The cultural economy of human waste reuse: perspectives from peri-urban Karnataka, India


 Safely managed waste reuse may be a sustainable way to protect human health and livelihoods in agrarian-based countries without adequate sewerage. The safe recovery and reuse of fecal sludge-derived fertilizer (FSF) has become an important policy discussion in low-income economies as a way to manage urban sanitation to benefit peri-urban agriculture. But what drives the user acceptance of composted fecal sludge? We develop a preference-ranking model to understand the attributes of FSF that contribute to its acceptance in Karnataka, India. We use this traditionally economic modeling method to uncover cultural practices and power disparities underlying the waste economy. We model farmowners and farmworkers separately, as the choice to use FSF as an employer versus as an employee is fundamentally different. We find that farmers who are willing to use FSF prefer to conceal its origins from their workers and from their own caste group. This is particularly the case for caste-adhering, vegetarian farmowners. We find that workers are open to using FSF if its attributes resemble cow manure, which they are comfortable handling. The waste economy in rural India remains shaped by caste hierarchies and practices, but these remain unacknowledged in policies promoting sustainable ‘business’ models for safe reuse. Current efforts under consideration toward formalizing the reuse sector should explicitly acknowledge caste practices in the waste economy, or they may perpetuate the size and scope of the caste-based informal sector.


GRAPHICAL ABSTRACT INTRODUCTION
Safe fecal sludge (FS) management is a necessity for 1.8 billion people in low-and middle-income countries (LMICs) that depend on septic tanks and pits for their sanitation needs (Berendes et al. ). These systems accumulate FS, which needs to be collected, transported, treated, and disposed of safely. If treated to the necessary standards, FS can be reused as a source of energy and/or nutrients in agriculture instead of being disposed of as waste.
At the same time, unregulated reuse carries many risks, such as endangering public health through inadequate safeguards, endangering worker health, and reinforcing the social stigmas associated with the work of human waste management.
In this paper, we estimate the factors influencing the latent demand for FS fertilizer (FSF) among farmowners and farmworkers in Karnataka, India -a significantly agriculture-based country that has made great progress with respect to latrine coverage but much less so with respect to post-latrine waste management (Coffey & Spears ; WHO/UNICEF/JMP ). In 2017, 93.7% of urban households in India had access to a latrine (WHO/UNICEF/JMP ). About half of all urban toilets are connected to a soak pit or a septic tank; however, as of 2019, out of 9,391 towns and cities in India, only 30 have functional FS treatment plants (Rao et al. ).
The term FSF usually refers to FS that has been treated for safe reuse as fertilizer or other uses. However, there are no studies that have tested the extent to which the FSF available at our study sites were fully or only partially composted; therefore, we use the term FSF to mean at least partially treated waste. Any mention of FS refers to untreated waste. The local term for treated or untreated human waste is bhangi gobbara. Bhangi is a derogatory term used for a subcommunity among Dalits who occupy the lowest rungs of the Hindu caste ladder. We repudiate the use of derogatory labels for any community, and we use the term in this paper only to reflect the ground realities of FS reuse and, by association, of sanitation work in India.
The literature on safe sanitation and sustainable agriculture has argued that FS should be managed as a resource rather than as waste (Keraita et al. ; Nikiema et al. ). The reuse of (treated) human manure for agriculture is actively being debated in India. Researchers have argued that productive waste reuse can generate revenues to partially recover the cost of waste collection and treatment (e.g. Murray & Ray ; Nikiema et al. ).
Others have developed innovative business models for waste collection, treatment, and reuse for agriculture or fuels, with estimated demand scenarios and stated willingness-to-pay studies (e.g. Danso et al. ). For any business models and reuse recommendations to be implemented, however, more needs to be understood about the preferences, priorities and social relationships that could drive demand for treated FSF. Here, we address this question using qualitative interviews with farmers in Karnataka, and a discrete choice preference-ranking model guided by these interviews.
