Performance Evaluation of Economic Relocation Effect for Environmental Non-Governmental Organizations: Evidence from China

2.1 Literature Reviews

Informal environmental regulations play a significant role in industrial transfer and upgrading. Informal environmental regulations influence the decision-making process of enterprises, promoting technological innovation and environmental protection behaviors (Féres & Reynaud, 2012; Zhang et al., 2023), thereby driving the green transformation of industries to some extent (Pargal & Wheeler, 1996; Pargal et al., 1997). At the same time, these studies also emphasize the key role of informal environmental regulations in achieving a win–win situation for economic development and environmental protection (Ouyang et al., 2019; Sharghi et al., 2024; Wu, 2023). ENGOs, as a form of informal environmental regulation, have attracted academic attention for their role in environmental governance. Li et al. (2018a,b) used urban panel data from China to quantitatively study the significant improvement in urban environmental quality due to the development of ENGOs in China. Anderson et al. (2019) conducted research in 50 cities in China and found that the supervision of non-governmental organizations (NGOs) can promote the implementation of local government environmental policies. In the process of climate change, ENGOs directly or indirectly participate in the formulation of climate change policies, becoming an important partner for governments in climate change governance (Haris et al., 2020). Partelow et al. (2020) argue that the diversity of ENGOs exceeds traditional perceptions, and their influence in global environmental politics and discourse is continuously increasing, mainly including environmental discourse categories: environmental management, climate politics, environmental justice, and ecological modernization. Li et al. (2021) used data from OECD countries and found that ENGOs have a positive impact on improving environmental quality by increasing environmental protection investments.

2.2 Theoretical Mechanisms

As a non-profit organization, NGOs play an important role in the process of economic and social development. Among them, ENGOs, as a type of NGO, take environmental protection as their responsibility and have become an important component of the environmental protection process in various countries around the world. The rise of ENGOs is putting political pressure on local government officials with respect to environmental issues. Over the past 20 years, scholars have extensively studied the role of ENGOs on environmental laws and policies. Globally, the effects of pollution on human health can result in a strong public demand for environmental and energy policies in order to accelerate governmental action and improve accountability (Grano, 2012). ENGOs have many means, such as media, to influence environmental governance and policies, highlight the need for implementation of laws, and represent the interests of vulnerable populations.

Despite these facts, ENGOs in China are among the least likely to have an influence on policymakers (Lu, 2007). Indeed, ENGOs are perceived as social movements that focus on individual interests and act on behalf of mankind (Wang & Liu, 2009). ENGOs are more likely to influence green policies when these do not oppose stakeholder interests (Betsill & Corell, 2008). Therefore, the lower the political stakes, the higher the ENGO influence. In particular, ENGOs usually have a better chance of playing an important role in the negotiation process during the early stages of environmental policy making rather than the later stages of development actions (Betsill & Corell, 2008). However, it is worth noting that these organizations can address and reflect the needs of local residents and community groups through grassroots communication, by creating trust and reducing distrust in bureaucratic encounters.

Due to the authoritarian constraints to which ENGOs in China are subjected, some of them can push their agendas to the attention of authorities without openly criticize the local government, as opposed to other ENGOs around the world. However, the non-political and non-antagonistic nature of Chinese ENGOs can still have wide-ranging civic and political implications, as they learn the subtlety of domestic politics by cautiously pushing the boundaries of advocacy (Young, 2001).

Given the inherently superior ability of ENGOs, some successes have been achieved via (i) legal suits that generate pressure on the authorities; (ii) the involvement of key actors, such as the media, which contribute to make the issue more well-known, and (iii) the conventional method of utilizing bureaucratic political channels and conflicts between politicians and businesses. The long-term battle is not only to defeat a single corrupt official or relocate polluting factories but also to build a more transparent and open system in which citizens are actively involved.

Moreover, ENGOs can also exert positive influences on the firms’ environmental regulations (Heyes et al., 2018), environmental inspections, and enforcement actions, including issuing new guidance and initiating social movements (Li et al., 2018a,b). In this regard, cognitive theory has emphasized the importance of attitudes, beliefs, ideology, and personal values in driving corporate behavior. Specially, when the firms are situated in a community that advocates for a sustainable environment, they may tend to implement similar measures to strengthen social cohesion and remove the barriers among themselves (Sun et al., 2019).

The ability of ENGOs can bring pressure to bear on the firms for the implementation of environmental standards and indicators (McCarthy & Zen, 2010). If a company refuses to comply with the pollution criteria, ENGOs begin to affect state policy in response to the threat of negative actions, including consumers’ boycotts (Gunningham et al., 2004). In this sense, the involvement of ENGOs in green decision-making and enforcement remains crucial.

