1. Introduction
Environmental injustices in India have most often been discussed in terms of extreme events (e.g., the 1984 Bhopal Gas Tragedy) or struggles launched by social movements (e.g., against bauxite mining in Niyamgiri hills, Orissa) [
1]. In the process, chronic contexts of pollution that are not characterized by visible forms of resistance often receive less attention even as pollution maybe equally, or even more, inequitably distributed here. This argument especially holds true for hazardous industrial pollution which needs to be studied in terms of the social characteristics of the population that resides in industrial regions and is likely to be exposed to the harmful effects of hazardous waste production.
India’s position as an “emerging economy” is partly linked to its industrial performance [
2,
3], and industrialization here has a highly uneven distribution across the country [
4]. Gujarat in western India is one of the leading states in terms of industrial production [
5], as well as one of the top states in terms of the production of hazardous waste [
6]. A number of studies have focused on explicating the causes of successful industrial development in Gujarat [
7,
8,
9] or drawn attention to deepening economic and social polarization within the state [
10,
11]. A distributive environmental justice (EJ) analysis of hazardous industrial pollution would contribute to this body of research by linking the spatial patterns of industrial development with social inequalities.
This article aims to address an important research gap in EJ studies in India, as well as contribute more broadly to studies of industrial development, by analyzing the relationship between the spatial distribution of hazardous industrial facilities classified as major accident hazard (MAH) units and pertinent socio-demographic factors in the highly industrialized state of Gujarat. Specifically, the objective is to determine if socially disadvantaged communities in Gujarat reside disproportionately in areas burdened by higher densities of MAH units, further subdivided by the type of MAH unit as defined by capacity (large versus medium/small) and sector (private versus public). Data utilized for this study combines the locations and characteristics of Gujarat’s MAH units in 2014 with population and housing information obtained from the 2011 Census of India. The unit of analysis is the subdistrict (known as
taluka in Gujarat), an administrative division in India below the state and district levels, and the smallest geographic unit for which data on population and housing characteristics pertinent to our study are available in the 2011 Census. The subdistrict becomes a useful unit of analysis for Gujarat due to the concentration of industrial development around urban centers which could become masked at the relatively large scale of the district. Our statistical analysis is based on generalized estimating equations (GEEs) that account for geographic clustering of subdistricts within districts and provide statistically valid insights on the association between MAH unit density and specific socio-demographic characteristics of the population. Overall, our study becomes significant for gaining a better understanding of the negative aspects of industrialization in Gujarat to balance against the often unequivocal highlighting of positive aspects in governmental and corporate reports [
5,
12].
3. Results
The density of MAH units at the subdistrict level is depicted as a classified choropleth map in
Figure 1, which also shows district boundaries in the state. The MAH unit density value of each subdistrict is used to group subdistricts into four quantiles. Subdistricts with the greatest MAH unit density (highest quartile or top 25%) are located primarily in three districts: Bharuch in south Gujarat, and Ahmedabad and Vadodara in central Gujarat. These three districts collectively contain 254 of 402, or about 53% of all MAH units analyzed in this study.
We began our statistical analysis by examining bivariate linear correlations between each independent variable and five MAH unit density variables, respectively. The Pearson’s correlation coefficients associated with each pair of variables are presented in
Table 2. The density of all MAH units was significantly and positively correlated with population density, urban population proportion, and literacy rate, but negatively correlated with the proportion of home owning households. A similar pattern was observed for all four subcategories of MAH units in the state. The proportion of SCs or STs, however, was not significantly correlated with any of the five dependent variables.
The results from the multivariate GEEs are summarized in
Table 3,
Table 4 and
Table 5. The first model (
Table 3) used the density of all MAH units as the dependent variable. This table indicates that MAH unit density was significantly related to all independent variables (
p < 0.05), except for literacy rate. After controlling for the effects of other explanatory variables, the density of MAH units was significantly greater in subdistricts with higher proportions of SCs, STs, and urban population, but significantly smaller in districts that were more densely populated and had a higher home ownership rate. Although literacy rate showed a positive association with MAH unit density, this relationship was not statistically significant.
For the GEEs in
Table 4, MAH units were classified based on production capacity. The density of large capacity industries was significantly and positively related to the proportions of the urban and ST population, but indicated a significantly negative association with home ownership rate. For density of medium/small capacity industries, all independent variables showed a significant relationship. More specifically, the density of medium/small industries was positively related to the proportions of the urban population, SCs, STs, and literates, and negatively related to population density and home ownership rate.
The GEE models in
Table 5 allowed us to compare the socio-demographic distribution of MAH units by sector. The density of private sector industries was significantly higher in subdistricts with higher proportions of the urban, SC, and ST population, and lower proportion of home ownership. In contrast, for the GEEs representing public sector industries, SC and ST proportions were not significant. Instead, the density of public industries was positively related to the urban proportion and literacy rate, but negatively associated with population density and home ownership rate.
