existing studies show that higher charges increase the cost of emissions for firms remarkably (Ekins, et al., 2011; Pal and Saha, 2015; Leslie, 2018; Cheng et al., 2022). Higher tax rates can thus force firms to find ways for emission reduction, which in turn contribute to improved air quality (Gray and Shadbegian, 2003; Ekins, et al., 2011; Pal and Saha, 2015; Cheng et al., 2022). In additon, previous studies show that firms are less likely to invest in regions with higher tax rates (Dean et al., 2009). Therefore, we propose the following hypothesis.

H1

The reform can mitigate air pollution.

In regions with stricter tax enforcement, tax avoidance activities are more likely to be detected by tax authorities. The consequences of tax avoidance activities are also more severe for firms in these regions. Tax avoidance may not only fail to create interests for firms, but can also bring them losses in reputation. In regions with stricter tax enforcement, the benefit-cost ratio of tax avoidance becomes lower, which further discourages firms’ tax avoidance activities. Therefore, we propose the following hypothesis.

H2

Stricter tax enforcement can enhance the effect of the reform on emission reduction.

Fiscal stress contributes to laxer environmental regulation and declines in governments’ investment in the environment, and further leads to air pollution. Local governments are inclined to support industries that contribute more tax revenues facing higher fiscal stress (for instance, manufacturing and construction industry), and are more likely to relax environmental regulation for economic development (Han and Kung, 2015; Bai et al., 2019; Wen and Zhang, 2022). Laxer environmental regulations can make local firms emit more pollutants, and can attract new polluters as well (Dean et al., 2009). In addition, economic performance is particularly important for the promotion of local officials (Li and Zhou, 2005; Chen et al., 2021). The need for economic growth can distort the structure of local government expenditures (Bai et al., 2019). Facing higher fiscal stress, governments are inclined to reduce investment in environmental protection, which can adversely affect pollution abatement (Bai et al., 2019). Accordingly, we propose the following hypothesis.

H3

The reform is less effective in regions with higher levels of fiscal stress.

Initial pollution levels can also vary the effect of the reform. This is because regions with clean air have less potential to improve air quality (Gendron-Carrier et al., 2018). If the air pollution level does not reach a certain level of severity, the reform ought to be less effective. In addition, local governments will trade-off between economic development and the environment (Pang et al., 2019). For less polluted cities, the governments are not willing to sacrifice economic development for environmental improvement, leading to weaker effects of environmental regulations. While for heavily polluted cities, local governments are more inclined to improve air quality to avoid punishment from the central government. Therefore, we propose the following hypothesis.

H4

The reform is more influential in regions with higher initial pollution levels.

As we discussed earlier, building on the emission charging system, 14 provinces have raised the charges according to the EPTL, which can further contribute to enhanced effectiveness of the EPTL in the provinces. Also, the appliance of exogenous policy shocks in this paper can reduce the problem of endogeneity bias (Friedrich, 2022). The environmental tax reform is implemented by the central government. Firms are thus almost impossible to influence the environmental tax reform. This addresses the reverse causality problem in the inferences of the paper. Hence, using environmental tax reform as a quasi-natural experiment and employing the difference-in-difference (DID) method can reduce endogeneity bias and increase the reliability of the results in this paper. To test the effect of the reform on air pollution, we adopt the difference-in-differences (DID) method and employ the following model: A Q I j t = β 0 + β 1 T r e a t i × P o s t t + β 2 W j t + θ j t + ε i j t (1)

In Eq. 1, AQI _jt is air quality index for city j in date t. Treat _i is an indicator for cities in the 14 provinces that have raised the tax rates ² . Post _t is an indicator variable which equals one for years of 2018–2021, and zero otherwise. This is because the EPTL came into effect in January 2018. W _jt is a series of weather controls, including Wind speed, Wind direction, Temperature, Precipitation, Dew point, and Sea-level pressure, as well as the corresponding quadratic terms (Zhang et al., 2018). Weather variables are included because the existing studies show that meteorological factors can influence the aggregation and dissipation of air pollutants (Seaman, 2000; Arain et al., 2007; Greenstone et al., 2022). Following Agarwal et al. (2019) and Li et al. (2020), we introduce weather variables in the model. The definitions are presented in Table A1 θ _it is a set of fixed effects, including city and date fixed effects. We include city and date fixed effects to control for both ctiy- and time-level unobservables. We cluster standard errors at the city level. All continuous variables are winsorized at the .5 and 99.5 percent.

Data on air pollution comes from the Ministry of Ecology and Environment of China (CMEE). We obtain hourly data on AQI, PM _2.5 , and PM ₁₀ from the CMEE, and convert the data into daily mean values at the city level.

Data on meteorological factors comes from the National Oceanic and Atmospheric Administration (NOAA). Original data on weather are at three-hour intervals. Dew point is used to control for relative humidity (Zhang et al., 2018). We combine the data into daily mean levels. Quadratic terms of weather variables are constructed using daily mean values. An inverse-distance weighting method using weather stations within 200 km is applied to compute city level weighted weather data (Zhang et al., 2018). Distances are measured in great-circle distance for higher accuracy (Sager 2019). We convert Greenwich Mean Time (GMT) used in the NOAA data to Beijing time, and then match weather variables with pollution data. Descriptive statistics are shown in Table 1.

Table 2 reports the results from estimating Eq. 1. In column 1, we exclude control variables and fixed effects. The coefficient of Treat × Post is significantly negative at the 1% level. In column 2, fixed effects are introduced. The coefficient of Treat × Post is negative and significant at the 1% level, and declines from −4.523 to −2.634. In column 3, we exclude the quadratic terms of each weather variable. The coefficient of Treat × Post remains significant. Compared with column 3, the coefficient of Treat × Post in column 4 is only slightly disturbed. In Table 2, all the coefficients of Treat × Post are negative and statistically significant. This indicates that, to some extent, the reform is exogenous.