The partnership between ENGOs and advocacy campaigns plays important roles in raising public awareness, increasing consumer satisfaction, and developing new environmental regulations (Hopkins et al., 2011). To a certain extent, the accelerated development of ENGOs reflects a need for more responsible environmental management and the increasing demand for corporate leaders toward green value creation (Dauvergne & Lister, 2012). Through their demonstrating commitment, companies have gained legitimacy and are influenced by ENGOs with respect to corporate production and environmental governance. Besides, climate changes and global environmental issues also create conducive conditions for policy decision-making.

There may be some intriguing possibilities for ENGOs to leverage company initiatives, reach into corporate networks, and accelerate economic reforms, at unprecedented speed and scale. The lever for ENGOs to drive corporate environmental improvements is much greater now that companies have tied their products so openly to far-reaching sustainability commitments. The costs resulting from failing to meet the goal of company reputation are considerably high, and even higher if competitors have made good progress toward corporate sustainability (Dauvergne & Lister, 2012).

It is insightful to reveal the importance of ENGOs in shaping the context of private environmental governance mechanisms, such as eco-certification (Lister, 2011). Previous research has shown that ENGOs can start a partnership with industry frames and play a critical role in the face of conflict over corporate legitimacy, thus creating new forms (state-led and market-led) of environmental governance (Dauvergne & Lister, 2012).

The environmental issues (such as climate change and air pollution) may be even worse, as the state capacity or willingness to address the causes is low, leaving the government failure to do much about it. Thus, ENGOs should explore creative ways to leverage the scale, speed, and innovative ideas of corporate governance to increase regulatory efforts. ENGOs also need to focus on the roles of companies in protecting the environment at different stages of the product life cycle, ranging from material sourcing to manufacturing, retailing, and consumer use. Most companies are forced to decrease the intensity of environmental impact per unit of output, mainly through energy-efficient appliances and energy-conserving practices encouraged by ENGOs (Dauvergne & Lister, 2012).

ENGO–corporate partnerships are anticipated to become more important, along with the acceleration of sustainable development. This may either raise a warning flag or initiate positive developments (MacDonald, 2008). Hence, we believe that the growth of ENGOs could on one hand directly lift environmental supervision, information disclosure, environmental education, and environmental litigation. On the other hand, they indirectly motivate governments to strengthen formal environmental regulations. These could further force enterprises to enhance investment in environmental governance and finally crowd out polluting industries (Figure 1).

Figure 1

Mechanism of ENGOs’ crowding-out effect on polluting industries.

In all, the mechanism of ENGOs’ crowding-out effect on polluting industries can be explained through the following mechanisms:

Advocacy and policy influence: ENGOs often engage in advocacy efforts to promote environmental policies and regulations that limit the activities of polluting industries. By successfully lobbying for stricter environmental standards, ENGOs can crowd out polluting industries by making it more difficult or expensive for them to operate.

Public awareness and pressure: ENGOs play a crucial role in raising public awareness about environmental issues and the negative impacts of polluting industries. By increasing public scrutiny and pressure on these industries, ENGOs can make it more challenging for them to operate, thereby crowding them out.

Economic incentives: ENGOs may also work to promote sustainable business practices and green technologies, which can provide economic incentives for companies to shift away from polluting activities. As more companies adopt these practices, polluting industries may find it increasingly difficult to compete, leading to their crowding out.

Funding and resources: ENGOs often rely on donations and grants from donors who support environmental causes. By diverting funding and resources away from polluting industries and toward environmental initiatives, ENGOs can limit the growth and expansion of these industries.

Partnerships and collaborations: ENGOs may form partnerships with governments, businesses, and other organizations to promote environmental sustainability. These collaborations can result in the development of new policies, technologies, and practices that make it more difficult for polluting industries to operate, contributing to their crowding out.

Based on the above analysis, this article proposes the following research hypotheses:

Hypothesis 1: The development of ENGOs will promote the relocation of polluting industries.

Hypothesis 2: The development of ENGOs and the increase in regional environmental protection investment jointly promote the relocation of polluting industries.

