Do users' perceptions of neighbors' cleaning behavior of shared sanitation facilities act as a credible threat in urban low-income communities? A case study of Khulna City, Bangladesh

Shared sanitation facilities (SSFs), which are common in low-income urban communities, are used by 540 million people worldwide, 330 million of whom live in urban areas in low- and middle-income countries (UNICEF & WHO 2019). SSFs, which refer to sanitation facilities shared by two or more households, have been reported to be less hygienic than individual household facilities (Kwiringira 2014; Heijnen et al. 2015; Aluko et al. 2018). Households that rely on SSFs are less educated and have a lower socio-economic status than those that do not (Heijnen et al. 2015). They have less interest in cleaning SSFs, which disposition is associated with lower education (Aluko et al. 2018). Furthermore, SSFs present a challenge in terms of sustainability. A study conducted in Uganda found that about 45% of respondents said they had abandoned their shared toilets and moved to new ones within five years due to malfunctioning or overflowing (Günther et al. 2011).

Previous studies have investigated the psychological and socio-economic factors influencing users' cleaning and maintenance behaviors with respect to SSFs. This includes an examination of normative factors, social dynamics, and social capital (Tumwebaze et al. 2014; Tumwebaze & Mosler 2015; Aluko et al. 2018). However, there is another aspect of the management of SSFs that encourages the avoidance of cleaning responsibilities or free-riding. Different types and management systems of SSFs are found in low-income communities (Mazeau et al. 2014). For example, in communities with landlords, landlords bear the responsibility for investing in SSFs, while their tenants provide labor to maintain them (Simiyu et al. 2017; Chipungu et al. 2019). In communities without landlords, community-based organizations (CBOs) may assume the role of managing SSFs (Mazeau et al. 2014; Simiyu et al. 2017). CBOs and community groups are often mobilized by non-governmental organizations (NGOs) and local governments to use and maintain SSFs (World Bank 2009). In this type, SSFs are constructed by NGOs or local governments at little or no cost to the community, but it is common for communities to take responsibility for their subsequent management, such as regular cleaning and collecting money from users for repairs, without guidance from an external authority. Therefore, in both cases, SSF users may have an incentive to avoid contributing to the maintenance of these resources. This can be observed in behaviors such as delaying the payment of maintenance fees or avoiding cleaning responsibilities. This observation indicates that SSFs are akin to common pool resources (CPRs) (Ostrom 1999, 2002; Simiyu et al. 2017).

CPRs, similar to natural resources such as shared fisheries, forest, or water resources, are prone to social dilemmas that place users in a conflict between cooperation and non-cooperation. Thus, CPRs face the risk of ‘the tragedy of the commons’ (Hardin 1968). The ‘tragedy of the commons’ often occurs in the absence of a self-governing system or external regulations. There is a high probability that, when applied to communities that share SSFs, the absence of rules or systems for users of SSFs to pay penalties or contribute to maintenance may result in the ‘tragedy of the commons’. In the absence of such rules or systems, users of SSFs have an incentive to avoid contributing to the maintenance of the SSFs.

Some studies examine SSF users' behavior based on the assumption that SSFs are analogous to CPRs. A review identified social dilemma factors in SSF cleaning behavior, such as group size, gender, social norms, and trust among the same SSF users (Tumwebaze & Mosler 2014a). Tumwebaze & Mosler (2015) showed that group discussions and commitment, such as declaring cleaning responsibility, are associated with an increase in self-reported cleaning behaviors.

Simiyu et al. (2017) applied the eight principles of CPRs (Ostrom 1993, 2002) to explain the quality of SSFs. For example, clear boundaries, such as fences and locks, published rules, and collective decision-making, influence the quality of SSFs. Similarly, Chipungu et al. (2019) applied the eight principles to investigate the social dynamics of SSF users. In their study, the presence of physical boundaries and the implementation of penalties for those who violate the rules were identified as key factors influencing the maintenance of SSFs.

