The World Bank Economic Review, 38(4), 2024, 824–845 https://doi.org10.1093/wber/lhae011 Article Group Incentives for the Public Good: A Field Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Experiment on Improving the Urban Environment Carol Newman , Tara Mitchell, Marcus Holmlund, and Chloë Fernandez Abstract What strategies can help communities to overcome the public goods problem in the maintenance of communal spaces and infrastructure in urban environments? This paper investigates whether an intervention targeted at Community-Based Organizations can motivate them to make increased contributions to the public good, thereby improving outcomes for the community as a whole. Using a randomized controlled trial conducted in Dakar, Senegal, the analysis tests the effectiveness of a program that provides incentives to community groups to encourage them to keep their neighborhoods clean, with the ultimate goal of reducing flooding. After one year, the intervention proved to be effective in engaging communities, improving cleanliness, and reducing flooding. JEL classification: O12, O13, O18, H41, D71 Keywords: infrastructure, public goods, community-based organizations, flooding Carol Newman (corresponding author) is a professor in the Department of Economics and Trinity Impact Evaluation re- search center (TIME), Trinity College Dublin, Dublin, Ireland; her email address is cnewman@tcd.ie. Tara Mitchell is a professor in the Department of Economics and Trinity Impact Evaluation research center (TIME), Trinity College Dublin, Dublin, Ireland; her email address is mitchet@tcd.ie. Marcus Holmlund is a Research Manager and Senior Economist at the Development Impact (DIME), World Bank, Washington, DC, USA; his email address is mholmlund@worldbank.org. Chloë Fernandez is a Research Officer at the Development Impact (DIME), World Bank, Washington, DC, USA; her email address is cfernandez2@worldbank.org. This research is part of the World Bank–assisted Senegal Stormwater Management and Cli- mate Change Adaptation Project (P122845) (PROGEP), which was funded by World Bank, the government of Senegal and the Nordic Development Fund. The authors are very grateful to the editor, Eric Edmonds, and three anonymous reviewers who provided valuable comments and suggestions that have significantly improved the paper. The authors are also thankful to their partners at the Municipal Development Agency (ADM), Pierre Coly and Mouhamadou Ba. They are also grate- ful to Andrea Guariso, Denis Jordy, Isabelle Kane, Arianna Legovini, Gabriele Rechbauer, Fatim Seck, Eric Lancelot, Amos Abu, Farouk Banna, an anonymous DIME reviewer, and participants at conferences and seminars including the NEUDC at Tufts University, the CSAE conference at Oxford, as well as Queens University Belfast, Groningen University, the University of Copenhagen, the Graduate Institute Geneva, and the Copenhagen Business School for valuable comments and insights. The authors thank Molly Offer-Westort, Violaine Pierre, Felipe Dunsch, and Ababacar Gueye for invaluable assistance with project coordination and data collection, as well as research assistants Margaryta Klymak, Michell Dong, and Qiao Wang. This work benefitted from generous funding from the UK/FCDO through the i2i Trust Fund, the Bank-Netherlands Part- nership Program, the Nordic Development Fund, and the World Bank. The RCT was registered on the AEA RCT Registry on July 7, 2015 (AEARCTR-0000754). Computational reproducibility was confirmed by DIME Analytics. A supplementary online appendix for this article can be found at The World Bank Economic Review website. C The World Bank 2024. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. This is an Open Access article distributed under the terms of the Creative Commons Attribution 3.0 IGO License (https://creativecommons.org/licenses/by/3.0/igo/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. The World Bank Economic Review 825 1. Introduction Public investment in infrastructure has accelerated in low- and middle-income countries (LMICs) over the last two decades, but large gaps still remain. The need for public infrastructure is also growing at a fast pace in LMICs in particular, given the disproportionate impact of climate change in the poorest regions of the world and the role that climate-resilient infrastructure can play in adaptation and mitigation strategies. Governments in LMICs already invest around $1 trillion (3.4 to 5 percent of gross domestic product) in Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 infrastructure every year (Hallegatte, Rentschler, and Rozenberg 2019). However, infrastructure adequacy, quality, operations, and maintenance remain insufficient, and infrastructure-related disruptions still cost households and firms in LMICs at least $390 billion per year. The maintenance of public infrastructure presents a local public-goods problem, and as such consider- ation must be given to the actions of local communities. Developing strategies to encourage communities to make contributions to this public good is essential for the effective functioning of the infrastructure and the sustainability of public infrastructure investments over the longer term. In fact, even the existing aggregate volume of investment could be of a sufficient order to