3.1 Model

In our research cycle, ENGOs were not present at the beginning. Over time, some regions had ENGOs, while others had relatively lagging development of ENGOs. By the end of the research cycle, not all regions have ENGOs. We consider the emergence of ENGOs as a quasi-natural implementation, with regions with ENGOs as the treatment group and regions without ENGOs as the control group. The time after the appearance of ENGOs is used as the experimental time. The above analysis meets the setting principles of the difference-in-difference (DID) model. The dependent variable is the development level of the polluting industry under the influence of ENGOs, and the core variables are engodata and engopop: the number of ENGOs and their employees in the region, respectively; engodata and engopop are continuous variables between 0 and 1, rather than 0–1 dummy variables. Therefore, this article adopts the continuous DID method for estimation, with the following model settings:

where the subscripts i and t denote different cities and years, respectively; poll_loc is the overall development level of the polluting industry, which is represented by its location quotient; and Z represents other control variables that affect the location quotient of the polluting industry. γ is the corresponding coefficient matrix for the control variables; ξ is the random disturbance term; μ and υ are regional fixed effects and time fixed effects, respectively, to control for factors that change over time and across regions.

It has been acknowledged in the literature that the relationship between ENGOs and the location quotient of polluting industries is indirect. Based on previous research, we know that ENGOs have an impact on two major environmental governance investments: government investments and business investments made at the request of the government. Moreover, ENGOs have a direct influence on pollution industry governance investments. Therefore, to investigate whether ENGOs play a role in the relocation of polluting industries through environmental governance investments, interactions between the ENGO variables and environmental governance investments are introduced into the econometric equation. Specifically, the econometric model is as follows:

In the above two models, the coefficient of interest is ϕ . When ϕ is significantly negative, it indicates that ENGOs have indeed reduced the location quotient of polluting industries in the area through environmental governance investments, which also indirectly verifies the economic relocation effects and industrial upgrading effects of ENGOs. It is worth noting that environmental governance investments in this article are still measured by indices such as aggregate investment in environmental pollution treatments, investment in urban environmental infrastructure, investment in industrial pollution controls, three simultaneous environmental investments into construction items, and aggregate investments in environmental pollution treatments as a share of GDP.

3.2 Polluting Industries

Most studies on the relocation of polluting industries mainly draw from industry-related data, while determining the definition of a polluting industry is the first step in this study. Following different standards, there are three main classification methods for dirty industries in international academic circles.

The first method is environmental cost classification. Although it is difficult to calculate the pollution abatement costs of every industry, it is possible to calculate their pollution-control costs. By calculating the ratio of pollution-control costs to total value added, we can determine whether an industry is a polluting industry. It is easier to measure the ratio of pollution-control costs to total sales and rely on that ratio to define polluting industries. Through this method, it could be classified into five industries, including the papermaking, mining, nonferrous metals, steel, and chemical industries, as polluting industries. Specifically, we calculated each industry’s ratio of pollution-control costs to total production costs, and then through an empirical study, we determined the cutoff for this ratio to be 1.85%, meaning an industry is regarded as a polluting industry when its ratio is larger than 1.85%; otherwise, it is not. This method may be applicable to an entire country but not to a region, as its industrial structure is not necessarily complete.

The second method is pollution damage classification, which takes damage to the natural ecology and public health as its criteria. If an industry causes greater damage to these two features, it is classified as a polluting industry; otherwise, it is not. Through this method, McGuire (1982) ultimately classified 17 industries as polluting industries, such as mining, food manufacturing, tobacco and beverage manufacturing, textiles (including clothing, footwear, and hat manufacturing), fur and leather products, and papermaking.

The third method is pollution intensity classification using the Industrial Pollution Projection System (IPPS), which was proposed by the World Bank in 1994 and has mainly been used to evaluate the extent of industrial pollution. It is currently the most mature and widely used classification standard for polluting industries in the world. It is based on the calculation of pollution intensity for each firm, which is then used to classify industries as polluting or non-polluting. The IPPS is considered one of the most mature and widely used classification systems for identifying polluting industries globally. The IPPS methodology involves the following steps:

Calculation of pollution intensity: Using the four-digit Standard Industrial Classification (SIC) code, the pollution intensity of each firm is calculated. Pollution intensity is typically measured as the amount of pollution released per unit of production or revenue.

Comparison with critical value: The calculated pollution intensity is compared to a critical value, which serves as a benchmark for determining whether an industry is polluting or not. If the emission intensity of an industry exceeds this critical value, it is classified as a polluting industry.

Classification of industries: Based on the comparison, industries are categorized into polluting and non-polluting sectors. This classification helps in understanding the environmental impact of different industries and can inform policy decisions related to environmental regulations and sustainability initiatives.

Referring to the International Standard Industrial Classification (ISIC2.0) level, the method calculates each firm’s pollution intensity by four-digit Standard Industrial Classification (SIC) code, and pollution intensity is used as a criterion for defining polluting industries. If an industry’s emission intensity exceeds the critical value, that industry is considered a polluting industry; otherwise, it is not.