Although these studies have focused on identifying factors associated with the behavior of SSF users or the quality of SSFs, the discussion has yet to focus on the structures that cause the tragedy of the commons in CPRs. By focusing on the structure that users are involved in regarding SSFs, it may be possible to identify factors that could overcome social dilemmas and prevent the tragedy of the commons.

Like the traditional theory of the structure of CPR, game theory provides models of how individuals act and cooperate with each other when facing a social dilemma (Osborne & Rubinstein 1994). When a CPR is abandoned, individuals are described as choosing ‘non-cooperative’ behavior as their rational choice as individuals. At this point, the structure of the game becomes a one-shot or finitely repeated game. This implies that the game will be played only one shot or always end at some point, and non-cooperation is explained as the most rational choice for the individual. When a CPR is maintained, however, the structure of the game is not one shot anymore but repeated infinitely. When the game is repeated, i.e., not terminated, cooperative behavior becomes the rational choice for the individuals (Dal Bó 2005; Dal Bó & Fréchette 2018).

This mechanism has been supported by the Folk Theorem (Osborne & Rubinstein 1994). In infinitely repeated games, individuals estimate the future payoff compared to their current payoff. This future payoff is calculated from a discount factor R (0 < R < 1). If R is sufficiently close enough to 1, the players estimate a higher payoff in the future when they choose to cooperate than defect. In other words, if R is close enough to 1, the players are patient enough to cooperate even if the other players choose to defect. Consequently, a high discount factor allows a player to choose cooperative action as an optimal choice.

The determinant that increases cooperative behavior, i.e., the discount factor, has been discussed in virtual games of cooperation conducted in laboratory settings (Dal Bó 2005), virtual resource management (Osés-Eraso & Viladrich-Grau 2011), or resource consumption in field settings (Amirova et al. 2019). The discount factor is affected by the presence of ‘credible threats’, such as sanctions or punishments from other users or authorities. This study focuses on the question of whether users perceive the behavior of others as a credible threat. The act of flushing and cleaning a toilet after use is a private behavior. However, when performed in an SSF, it can be regarded as cooperative behavior, as flushing implies not only the user's own hygiene but also the cleaning of the next person in line, including their family members. Additionally, obtaining water from outside places is a burden on users, which may encourage them to skip or free-ride.

Previous studies have provided empirical evidence of self-governed CPRs from the field (Ostrom 1990), such as governance or the rules established by farmers, fishers, or the appropriators of the CPRs. However, there has been no clear evidence of the game structure of self-governed CPRs. Focusing on the game structure may help identify its credible threat that influences users' cooperative behavior in CPR management.

This study was conducted on the premise that SSFs can be considered a CPR and that the emergence of free riders and the tragedy of the commons can be attributed to the one-shot or finite iteration game structure. The objective was to examine whether the game structure of SSF users' behavior, which does not fall into the tragedy of the commons, follows an infinitely repeated game structure. Furthermore, this study examined whether users' perceptions of their neighbors' SSF behavior could serve as a credible threat to their own behavior.

The specific hypotheses of this study are derived.

• H1. In communities where SSFs are maintained without the intervention of others, users' behavior is structured as an infinitely repeated game.

• H2. In communities where cooperation is maintained through an infinitely repeated game structure, the perception that others are cooperating constitutes a credible threat to the users.

• H3. When the perception that others are cooperating is not a credible threat to the users, the users' perception of others' behavior is associated with their socio-psychological factors.