Reusing human waste in India, a society rooted in a slowly changing caste-based hierarchy comes with special significance attached to those who handle human waste (Teltumbde ; Gatade a Because concepts of purity and pollution form a central part of Hindu religious beliefs and many elements of the caste system, any comprehensive discussion of FS reuse in India must engage with both ancient systems of oppression and modern efforts toward sustainable agriculture. The objective of our study is therefore to estimate farmowners' preferences for, and to gauge farmworkers' willingness to work with, FSF. If reuse becomes an official policy, composted FSF is the most likely treatment process to be used, since it can provide effective treatment while retaining the chemical components and structural integrity most prized by farmers. In addition to using discrete choice modeling to rank farmowner and farmworker preferences, we use these models to infer the influence of social pressures, caste-adherence, and truth in advertising on the use of FSF.

BACKGROUND
The use of human excreta as fertilizer is an age-old tradition (WHO ). Economically developed countries have largely disallowed the practice of using untreated or partially treated human waste in agriculture. In contrast, the reuse of untreated human waste, and specifically wastewater, in LMICs has flourished. Farmers in many countries continue to use wastewater as a source of water and nutrients (Radcliffe ). In addition, onsite sanitation systems, in which the toilet empties into a soak pit or septic tank, have grown rapidly in India. As manual emptying has declined as a practice, urban pits and septic tanks are increasingly ser- The reuse of human excreta in irrigation with at best partially treated, diluted wastewater has been estimated at up to 9 million ha in India (Thebo et al. ). The use of untreated FS is also prevalent, largely through unregulated and informal channels. Aside from the reuse of raw sludge delivered from septic tanks, several FSF entrepreneurs are in business, selling fully or partially composted waste to those farmers who accept its reuse. One recent study on farmers' attitudes toward human waste-based fertilizer products in South India found that more farmers were receptive to urine reuse over FS reuse (Simha et al. ).
Without treatment and safety regulations, however, informal reuse occurs in the shadows and remains risky for the health of farmers and the environment. We chose these cities because they both had systems for FSF reuse in their rural periphery, and a significant proportion of their urban populations used onsite systems. In the rest of this paper, we present our exploratory findings on the status of FS reuse in peri-urban Karnataka and on farmowners' and farmworkers' attitudes toward the use of treated human waste-based fertilizers. We conclude with the policy implications of these findings.

Initial interviews
We conducted our initial interviews to understand how and where FS reuse (raw or treated) was or was not taking place, and what farmowners' and farmworkers' perceptions of (at least partially treated) FSF might be. These interviews informed our later data collection efforts and helped us to design our stated preference survey. We interviewed 23 farmowners and 38 farmworkers from seven villages surrounding Dharwad and Bangalore. Additionally, we conducted two group interviews with farmowners and three group interviews with farmworkers, all in places at which farmers and farmworkers regularly congregated. As these interviews were, in effect, focus group discussions, their analysis is of a qualitative nature.
Apart from farmowners and workers, two sludge-selling entrepreneurs and seven truck operators who release waste onto farmlands were also interviewed. On 12 occasions, truck operators were accompanied in the act of collecting, transporting, and disposing of FS. The first author also visited farms to observe how the sludge was discharged, stored or applied to farmlands, and the behavior of farmworkers during work, meals, and rest breaks. Taken together, these interviews and observations helped us to understand the practices of and around FS reuse in peri-urban Karnataka, and acted as pilots for the surveys we designed to estimate willingness to use, and to work with, FSF.