In column 4, the coefficient of Treat × Post indicates that in provinces which have raised tax rates, AQI is 2.361 lower than other provinces after the reform. Air quality in the provinces with higher tax rates is significantly improved. Hypothesis 1 is thus confirmed. Firms in regions with higher tax rates may have higher emission costs. The firms are therefore more inclined to reduce production and apply cleaner technologies.

Stricter enforcement of emission-related policies can make it increasingly harder for firms to avoid paying environmental taxes. To test the impact of tax enforcement on the causality between the reform and air pollution, we estimate Eq. 1 after introducing the interactions between TE and Treat × Post. TE refers to tax enforcement, defined as current tax burden minus the expected tax burden. We use the data from 2016 to calculate TE. The expected tax burden is calculated using the residual of Eq. 2.

In Eq. 2, T _jt is tax revenue for the city j in year t. GDP _jt is gross domestic product of city j in year t. IND1 and IND2 are the proportion of first and secondary industry in GDP, respectively. Eq. 3 is used to measure the extent of tax enforcement. A higher value of TE indicates stricter tax enforcement. T E = T j t G D P j t − P r e T j t G D P j t (3)

The results are presented in Table 3. Introducing TE in the model changes the number of observations available for regression, we therefore drop observations for which TE is missing. The result in column 1 shows that although the sample size is reduced, the effect of the reform on air pollution remains significantly negative. In column 2, the coefficient of Treat × Post × TE is negative and significant. This suggests that stricter tax enforcement contributes to enhanced effect of the reform on emission reduction. The result is consistent with Hypothesis 2.

The findings in Table 3 show that tax enforcement can vary the effectiveness of the EPTL. And as we discussed in Section 2, during the implementation of the emission charging system, lax enforcement of the policy had contributed to weakened efficiency. Similarly, after the reform, if local governments’ do not enforce the law effectively, the implementation of the EPTL may not achieve its primary target, namely, reducing environmental pollution.

Tax competition is the phenomenon that local governments compete with each other for tax sources. Tax competition among local governments can lead to serious air pollution in multiple ways ³ . Tax competition can contribute to laxer environmental regulations and inefficiency of environment-related policies, and result in environmental pollution. In order to attract investment and labor, local governments may lower tax rates to engage in tax competition (Cremer and Gahvari, 2004; Bierbrauer et al., 2013). To maximize profits, firms may migrate from high-tax-rate regions to lower ones, which causes increased pollution in the latter. In China, local governments are not authorized to adjust tax rates without the approval of the central government, but are authorized to offer tax incentives, which further results in lower effective tax rate. This further helps local governments to attract more investment. Therefore, we can predict that the reform is more effective in regions with a lower level of tax competition.

According to Bai et al. (2019), tax competition is defined as income from value added tax divided by GDP. We then creat two subsamples according to the median value of tax competition. Column 1 and 2 of Table 4 displays the results for subsamples in which tax competition is above and below the median value, respectively. As predicted, the reform improves air quality in regions with lower tax competition.

The results show that a higher level of tax competition restricts the effectiveness of the reform. Existing studies also show that tax competition between governments has a negative impact on pollution for both local and surrounding regions (Wilson, 1999; Cremer and Gahvari, 2004; Bai et al., 2019). If local governments are overly involved in tax competition, the reform would be less effective.

Fiscal stress can lead local governments to apply laxer environmental regulations so as to attract investment. Also, fiscal stress is responsible for reduced expenditures on environment-related public goods. Hence, local governments’ fiscal stress may undermine the effectiveness of the reform. To investigate whether fiscal stress matters for the causality between the reform and air pollution, we apply interactions between FS, FS1 and Treat × Post in the model. The definitions of FS and FS1 are presented in Table A1. We use the data from 2016 to calculate FS and FS1.

The results are presented in Table 5. Introducing FS and FS1 in the model also changes the number of observations available for regression, hence, we drop observations for which data on FS or FS1 is missing. Column 1 shows the baseline results, the coefficient of Treat × Post remains negative and significant. Column 2 and 3 report the results after introducing the interactions. The coefficients of Treat × Post × FS and Treat × Post × FS1 are positive and significant. The findings show that the effectiveness of the reform can be reduced by local governments’ fiscal stress.

Initial pollution levels can vary the influence of the reform on air pollution. To examine the role of initial pollution levels, we induce the interaction between Treat × Post and Initial. Initial is the annual average AQI in 2017, the year before the reform. Annual mean values of AQI are calculated using daily data. Cities with daily data less than 100 days are excluded.

Table 6 displays the results. Due to missing values of Initial for a few cities, the sample size is smaller than the baseline regression. Hence, we re-estimate Eq. 1 using the reduced sample. In column 1, the coefficient of Treat × Post is negative and significant, indicating that the result is not influenced by the reduction in sample size. In column 2, the coefficient of Treat × Post × Initial is significantly negative. The result shows that the reform is particularly significant in heavily polluted regions. Regions with higher initial pollution levels have greater potential for environmental improvement.

Results in Table 6 indicate that initial pollution levels are able to vary the causality between the reform and air pollution. Regions with higher initial pollution levels are more sensitive to the reform. Gendron-Carrier et al. (2018) investigate the expansion of subway system on air pollution and find that, in cities with higher pollution levels, subway systems contribute to improvement of air quality. While the effect is weaker in less polluted cities. To some extent, our result is consistent with Gendron-Carrier et al. (2018).