In this program, flushing SSF after use is one of the targeted hygiene behaviors, along with washing hands before eating and after using the toilet. The four communities targeted for the project were selected based on the conditions of their SSFs, the communities' size, and their attitude toward outsiders. The SSFs were relatively clean, and the communities had been maintaining SSFs without external support (Supplementary material, Figure S1). The average community size was approximately 100 households, and the residents were generally cooperative with outsiders. The most common type of SSFs observed in these areas consisted of two to four pit latrines built on an elevated platform to prevent the pits from flooding during the rainy season and floods. The workshop program targeted female residents who are or have been caregivers because they spend more time in the community and, therefore, use SSFs more than others. In addition to these women, the study included female residents who spent the most time in the community and who were willing to participate in the survey. A total of 403 respondents participated. Their demographic information is presented in Table 1.

This study used a structured questionnaire based on the RANAS model (Mosler 2012). The RANAS model is a behavior model for hygiene and sanitation that integrates several behavioral theories and enables us to estimate users' readiness for targeted hygiene behavior (Mosler 2012; Tumwebaze & Mosler 2014b; Tumwebaze et al. 2014; Seimetz et al. 2016). ‘RANAS’ refers to five factors, such as risk, attitude, norm, ability, and self-regulation, which consist of five-factor blocks to develop a behavior change program. The questionnaire included questions related to SSFs, such as the practice of flushing SSFs after use and the frequency of handwashing after using SSFs. It also encompassed hygiene behaviors, such as the practice of handwashing before meals and the safe storage of drinking water at home. Socio-psychological factors such as social pressure, and impressions of daily life were also included in the questionnaire. To gain insights into the game structure, this study selected questions pertaining to users' own behavior, the behavior of others, and socio-psychological factors. The questionnaire consists of the three hygiene behaviors based on the RANAS model and questions related to socio-psychological factors. However, only the questions regarding self-regulation of flushing SSFs after use and social pressure were used for analysis (Supplementary material, Table S1). The first question indicates ‘self-cooperative behavior’, the second indicates users' perceptions of others' cooperative behavior, and the third refers to users' experience of being told to clean SSFs after use. Questions regarding socio-psychological factors were also used to explore the factors associated with users' perceptions of their own habits of cleaning SSFs, which are shown in Supplementary material, Table S2.

The surveys were conducted in June 2018 (Communities A and B) and September 2019 (Communities C and D). Surveys in Communities C and D followed the same methodology as those in Communities A and B, although they were only conducted after ethical approval for the former. Each household was visited by enumerators, and female respondents who agreed to be interviewed were included in the survey, followed by an interview at the participant's home. The printed questionnaire was completed by enumerators who interviewed the respondents. The number of respondents from Communities A to D was 102, 98, 102, and 101, respectively, giving a total of 403 female residents aged between 17 and 75 years who spent most of their time in the community. One respondent from Community B was excluded from the analysis owing to a lack of demographic information. Thus, data from 402 respondents were ultimately used for the analysis.

Second, a linear probability model was used to regress a user's socio-psychological factors against their perception of others' SSF cleaning behavior. This study focused on the game structure of the community that uses SSFs; thus, it is necessary to identify the important parameters that may influence the game. This study examined the impact of users' perceptions of others' behavior as a credible threat to their own cleaning habits.

This study used a linear probability model with a dichotomous dependent variable instead of a logistic regression model because of its interpretability compared to a logistic regression model, although these two models can show almost identical results (Hellevik 2009). To avoid the risk of violating residual heteroscedasticity, robust standard errors were used in this study.

Analyses were conducted using the Stata statistical software (version 17.0; Stata Corp., College Station, TX, USA).

The survey of Communities A and B was granted ethical approval by the University of Marketing and Distribution Sciences, Kobe, Japan (No. 2018005). The survey of Communities C and D, which was conducted after ethical approval had been obtained, followed the same ethical considerations as the previous survey. All participants received information on the nature and purpose of the study, and the survey was conducted only with those who signed consent forms.

Table 1 presents information on the 402 respondents in the four communities. Monthly expenditures and refrigerator ownership were included in the questionnaire to determine the economic status of each household. Community D had the highest expenditures of all communities; compared to A and B, monthly expenditures in C and D varied more among households.