Stated preference surveys
To more precisely estimate farmers' perceptions of using FS and gauge farmworkers' willingness to work with FSF, we conducted two stated preference studies, one that looked at farmowner willingness to pay and the other at farmworker willingness to use. To identify and recruit farmowners, we worked from a list from the Revenue Office in Dharwad.  ). Guided by our interviews with farmers, workers, and the two entrepreneurs, we identified six key attributes of FSF: Label, Smell, Health, Wetness, Texture, and Price of the fertilizer (or daily wages for workers, to understand whether they would be more open to using FSF if they were paid more). Each attribute, other than Price, had two levels, one with higher utility than the other. For example, many farmowners believed that packing FSF in bags labeled 'organic' would encourage their workers to handle FSF, whereas labeling it bhangi gobbara (or some synonym thereof) would discourage them. Therefore, we had two levels for the attribute Label; both said 'organic manure' but one also said 'bhangi gobbara'. Our choice instrument included two levels (present and absent) for the attribute Smell and two levels (wet and dry) for Wetness (see Supplementary Table S1 and Figure S1 for the full list). The choices also included either cow manure or chemical fertilizers as non-FSF options. Of the 2,306 farmowners, 1,807 completed the entire survey; of the 839 workers, 674 completed it.
We collected separate data (and estimated separate models) for owners and workers because the choice to buy FSF is fundamentally different from the choice to sell one's labor and be asked to use FSF. These surveys were meant to rank attributes of FSF as more or less acceptable; they were not intended to derive demand estimates for FSF. All the surveys and interviews were conducted in Kannada, the primary language in Karnataka.
Discrete choice surveys present respondents with distinct and mutually exclusive 'sets' from which they select a preferred option. We presented farmers and workers with three choice sets, each containing three options, and asked them to choose their favored option from each choice set.
Each choice set contained two FSF options with all the FSF attributes (at specified, randomly assigned, levels) (see Supplementary Figure S1 for the full list for an example).
The two options, guided by our earlier interviews, were designed such that no one option was deemed 'better' than the other across all attributes. Based on locally prevalent practices, the third option in each choice set was either cow manure or chemical fertilizer, also randomly assigned, at current market prices or at current daily wages (see Supplementary Table S1 for the full list).
Following Train (), we employed a multinomial logit specification to model the choices of the respondents and infer how they valued different attributes relative to each other. For a multinomial logit model with a linear-inparameter model specification, the utility of alternative j over choice set t as perceived by individual n, denoted u ntj , is written: where x ntj is a column vector of explanatory variables, such as the attributes of the options presented and the characteristics of the individual; β is a column vector of coefficients for these attributes (also known as 'taste parameters'), where each coefficient stands for the relative influence that its associated attribute has on a discrete choice decision; and ε ntj is the stochastic component of the utility. Let y ntj denote the choice indicator, equal to one if individual n chooses alternative j over choice set t, and zero otherwise. Under these assumptions, and assuming further that individuals are utility-maximizing, the probability that individual n chooses a sequence of choices yñ ¼ 〈y n11 , . . . , y nTJ 〉, where T denotes the number of choice sets faced by a single individual (equal to three in our case) and J denotes the number of options in any one choice set (equal to three in our case), may be given as follows: that the land could be fertilized (see also Wade ).
The growth of pit latrines in rural India has increased the opportunity to reuse human waste. Truck-based (as opposed to manual) pit emptying has also burgeoned in the last 20 years, but the truck operators we spoke to complained that the disposal of loads of sludge was both difficult and risky (see also Sharada Prasad & Ray ).
As a result, they have been encouraging farmers whose pits they empty to reuse their own FS. If a truck operator empties the pit of a farmer, he encourages the farmer to compost and reuse the sludge on his own farm, while the driver saves on transport time and costs.
Only seven of out of 23 farmers in our sample had used any form of FS in the past. They used it mainly for mango orchards, banana plantations and sugarcane, and usually right before the monsoon season. Though all seven agreed that using manure could protect the fertility of the soil, they felt that FSF use would reduce but not eliminate the use of chemical fertilizers. Sarkaari gobbara was apparently indispensable for obtaining a good yield. As FS was not easy to come by, some of these farmers provided their land for FS offloading and did not charge the truck operator. Others paid the truck operator to discharge the waste on their field, especially if they were not on the regular route. Both farmowners and workers who had used bhangi gobbara perceived it to be an organic manure similar to kottige gobbara (cow manure). It was more disgusting, certainly, but also more potent (olle powerru (good power); tumba fastu (very fast)). For yields, it was a 'super hit'.