4. Discussion
The statistical results of this study provide several insights on social inequalities associated with the distribution of hazardous industrial facilities in Gujarat. Overall, a greater concentration of MAH units was significantly more likely to be found in subdistricts that were more urbanized, less densely populated, contained a higher proportion of socially disadvantaged residents (both SCs and STs), and a lower proportion of home-owning households, after accounting for geographic clustering in the data. When MAH units in Gujarat were classified by capacity and sector, almost similar distribution patterns and social inequities were observed for large capacity industries (except for the SC population), medium/small capacity industries, and those belonging to the private sector, respectively. Public sector industries represent the only subcategory that did not indicate a significant statistical association with the proportions of the SC and ST population.
With regard to the socially disadvantaged groups, bivariate correlation analysis did not indicate significant associations between MAH unit density and proportion of SCs or STs. After controlling for the effects of clustering and other independent variables in our multivariate GEEs; however, we found a significantly positive relationship between the SC proportion and the overall density of MAH units, as well as the densities of medium/small capacity and private sector units. We also found density of all MAH units, as well as the densities of large capacity, medium/small capacity, and private sector units, to be significantly greater in subdistricts with a higher proportion of the ST population. These results indicate the need to more carefully understand the distribution of SC/ST groups in Gujarat to determine whether they have migrated towards the employment opportunities provided by hazardous industrial facilities or if these industries have found it easier to locate in areas where socially disadvantaged groups reside. Given that SC and ST groups in Gujarat are more likely to be found in rural areas [
48,
49], their significant presence in subdistricts with higher MAH unit density which are also urbanized suggests an environmentally inequitable distribution.
When variables denoting socioeconomic status are considered, literacy rate suggested a positive association with several MAH unit density subcategories, after controlling for urbanization and other explanatory variables. This could imply that MAH units tend to concentrate in areas with higher availability of educated laborers for industrial jobs. The proportion of home owners, however, indicated a consistent and significantly negative association with the overall density of MAH units and all subcategories examined. While these results suggest that economically disadvantaged residents who cannot afford to purchase a home reside near hazardous industries, this finding could also reflect lower rates of home ownership in urban subdistricts of Gujarat with fewer affordable housing options. As mentioned previously, lack of rental housing in India is viewed as an impediment to the mobility of workers who may not want to purchase a house [
54]. Our results suggest that rental housing stock is coincident with highly polluted subdistricts, which points to either the high costs of home ownership around industrial facilities, or the unwillingness of those with the means to purchase housing to reside near hazardous industries. Overall, this leads to the conclusion that home ownership is a very useful variable to pursue in future analyses of distributive EJ in India.
In terms of the control variables of this study, densities of all MAH units, medium/small capacity industries, and public sector industries were found to decline with an increase in population density. This finding is similar to those reported in national-scale EJ studies conducted in India and the U.S., which demonstrate a negative association between population density and hazardous industrial pollution after controlling for urbanization [
37,
38,
41]. With respect to Gujarat, medium/small capacity and public sector MAH units were more likely to locate in urbanized subdistricts that were sparsely populated, possibly due to these having higher availability of vacant land that were proximate and accessible to major urban centers. This result coincides with previous EJ research that has depicted sparsely populated urban areas as having a lower ability to control the presence of industrial pollution in their vicinity [
41]. Large and private sector industries in Gujarat, however, are concentrated in larger urban subdistricts and do indicate a statistically significant association with population density. The extent of urbanization, as measured by the urban proportion, significantly influenced the distribution of all MAH units and the four subcategories examined, even after controlling for other socio-demographic variables. Thus, urbanization continues to attract industrialization in Gujarat despite government efforts to shift industries to rural and less populated and polluted areas [
14].
Finally, it is important to consider specific limitations of our study that are related to the unavailability of more detailed information on hazardous industries and potentially affected populations. First, data on industries and industrial pollution continues to be difficult to access in the context of India despite some steps taken towards rectifying the situation through the Environment (Protection) Act 1986 and Central and State Pollution Control Boards. Thus, the quantity or quality of pollutants emitted by each MAH unit are not available and this prevents us from assessing human health risks posed by hazardous industries based on exposure and toxicity. Second, although MAH units store or transport the highest quantities of toxic chemicals and pose the greatest health risks for local residents compared to other facilities, they are not the only source of industrial pollution in Gujarat. For a more comprehensive EJ assessment, it is also necessary to analyze industries that manage smaller volumes of toxic substances, as well as facilities that are involved in the treatment, storage, and disposal of industrial hazardous waste. Third, our study is based on socio-demographic variables from the Census of India, which represent residential characteristics of subdistricts. The hazardous industries examined in this study can have adverse effects on not just where people live, but also where people work and conduct other daily activities. This implies that even if a socially disadvantaged subdistrict contains few or no hazardous industries, residents of the subdistrict could be exposed to pollution generated by these industries in non-residential locations such as places of work, education, and shopping. It is thus important to explore additional data sources, including surveys at the household level, that can provide a more fine-grained analysis of the EJ implications of industrial development in India.