Then, Mani and Wheeler (1998) applied this IPPS methodology to classify American industries at the three-digit International Standard Industrial Classification (ISIC) level. Their study identified five sectors as the leading candidates for dirty industry status, which are the following:

Iron and steel: This sector involves the production of iron and steel products and is known for its high energy consumption and emissions of pollutants such as sulfur dioxide and nitrogen oxides.

Nonferrous metals: This sector includes the production of metals like copper, aluminum, and lead, which often involve the release of heavy metals and other hazardous substances.

Industrial chemicals: The production of chemicals like plastics, fertilizers, and pesticides can result in the emission of various toxic substances and greenhouse gases.

Pulp and paper: The paper industry is associated with water pollution due to the discharge of chemicals and wastewater from pulping and papermaking processes.

Nonmetallic mineral products: This sector includes the production of cement, glass, and other mineral products, which can contribute to air pollution through the release of particulate matter and carbon dioxide.

These sectors are considered to have higher pollution intensity than other industries, indicating that they have a significant environmental impact. The identification of these sectors as “dirty” can lead to targeted environmental policies and incentives to reduce pollution and promote cleaner production techniques.

As China’s industrial categorization differs from other countries to some extent, some scholars have proposed their own classification criteria. Otsuki et al. (2004) divided 20 two-digit manufacturing industries from the National Economical Industry Classification (GB/4754-2011) into three categories based on their average pollutant concentration: severely polluting, moderately polluting, and lightly polluting industries.

In terms of national policymaking, the Chinese State Council divided pollution sources into industrial pollution sources, agricultural nonpoint source pollution sources, domestic pollution sources, and centralized pollution sources in the First National pollution source survey program issued in 2006. Of these sources, industrial pollution sources include all secondary industries except for the construction industry and can be further divided into major and general pollution sources.

Based on international standards and existing studies in China, this article chose six polluting industries, i.e., mining, papermaking, chemical fiber manufacturing, nonmetallic mineral products, the smelting and processing of ferrous metals, and the production and supply of electric power. Additionally, since this article studies ENGOs’ role in environmental governance, we need to consider whether pollution behavior is easy to observe and therefore neglect industry-level pollution costs, damages, and intensity. The reasons we chose these six industries are as follows. For the mining industry, its major pollutant is the waste residue generated during the mining process that is piled up haphazardly and causes environmental problems, which are particularly easy for ENGOs to observe. The mining industry here includes all its subsectors. As the most traditional dirty industries, papermaking, and chemical fiber manufacturing have always been a focus of ENGOs because they usually produce a huge amount of sewage and odor. Nonmetallic mineral products and the smelting and processing of ferrous metals have also been representative of heavy industries, resulting in severe air, water, and soil pollution, another major cause of concern for ENGOs. The production and supply of electric power (mainly referring to thermal power plants and heating companies) are mainly fueled by coal, which also draws ENGOs’ attention.

3.3 Industrial Relocation

Existing studies have found strong evidence for the existence of pollution havens in China. Research integrating environmental regulations on polluting industries has found strong evidence in favor of the fact that polluting industries tend to relocate to regions with looser environmental regulations as well as to Midwest China. Once environmental regulations in adjacent regions are strengthened, polluting companies in those places may relocate to local areas, turning these areas into pollution havens. However, the economic relocation effect of ENGOs has not yet received much attention in academia. By contrast, this article aims to fill this research gap through an empirical examination. After defining the polluting industries of concern to ENGOs as above, we then focus on the relocation index for polluting industries. Previous studies have adopted the location quotient of polluting industries or the share of the output value of a region’s polluting industries in the output of the country as a measure of polluting industries. These indicators partly reflect the development level of certain polluting industries in a region. As we aim to determine whether the relocation of polluting industries is influenced by the development of ENGOs, industrial and growth indicators fail to reflect whether these industries have relocated or not. Ultimately, we chose the employment location quotient, denoting the study area as a region (j) within a nation (n) and using employment (worker) as the measure of economic activity. Then, the location quotient for industry i may be expressed as follows:

where worker ij represents employment in industry i and region j; worker in represents employment in industry i throughout the whole nation. It should be noted that industry refers to the six polluting industries and region refers to the 31 Chinese provinces in this article. Formula (5) compares the relative concentrations of employment in industry i in two economies (i.e., the industry’s share of employment in each economy). If the location quotient for an industry is greater than 1, it is assumed that the industry occupies an important position in this province. If it is equal to or less than 1 for an industry, it is assumed that this industry is not dominant compared with the average national level.

