The elimination of Open Defecation (OD) is a critical initial step towards ensuring safe sanitation, with this being a basic prerequisite for improving human health [1]. The eradication of this practice has been central to achieving the target of universal access to sanitation by 2030, as was declared under the sixth SDG (Sustainable Development Goal). However, while such SDGs and global efforts to ensure safe and hygienic sanitation has no doubt bettered conditions, it is unlikely that the above ambitious goal shall be achieved within the stipulated timeframe [2, 3]. To stress the daunting task further, in 2017, OD was practiced by more than 673 million people worldwide, mostly in Sub-Saharan Africa and South Asia, with almost 50% of all such cases being from India alone [4].
Sanitation systems reveal especially high failure rates [5, 6, 7, 8] and sanitation interventions often lag behind initial policy expectations [9, 10, 11]. This is particularly true for past sanitation campaigns in India, such as the Central Rural Sanitation Programme (1986) or the Total Sanitation Campaign (1999–2012). The most recent endeavour, the Swachh Bharat Mission (2014–2019) (a restructured form of the Nirmal Bharat Abhiyan which ran from 2012–2014), has substantially increased toilet coverage across the country. However it remains to be seen whether this shall actually eliminate OD and foster sustainable sanitation changes, with toilet adoption still being far from universal [12, 13, 14, 15].
Possibly, the repeated failures or lowered effectiveness of past intervention policies have stemmed from sanitation practices being highly context sensitive and deeply embedded in human-environment interactions. Local sanitation conditions result from complex interactions between historically and culturally conditioned social behaviour, structural societal inequalities, multi-scalar sanitation politics and various natural environment parameters [16, 17, 18]. Thus the same factors have distinct and even opposite effects on sanitation outcomes in different geographical settings and furthermore, the interrelationships between distinct sanitation drivers are also context sensitive [19, 20, 10].
The implications of the contextual nature of sanitation have still to be fully understood. It makes the transferability of both, the policy interventions and any findings accrued from sanitation research, quite complex and difficult to generalize. Various measures have thus been recommended for adjusting the planned interventions to the targeted local environments [21, 22, 7]. This contextual sensitivity also amplifies an inverse relationship between the internal and external validity of garnered research findings. The experimental impact-evaluations represent an exemplar case in this respect. Although they attain high internal validity due to the effective elimination of contextual influences from the examined relationship between an intervention and its outcome, the methodological disregard of contextual factors means there is a weaker external validity (generalizability) of their results [23, 19, 11]. Contrarily, observational studies usually focus solely on statistical associations. As such, their internal validity is weaker. However, they can provide richer information on the roles of various observable contextual parameters and may also cover larger populations, which then adds to their external validity. On the other hand, inferences from larger-scale (i.e. ecological or population-level) studies are more susceptible to bias caused by various heterogeneities within the analyzed population. This challenges their ecological validity in terms of the extent to which such research results can be generalized to real-life settings [24].
The present article addresses the latter issue by examining how the inferences usually drawn about the correlates of OD practices in India are affected by heterogeneities demonstrated in space. More specifically, we focus on the problem of spatially varying relationships between OD and its covariates, which can be referred to as spatial non-stationarity or process spatial heterogeneity [25, 26]. With this in mind, our paper has two goals. Firstly, to examine the district-level correlates of OD in India, and secondly, to assess the role of spatial heterogeneity in these correlates by comparing the results obtained using a standard non-spatial regression model with those obtained from spatially explicit regressions and a Multi-scale Geographically Weighted Regression (MGWR). Our analysis uses data from the Census of India (2011) covering a contiguous area of 266 districts (their rural parts) across seven northern and eastern states of India, wherein 65% of all households practicing OD across the country reside. Although at present, the 2011 Census data may be already somewhat outdated, the more recent data required for such district-level analysis of OD is not yet available. To the best of our knowledge, this paper presents the first spatially explicit analysis of district-level sanitation correlates in India as well as the first quantitative analysis of the spatial non-stationarity of sanitation determinants globally. This article thus seeks to make not only an empirical but also a methodological and epistemological contribution. It demonstrates an approach that can be adopted for a meso-scale analysis of sanitation drivers in India, as and when more recent district-level data is available. Moreover, despite the recent decline in OD in India, the prevalence of this practice is still likely to be substantial [13, 27] and possibly has a pronounced spatial variation. Therefore, the ascertained empirical evidence about the OD determinants in this study can still be valuable for addressing sanitation issues in India and other countries.
The variables used in the present study (which act as the correlates of OD) contain measures of the general socioeconomic situation of the studied districts, their occupational structure and their socio-ecological conditions. Additionally, we also examine the role of the proportion of Muslims in districts’ population. Previous research has demonstrated that Muslim communities in India and elsewhere tend to have generally lower OD rates [28, 29, 30, 31]. In addition, Geruso and Spears (2018) [32] uncovered that, due to the comparatively better sanitation conditions of Muslim communities, the spatial co-concentrations of Muslim and non-Muslim neighbourhoods generate positive health externalities and decrease infant mortality rates in both Muslim and non-Muslim families. Inspired by this interesting finding, we have hypothesized that a greater proportion of Muslims in a district's population may not only lead to better health, but that it would also affect the sanitation conditions of non-Muslims. Since such externalities have a higher chance of arising in more densely populated areas [33], we expect that an area's population density shall moderate (amplify) the negative relationship between the share of Muslims and its prevalent OD rate.
Our results demonstrate that the parameters relating to the household ownership of assets, drinking water inaccessibility and literacy rate represent the most stable district-level correlates of the OD rate in India. Furthermore, our surmise about the negative association between the share of Muslim population and the district-level OD rates being contingent upon the level of population density was also confirmed. The latter finding supports our hypothesis about the positive sanitation effect of Muslim neighbourhoods and ultimately, the improved health externalities stemming from the spatial co-concentrations of Muslim and non-Muslim neighbourhoods. Overall, our analysis demonstrated a notable spatial clustering as well as a significant spatial non-stationarity of most of the examined variables. Therefore, studies that ignore the spatial heterogeneity of sanitation drivers possibly provide incomplete information, which may not be adequate for framing properly informed policies. While this claim of ours does have a general relevance with respect to most social and natural phenomena, we believe that sanitation represents a field in which geography matters particularly strongly.