Table 2 portrays the distributions of Category I, the four types of SSF users, across all communities. Community C had the highest proportion of selfC-othC (87.3%), whereas Community A had the lowest (50.5%). All communities had a high proportion of selfC-othC and selfC-othD, which means that more than half of the people in the community were cleaning after SSF use. Community C had the highest percentage (92.2%), whereas Community A had the lowest (55.9%).

Table 3 shows the distribution of Category II based on users' perceptions of their own behavior and their experience of being asked by others. Communities A and B had a lower proportion of selfC-othS (15.7 and 13.0%, respectively), whereas Communities C and D had a higher proportion (70.0 and 62.0%, respectively) of this occurring. Communities A and B had the largest proportion of selfC-othNS, whereas Communities C and D had the largest proportion of selfC-othNS; all communities had a small proportion of selfD-othS.

Table 4 shows the results of the linear probability model, which examined the impact of SSF users' socio-psychological factors on their perception of others' behavior in each community. The results for Communities B, C, and D showed no significant variables, but, for Community A, the variable ‘shame’ significantly increased the possibility of ‘othC’ (p < 0.000); however, the variables ‘monthly expenditure’ and ‘norm’ significantly decreased the possibility of ‘othC’ (p = 0.046 and 0.036, respectively).

This study examined the game structure of users of SSFs in low-income urban communities that are maintained without external support, focusing on users' perceptions of other users' behavior and their own habits of using SSFs. Based on the proportions of four types of respondents observed from the four communities, the game structure of the communities that maintain their SSFs is most likely an infinitely repeated game. This study also identified the differences between communities in which the perceptions of others' habits seemed to be a credible threat and a community in which the perceptions of others' habits were not perceived as a credible threat. Previous studies have focused primarily on socio-psychological or socio-economic factors when examining the management of SSFs (Tumwebaze et al. 2014; Tumwebaze & Mosler 2015; Aluko et al. 2018). However, this study adopted a game theory perspective and analyzed the structure of the game, which enabled the researchers to focus on the mechanisms that affect SSF users' behavior.

The proportions of the four types of respondents in Category I did not yield results in either community that would indicate the highest proportion of ‘selfD-othD’ combinations, i.e., the possibility of a social dilemma. On the other hand, the combination ‘selfC-othC’ was the most common combination of the four types in all communities. The combination ‘selfC-othC’ indicates the people who perceive that both neighbors and themselves flush SSFs after use. These results of the observed proportions indicate that the game structure of these communities is likely to be an infinitely repeated game, allowing ‘cooperation’ to be chosen as an individually rational choice and avoiding the ‘tragedy of the commons’. Furthermore, all types of respondents were observed in all communities, which may also be a characteristic of an infinitely repeated game. As shown by the Folk Theorem, infinitely repeated games allow both cooperation and defection as individually rational choices, because the repetition of the game allows individuals to sometimes deviate. These results support H1.

In Communities B, C, and D, the ‘selfC-othC’ combination was observed in over 70% of the cases, the highest percentage of the four types. As described above, the game structure of a community in which cooperation can be an individually rational choice should be infinitely repeated. In an infinitely repeated game, the discount factor should be high, which implies the presence of a credible threat. Thus, if the majority of CPR users choose to cooperate, there should be a credible threat associated with their choice. In this study, the perception that others would cooperate was consistent with the users' perception that they themselves would cooperate for more than 70% of the respondents. This suggests that the perception of neighbors flushing SSFs after use may pose a credible threat to respondents in communities or groups with self-governed CPRs, thereby supporting H2.

However, Community A showed a lower percentage of ‘selfC-othC’, which indicates that Community A may have a different game structure from the other three communities. In Community A, despite the identification of all user types, which suggests that Community A adheres to the principles of an infinitely repeated game, 50% of users were identified ‘selfC-othC’, which is clearly less than the other communities. It seems that from this proportion, the perception of others' behavior might not work as a credible threat to users' cooperative behavior in this community. This implies that there may be different factors influencing users' cooperative behavior when there are fewer possibilities for credible threats.