Since this article focuses on six polluting industries, we add up the six industries’ employment in the province to obtain total employment. For a country with j regions and i industries, let worker ij denote the employment in industry i in region j so that ∑ i = 1 6 worker ij is the total employment within the six industries in region j. Let worker in denote national employment in industry i, so that ∑ i = 1 6 worker in is the national employment within the six industries. The location quotient index for region j, poll_Loc j , is defined as in Formula (6).

It should be noted that the total location quotient does not add up to the sum of the location quotients for the six polluting industries. Rather, it is measured by comparing the share of employment in the six polluting industries at the province level with the share of employment in the polluting industries throughout the country.

In recent years, the utilization of the location quotient for estimating regional economic or industrial development levels has been widely used and has attracted an increasing number of scholars from various fields. Liu (2009) regarded the location quotient as a criterion for selecting regional pillar industries. In this article, the location quotient is used to measure the competitive advantage of a certain polluting industry in a province relative to the country, justifying its pillar position. Specifically, a decreasing location quotient indicates that this industry is losing its pillar position in the region, meaning it is relocating to other regions. In addition, instead of the output location quotient, this article adopts the employment location quotient index to measure the geographic concentration of the polluting industries across regions. If the employment location quotient declines, it is an indicator that this industry’s employment is less advantaged. This may be caused either by the industry relocating or by the industry introducing advanced production technology, leading to labor declines or transfers to other industries. Generally, most polluting industries are labor intensive; hence, using the employment share to calculate the location quotient index can serve as an indirect indicator of the employment effect and the economic relocation effect.

3.4 Explanatory Variables

In this article, we choose engopop and engodata as the core explanatory variables (equations (1) and (2)). However, ENGO data are not calculated in China. The data used in our study are mainly from the China Development Brief, a Chinese–English bilingual network platform founded in 1996 with a directory of NGOs. It provides professional observations, research, network platform support, and services to charity organizations, and the relevant information on most NGOs can be found through this platform (http://www.chinadevelopmentbrief.org.cn).

We define nongovernmental organizations engaged in the environmental protection business as ENGOs. The platform includes each ENGO’s information, such as its name, date of establishment, organization size (number of personnel), and business field, and identifies the ENGO’s registration place at the province–prefecture–county level. With these data, we can manually collect the relevant information on each ENGO and then pool those data to the province-prefecture level and obtain two-administrative-level ENGO data. It should be noted that the data currently available are relatively comprehensive, although they are not perfectly so. The data include whether there are any ENGOs (engo), the number of ENGOs (engodata), and the number of ENGO employees (engopop).

In this article, we use provincial-level data on the number of ENGOs (engodata) and the number of ENGO employees (engopop) as core explanatory variables. Then, we chose two years, 2000 and 2010, to analyze the trends in these two indicators. Figures 2 and 3 reveal the spatial distribution of the number of ENGOs in all provinces in 2000 and 2010, respectively. By comparing the two figures, ENGOs can be seen to have only existed in the coastal areas and Chongqing in 2000, where there were a larger number of ENGOs in the province-level municipalities of Beijing and Shanghai, and in Guangdong Province. However, ENGOs expanded into more provinces by 2010, although they were mainly still concentrated in coastal provinces and did not yet exist in every province. They began to appear in the southwestern region and Heilongjiang and Ningxia provinces. The central part of the country also saw the appearance of ENGOs. However, in Shanxi, Inner Mongolia, and Shaanxi, where environmental pollution is heavy, ENGO development lagged.

Figure 2

Spatial distribution of the ENGOs’ number of all provinces in 2000.

Figure 3

Spatial distribution of the ENGOs’ number of all provinces in 2010.

Figures 4 and 5 reveal the spatial distribution of ENGO employees in all provinces in 2000 and 2010, respectively. By comparing the distribution maps for 2010 and 2000, the regions with the largest number of ENGO employees can be seen to have spread from Beijing to Shandong, including Shanghai, Jiangsu, and Fujian Provinces. In addition, Hebei, Henan, Sichuan, Hubei, Zhejiang, Guangdong, and Guangxi provinces had the most ENGO employees. However, in 2010, Chongqing, Guizhou, and Heilongjiang provinces had relatively fewer employees. In Ningxia, Liaoning, Yunnan, Henan, and Jiangxi provinces, where there was only one ENGO each, the number of ENGO employees was small.

Figure 4

Spatial distribution of the ENGOs’ employees of all provinces in 2000.

Figure 5

Spatial distribution of the ENGOs’ employees of all provinces in 2010.