A primary concern for farmers was that their workers, 'How can you even think of such a disgusting question? … Men don't even wash the bottoms of their own children, they call their wives to do that business. Do you think they will touch someone else's shit?'

Preference modeling
As explained in the "Methods" section, these initial interviews helped us to design discrete choice surveys through which we modeled farmer and worker preferences for the use of FSF. We estimated separate models for chemical fertilizers and cow manure (there was no way to combine them). Both models yielded similar results for Label, Smell, Health, Wetness, Texture, and Price of fertilizer; we are presenting only the cow manure models for simplicity.
We estimated separate models for farmowners and workers since they are able to exercise agency in a fundamentally different manner. We also estimated separate models for male and female workers because these two groups face different social pressures and expectations. We found in our interviews that religious or caste identity was not the same as adherence to religious or caste practices; we included diet in the model estimation for both owners and workers, as diet, such as eating or not eating certain meats, is a significant proxy for adherence.
We estimated the preferences of farmowners across all FSF attributes in Model 1 (see Table 1). Only those choice sets where the third option presented cow manure were included (n ¼ 1,018). With the exception of price, estimated coefficients for all attributes were significant at the p < 0.01 level.
In Model 2, we estimated the differential preferences Note: We added a dummy variable indicator for FSF to account for any differences between cow manure and FSF not covered in the attributes specified. Out of 2,306 farmowners, only 1,807 participated in the preference modeling study. Each participant was presented with three choice sets (see, e.g., Figure S1). Each choice set had three options to choose. With 1,807 farmowners × 3 choice sets per owner, we get 5,421 choice sets in total. Half of those choice sets (5,421/2 ≈ 2,710) had cow manure as the third option, and the other half had chemical fertilizer. We have n ¼ 3,054 (instead of 2,710) because the cow manure and chemical fertilizer choices were randomly assigned and their allocation was probabilistically but not exactly close to 0.5. Our n was also influenced by the workers who dropped out.
vegetarians and a majority of them were higher-caste, while just over half the surveyed workers were omnivorous. In Model 3, we estimated the differential preferences regarding smell, health risk, and labeling by farmer wealth. Based on the initial interviews, two proxies for farmer wealth, yielding four levels, were used: tractor ownership (has tractor/no tractor) and farm size classification (large area/other area).
Tractors also provide a physical barrier between workers We observed a strong preference for cow manure over FSF, even when controlling for the included attributes. In Model 1, dry FSF, no health risks, no smell, and tea powder texture were all significantly preferred. The estimated coefficients for these attributes were robust across all three models. We found that landholdings and tractor possession did indeed affect preferences regarding smell; farmers without tractors had a strong preference that smell be absent, especially if they had small-to-medium rather than large landholdings (Table 1). The same pattern was observed for health risk preferences. These proxies of farmer wealth had no impact on preferences with regard to labeling. Labeling as 'organic manure'without specifying that it was FSFwas preferred across the whole sample in Model 1. In Model 2, we found that all farmers, regardless of diet, were strongly opposed to smell, but the preference for an 'organic' (without bhangi gobbara) label was stronger among vegetarian farmowners.
In Models 4 and 5, we estimated the preferences of farmworkers, across all FSF attributes (Table 2). Only those choice sets where Option Three presented cow manure were included, and separate models were estimated for male and female workers. In both models, we interacted diet with labeling and with smell, in order to compare them with the owner models.
The FSF coefficient estimate was not statistically significant for either men or women when compared with cow manure, indicating that a milder preference for cow manure than the owners had expressed. There was no significant effect of wages on willingness to work with FSF.