The results of Category II showed that the proportion of those who had experienced being told to clean SSF by other users was not consistent among the communities. The proportion of the users who had both a habit of cleaning and experience of being told to clean SSFs was not consistent either. This may indicate that requesting compliance from someone who has not followed proper hygiene practices might not be effective as a credible threat.

The influence of being directed by another individual to adhere to a particular behavior on users' behavior has not been previously examined. Previous studies have investigated the efficacy of punishment or risk-dominant conditions for cooperation, yet the findings are inconclusive, and the effectiveness of the punishments or risk-dominant conditions remains uncertain. One experimental study showed that the presence of costly punishment acts as a credible threat to free riders and increases the level of cooperation even in a one-shot game context; furthermore, this study found that strong emotions trigger punishment (Fehr & Gächter 2000). However, another study revealed that the effectiveness of punishment or risk dominance in inducing cooperation may be limited. Instead, the extent of increased cooperation appears to be influenced by players' perceptions of their payoff for cooperation or concern for the future (Dal Bó & Fréchette 2011). Alternatively, the perception of a high probability of punishment may be an effective credible threat to cooperation (Almeida 2023).

The results of the linear probability model showed differences between the respondents of Community A and the other communities. In Communities B, C, and D, perceptions of others' behavior work as a credible threat, but Community A may deviate from this structure. In addition, Community A showed a significant relationship between users' perceptions of others as cooperative and their social psychological factors, such as shame, guilt, or normative perceptions, whereas the other three communities did not. In other words, perceptions of others' habits seem to be independent of users' socio-psychological factors. However, for the latter community, users' perceptions of other users' behavior might have been influenced by their own socio-psychological factors. Although this study targeted female residents, it is worth considering because they are the majority of users of SSFs. Thus, the result supports H3. In other words, it is likely that the users of Community A judged the behavior of others more critically than users in other communities did. This is because ‘shame’ can lead to a heightened sensitivity to social norms and expectations. In communities without credible threats, some users may develop heightened social norms and feelings of shame, which may lead to a sense that SSF should be maintained.

This study has several limitations. First, as it focused only on a few communities that maintained their SSF without external support, limited generalizations can be made; moreover, biases based on the characteristics of the participants cannot be dismissed. Second, as the survey strategy involved face-to-face oral interviews, the answers might have been prone to social desirability bias. Further research should be conducted to better understand the game structure of SSF users by comparing more communities where SSFs are well managed to those where they are not. Uncovering the game structure of well-managed common pool resources may lead to the identification of community or collective structures that allow users to cooperate with each other.

This study analyzed the game structure of the community in which the SSFs are maintained by users without external support. The findings indicate that in communities that maintain their SSFs, the game structure is most likely to be an infinitely repeated game, and their perceptions of other users' cooperative behavior are a credible threat, which induces the users' cooperation regardless of their socio-psychological factors. However, in communities where a credible threat is absent, the users' socio-psychological factors, such as shame or norms, may influence their perceptions of other users' behavior and induce their negative evaluation of others' behavior, which, in turn, leads to their cooperative behavior. To gain a more comprehensive understanding of CPR users' cooperative behavior, it would be beneficial to consider not only individual factors but also the game structure of the community or group in which CPRs are shared.

The authors extend their gratitude to all study participants. The authors also thank Dr Akira Sakai, Professor Emeritus of the University of Marketing and Distribution Sciences, for providing a wealth of data for analysis and for his insightful commentary. Additionally, the authors extend their gratitude to Mr Md. Abdul Hadi for his assistance in conducting the surveys. Furthermore, the authors acknowledge the contributions of Professor Hiroyuki Nakata and Associate Professor Maiko Sakamoto, University of Tokyo, for their valuable comments.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.