3.5 Summary and Data

We structure the variables as follows:

Dependent variables: the employment location quotient for the six polluting industries in a region ( poll_loc ij ) and the total employment location quotient ( poll_Loc j ).

Key explanatory variables: the number of ENGOs (engodata) and the number of ENGO employees (engopop) in a region. Our main estimation strategy follows the same logic as a standard difference-in-differences (DD) strategy. Since not every region always has ENGOs, we use a continuous measure for intensity and thereby capture more variation in the data. It should be especially noted that ENGO development varies considerably across the different parts of China, so we add 1 to each data point and take its logarithm.

In line with the former research, the following variables are used as the main control variables: the natural logarithm of per capita GDP (lnpgdp) and the natural logarithm of per capita GDP squared (lnpgdp2), which indicate the level of economic development; the industrial structure (ind), which is measured by the proportion of the secondary industry; the capital–labor ratio (lncap_lab), which is the ratio of fixed assets to the labor force; foreign direct investment (FDI), which is the ratio of FDI to GDP and is multiplied by 100%; population density (lnden), which is measured by the natural logarithm of the number of people per unit of area; and energy efficiency (en), which is the natural logarithm of electricity consumption per unit of GDP.

In addition, the variables we selected for detecting the underlying mechanisms are as follows: aggregate investment in environmental pollution treatments, investment in urban environmental infrastructure, investment in industrial pollution controls, three simultaneous environmental investments in construction items, and aggregate investment in environmental pollution treatments as a share of GDP. The data for these variables come from the China Environmental Statistical Yearbook (2003–2016). The employment data come from the China Labor Statistical Yearbook (2000–2016). Other control variables are derived from the China Statistical Yearbook and China Regional Economic Statistical Yearbook. Some missing data are estimated by interpolation, while those areas with large amounts of missing data are indicated with a missing value in this paper. Since employment data for each manufacturing sector are at the provincial level only, our final data set includes the 31 provinces in mainland China from 2000 to 2016 and from 2003 to 2016. The descriptive statistics of the main variables are shown in Table 1.

4.1 Benchmark Regression

Table 2 reports the benchmark regression results based on equations (1) and (2). All four models are estimated by controlling for time-fixed effects, and the models’ F-test values are greater than 3, indicating that the models have reached a significance level of 1% or better. In column (1), we estimate the single impact of engodata on the total polluting industrial location quotient ( poll_loc ), while in column (2), we add in other control variables. The estimation results show that when controlling for time-fixed effects and region-fixed effects, regardless of other factors, the coefficients on the core explanatory variable (lnengodata) are −0.014 and −0.012, respectively, which are significantly negative, indicating that when engodata increases by a percentage point, poll_loc drops by approximately 0.012–0.014. This change of 0.012 is incredibly large considering that the location quotient index is either above or below 1, an indication that ENGOs do influence reducing the importance of polluting industries. Columns (3) and (4) investigate the impact of the number of ENGO employees (engopop) on the importance of polluting industries. The estimation results show that when controlling for time-fixed effects and region-fixed effects, regardless of other factors, the coefficients on the core explanatory variable (lnengodata) are significantly negative and basically remain the same at −0.028, indicating that increasing engodata significantly reduces the importance of polluting industries. These four models verify the pollution haven hypothesis. That is, polluting industries prefer to relocate to regions with laxer environmental regulations. In this article, we find that polluting industries gather in areas where NGOs are weaker, while in areas where ENGOs are relatively mature, the importance of the polluting industry is significantly reduced. The possible reason is that in an area where the ENGOs have developed, under the influence of ENGOs, residents have strengthened their environmental awareness and raised their needs for environmental quality. If there are polluting enterprises around, the residents will react to the government through means such as safeguarding their rights and reporting, so that the government can adopt more stringent environmental management measures. Under the pressure of environmental protection, polluting enterprises must choose to relocate to other areas.

The estimation results for the control variables show that when the location quotient of the polluting industries is used as the explained variable, the impact of the overall economic development level exhibits an inverted U-curve similar to an environmental Kuznets curve. As the economy grows in a region, the location quotient of the polluting industries in that region increases first before it decreases, indicating that the share of polluting industries increases because there is no marked preference for the industry at the beginning of the economic development, which leads to environmental deterioration. However, when the region strengthens its environmental regulations and requirements with the development of its economy, its polluting industries decline. In addition, the increase in the proportion of the secondary industry greatly reduces the location quotient of the polluting industries because such an increase is mainly due to nonpolluting industries, an indication that the industrial structure has upgraded in the region, which leads to a decline in the proportion of polluting industries. The capital–labor ratio represents the type of industries, and its increase shows that industry has become more capital intensive, with workers being replaced by machines, so more energy is required and more pollution is generated. The increase in the proportion of foreign investment significantly reduces the location quotient of polluting industries because regions tend to allow cleaner foreign investment, which reduces the importance of polluting industries. Industrial electricity consumption dramatically reduces the proportion of polluting industries because the consumption of electricity indicates that machines are replacing labor, and there is a shift to more advanced industries, which causes the importance of polluting industries to decline. There is no significant relationship between population density and the location quotient of polluting industries.