Preferences for an absence of health risks were statistically significant across men and women workers, as they had been for the owners. Vegetarian men and all women had strong preferences for an absence of smell in FSF. As with their employers, omnivorous men and women, when comparing FSF with cow manure, preferred that the term bhangi gobbara be excluded from the label. Vegetarian women, however, preferred labeling that included the term bhangi gobbara; the labeling preference was in the same direction for vegetarian men, but the coefficient was not statistically significant.
Health impacts were not explicitly measured as part of this study. About 88% of the farmworkers surveyed said that they wore no protective gear when they worked with either manure or chemicals, and no farmowner reported providing safety gear to the workers. Sixty-four percent of the workers surveyed reported minor injuries during their work, mainly With 308 male workers × 3 choice sets per worker, we get 924 choice sets in total for males. Half of those choice sets (924/2 ¼ 462) had cow manure as the third option, and the other half had chemical fertilizer. We have n as 456 (instead of 462) for males and 553 for females (instead of 549) because cow manure and chemical fertilizer were randomly assigned; the proportions were only probabilistically close to 0.5. Our n was also influenced by the workers who dropped out.
scratches and bruises to their hands and feet. In addition, 59% of the workers surveyed carried drinking water in a plastic bottle and drank directly out of it during work, without washing hands. Almost 40% of the workers reported that they did not wash their hands or feet after work due to a lack of water.
All the female workers said that they cooked meals for their family and did domestic chores, which commenced as soon as they got home. Given these multiple pathways of exposure, it is not surprising that, across all models with farmers and workers, the preference for FSF with no health risks was strong and statistically significant. We used a combination of qualitative interviews and structured surveys with over 2,000 farmowners to understand the practices of FSF reuse for agriculture in Karnataka. We built a discrete choice (stated preference) model to understand the ranking for each attribute of FSF compared with other attributes, with respect to willingness to use FSF. We found that, across the board, dryness, no smell and no health risks associated with FSF were conducive to the acceptance of FSF (see Table 1); these findings reflect previous work that has found perceptions of cleanliness and aesthetics to be key influences in attitudes toward waste (Ban et al. ). Our models indicate that farmowner wealth, indicated by large landholdings, hired labor, and tractor ownership, has a mitigating effect on the reluctance to use (treated) fecal waste. We attribute this to the physical barrier between farmers and the FS when there is a tractor on the farm.
Our reliance on preference modeling to estimate the importance of individual product attributes relative to one anotherthe product in our case being FSFrepresents a traditional use of such models. This kind of modeling is We also use our discrete choice models for a more unusual purpose: we show that they can be used to infer the underlying social relations and caste practices within which FSF use is embedded. These practices include employer's willingness to mislabel a product they and their workers will handle; differences in choice between farmowner and farmworker; and the relationships among FSF acceptance, health concerns, caste-adherence, and the maintenance of caste 'purity'.
In addition, creating hypothetical choice sets allowed us to give voice to workers' preferences for, and to infer the constraints under which they must work with, FSF. Our stated preference models indicate that, for all its sustainability and (possible) revenue generation potential, FSF reuse in Indian agriculture is still undergirded by caste hierarchies and caste practices, reflecting the influence of caste and religion in rural sanitation overall (see Vyas & Spears ).
Caste-adherence and within-caste social standing emerged as significant predictors of whether farmers would work overtly with FSF, and with the acceptance of FSF as long as the marketing and labeling obfuscated the source. Innovations in treatment technologies, and business models that are being developed without consideration of how stigma and sustainability intersect, could inadvertently continue the caste-based disparities, nontransparent practices, and unsafe waste handling that characterize informal reuse today.

DATA AVAILABILITY STATEMENT
Data cannot be made publicly available; readers should contact the corresponding author for details.

This project was supported by the CGIAR Research
Program on Water, Land and Ecosystems (WLE). We also wish to acknowledge all of our interviewees and thank them for sharing with us their time and knowledge.
Funding from the NIH T32 training grant at Columbia