4.2 Region-based Discussion

The vast territory and the unbalanced marketization of China (eastern China has a higher marketization level than midwestern China) result in remarkable regional differences (Chen et al., 2000), while the development of ENGOs varies significantly among regions (Bebbington, 2004). Therefore, ENGOs have heterogeneous influences on the location quotient of regional polluting industries. We chose 31 provincial-level administrative districts (according to China’s existing administrative system) in the mainland as geographic units and further divided them into two regions: East China, including Liaoning, Beijing, Tianjin, Hebei, Shandong, Jiangsu, Shanghai, Zhejiang, Fujian, Guangdong and Hainan Provinces, and midwestern China, including the remaining 20 provinces. Table 3 reports the estimation results for the above two regions. Columns (1) and (2) show the impact of ENGOs on East China. The results show that both coefficients are significantly positive, indicating that ENGOs have a crowding-out effect on polluting industries in the eastern region, which leads to a decline in the location quotient of total polluting industries in the region. Columns (3) and (4) show the impact of ENGOs on polluting industries in midwestern China. The estimation results show that the coefficients from the two models are neither significant nor different from the previous two, indicating that ENGOs do not significantly reduce the location quotient of polluting industries in these areas. This is caused by the fact that ENGOs in East China are relatively mature and have pushed the polluting industries into midwestern China. In East China, the relationship between the level of economic development and the location quotient of polluting industries follows a significant inverted U-shaped curve, whereas such a curve also exists in midwestern China, it is not significant. Some other control variables are not significant, but the signs on the coefficients are basically consistent with those from the benchmark regression.

4.3 Robustness Check

The fact that some areas suffer from substantial missing data at the provincial level and the number of employees in the polluting industries in those regions is relatively low may lead to inaccuracy in our estimation. Therefore, robustness tests are conducted in this section, in which observations from four regions, i.e., Tibet, Qinghai, Xinjiang, and Inner Mongolia Provinces were deleted for re-estimation. Table 4 reports the estimation results from the full samples and the subsamples. The first three columns present the impact of engodata on the polluting industries’ location quotient (poll_loc). The coefficients for the whole sample and for the eastern region are significantly negative and lower than those from the baseline regression, indicating that the estimation results are affected by the sample composition. For the midwestern China subsample, the estimated coefficient is positive but not significant. Once again, the results verify that in the full and eastern region samples, ENGOs have a crowding-out effect on polluting industries, an effect that does not exist in midwestern China, indicating that the midwest has become a pollution haven.

The last three columns present the impact of engopop on the polluting industrial location quotient (poll_loc). The results show that in the full sample and the eastern sample, the coefficients on the core explanatory variables are highly and significantly negative, but they are not significant in the midwestern sample. The six models all verify that ENGOs have a significant crowding-out effect on polluting industries in the full sample and the eastern region, while midwestern China has received some of the polluting industries from the eastern region and has become a refuge for polluting industries. The estimation results for the other control variables are in line with the estimation results from the benchmark regression and the previous regional regression, so they are not repeated here.

4.4 Industry-based Discussion

While the first three sections of this chapter mainly examine the total effects of polluting industries, this section considers the effect of each polluting industry individually. The variables poll_loc1-poll_loc6 represent the location quotient for mining, papermaking, chemical fiber manufacturing, nonmetallic mineral products, the smelting and processing of ferrous metals, and the production and supply of electric power, respectively. Table 5 reports the estimation results for the six polluting industries. The results show that engodata has a negative influence on the six polluting industries, while only in columns (2), (3), and (5) are the coefficients significantly negative, indicating that engodata has a negative impact on the location quotients of the following three industries: papermaking, chemical fiber manufacturing, and the smelting and processing of ferrous metals.

The impact of ENGOs on the other three industries, i.e., mining, nonmetallic mineral products, and the production and supply of electric power, is not significant. This may be because ENGOs have not closely monitored these three industries. First, compared with the other industries, the mining industry is not located in the center of a region, while ENGOs mainly serve urban areas or key cities. Then, the nonmetallic mineral products industry is dominated by cement production and includes glass, ceramics, gypsum, and other manufacturing industries, which are critical polluting enterprises that are often relatively evasive and difficult to supervise. Last, the supply of electric power mainly refers to thermal power plants and heating companies, which are industries that face rigid demand due to regional economic development and are mostly state-owned enterprises. In conclusion, ENGOs have relatively weak supervisory power over these companies, but their negative impact remains.

Table 6 reveals the impact of engopop on each polluting industry. The results show that engopop has a negative impact on every polluting industry, but its negative impact is significant on papermaking, chemical fiber manufacturing, and the smelting and processing of ferrous metals. Although the other coefficients fail to reach a significance level of 10%, the t values in columns (1) and (4) are both above 1.5; that is, the two models are significant within the 20% level. This indicates that engopop reduces the location quotients of the six polluting industries, causing a crowding-out effect that makes these industries relocate to regions with laxer environmental regulations.

4.5 Endogeneity Discussion

Endogeneity is an unavoidable issue in this article. Due to the continuous DID method used in this article, it can alleviate endogeneity issues to a certain extent, but it cannot fundamentally solve endogeneity problems. This article discusses endogeneity issues through the following strategies.

Lag of core explanatory variables: By lagging core explanatory variables (engopop and engodata) by one and two periods, it helps mitigate potential simultaneity bias. After changing the lagging variables, the results are still valid. These results indicate that the model has some robustness.

Parallel trend test and Placebo Tests: In the traditional DID model, the parallel trend hypothesis is the key, which requires that the processing group and the control group have the same trend before processing, while in our article, the continuous DID model may not be able to use traditional parallel trend tests.

In the continuous difference-difference (DID) model, the parallel trend hypothesis is a key premise for efficient estimation. This assumption requires that the processing group and the control group should have the same trend before the processing occurs. If this hypothesis is satisfied, then any observed change in the outcome after processing can reasonably be attributed to the processing itself, rather than other unobserved factors.

For traditional DID models, parallel trend testing usually involves comparing the differences between the treatment group and the control group before and after the treatment. However, in a continuous DID model, since processing changes on a continuous time scale, traditional parallel trend tests may not be applicable. This is because in continuous DID, it is difficult to define a clear “before” and “after” time period, because the processing effects may appear gradually.

Because the method of parallel trend test of continuous DID model is not mature, this test is not carried out in this paper. This means that the parallel trend hypothesis cannot be directly tested. Accordingly, placebo tests are difficult to perform because they often rely on being able to define a spurious processing time point and then check for the presence of a treatment effect before and after that time point.

4.6 External Validity

When discussing performance assessments of the economic relocation effects of ENGOs, external validity refers to the extent to which research findings can be generalized outside the study context. In other words, it concerns the universality of the findings. When assessing the economic relocation effects of ENGOs, this article considers how changes in the policy/industry context over time affect these relationships.

The following are some of the factors and consideration points that may influence the external validity of the performance assessment of ENGOs’ economic relocation effects and how these relationships change with changing policy/industry contexts:

Geographical and cultural differences: Different regions may have different cultures, values, and levels of environmental awareness, which can affect ENGOs’ operational efficiency and relocation decisions.

Policy environment: Policies such as tax incentives, subsidies, or other incentives may be different in different regions and at different times, which can affect the cost-benefit analysis of ENGO relocation.

Laws and regulations: Environmental regulations and land use laws may vary significantly in different regions and at different times, affecting ENGOs’ strategic choices and the effectiveness of their actions.

Economic conditions: Factors such as regional economic conditions such as GDP, employment rate, and economic diversity can affect ENGOs’ funding sources and the effect of relocation.

Technological developments: Technological advances could change the way ENGOs operates, for example, by reducing operating costs through improved energy efficiency or increasing its reach through social media.

Organizational maturity: More established ENGOs may have more resources and a greater ability to adapt to policy changes, while emerging organizations may have limited strategies and resources.

Global dynamics: International treaties and agreements, as well as changes in the global environmental agenda, can affect domestic policies and ENGOs’ framework for action.

Social attitudes and behavioral patterns: Changes in the level of public concern about environmental issues and behavioral habits, such as increased recycling, may also affect ENGOs’ strategy and effectiveness.

Funding flow: Changes in the source and amount of funding, including private donations, government grants, and international funding, can affect ENGOs’ relocation decisions and subsequent operations.

Data access and quality: The accessibility and quality of data can change over time, which affects the accuracy of long-term follow-up studies and performance evaluations.