achieve infrastructure-related Sustainable Development Goals if it were accompanied by appropriate policies (Rozenberg and Fay 2019). This paper demonstrates the effectiveness of a program providing primarily nonmonetary incentives, in the form of public recognition, targeted directly at key existing subgroups within the community in improving the maintenance of public infrastructure. This intervention relates to the maintenance of public drainage infrastructure in urban flood prone areas. Twenty-three percent of the world’s population faces significant flood risks, 89 percent of whom live in LMICs (Rentschler, Salhab, and Jafino 2022). Flooding leads to deaths and damage to property, but also has significant indirect effects such as interruptions to education because of school closures and an increased incidence of water-borne diseases. As a result of climate change, the effects of flooding are predicted to worsen in the coming years and will have the greatest impact on low-income regions. Urban areas are particularly vulnerable to flooding because of informal settlements on unsuitable land (Moser and Satterthwaite 2010). The setting for the present study is two low-income peri-urban areas of Dakar in Senegal: Pikine and Guédiawaye, which represent about 12 percent of the national population. These two municipalities are highly prone to flooding, and approximately half of the residents live in flood-prone areas.1 Some ma- jor investments in new drainage infrastructure have been made in these areas of Dakar in recent years. However, such infrastructure has historically suffered from a lack of maintenance and misuse, including the dumping of household waste. To overcome this public goods problem, this study designed an inter- vention called “Operation Clean Neighborhood” (Opération Quartier Propre or OQP) to try to shift behaviors and encourage local populations to keep their community and infrastructure clean. The anal- ysis uses a randomized controlled trial (RCT) to test the effectiveness of OQP, which targets established community-based organizations (CBOs) and encourages them, through social recognition and low-value in-kind incentives, to work toward keeping their neighborhoods clean. After one year, the study finds that OQP is effective in engaging communities and in improving their perceptions about the cleanliness of the neighborhood: CBOs are almost 27 percentage points more likely to include cleaning events as one of their activities (74 percent higher than the control sample mean at endline), households are 5 percentage points more likely to have received training on flood prevention (53 percent higher than the control sam- ple mean at endline), and households are 12.3 percentage points more likely to rate their neighborhood as clean (34 percent higher than the control sample mean at endline). Importantly, there is some evidence that this may also lead to a reduction in the reported levels of flooding (22 percent less than the control 1 This is not unusual in low-income settings. Pervin et al. (2020) describe similar problems in many parts of the world focusing on cities in Bangladesh and Nepal. Winter and Karvonen (2022) provide a review of the literature on flood gov- ernance in peri-urban areas revealing similar flood-related challenges in cities in Ghana, South Africa, Kenya, Thailand, India, China, Argentina, Mexico, and Brazil, among others. 826 Newman et al. sample mean at endline). This study uncovers important differences in the effectiveness of the program between areas that have had increased investment in infrastructure and those that have not, with the latter benefitting more from the intervention in terms of reduced flooding. There is also some evidence of im- pacts of reduced flooding on self-reported health: individuals in treated neighborhoods are 2 percentage points less likely to report that they were ill in the previous 30 days, 8 percent less than the control sample mean at endline. This paper also addresses the issue of spillovers, an important consideration in densely Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 populated urban centers. This study contributes to three strands of literature. The first is the extensive literature on the commu- nity provision of public goods. One branch of this literature focuses on the topic of community-driven development (CDD). CDD involves communities in the planning and implementation process for de- velopment projects (Beath, Christia, and Enikolopov 2015; Madajewicz, Tompsett, and Habib 2021; Quatrocchi et al. 2021). A number of studies show that community-based approaches can be effective at delivering local public goods (Casey 2018). Many community-based development programs involve significant information and facilitation campaigns, providing guidance on how best to provide the public good (Guiteras, Levinsohn, and Mobarak 2015; Cameron, Olivia, and Shah 2019; Abramovsky et al. 2023). Others aim to overcome the collective-action problem by providing incentives directly to individ- uals for providing a public service. For these types of interventions, the nature of the incentive has been found to be an important factor. The literature on the effectiveness of financial rewards in improving the impact of community-engagement projects is mixed. While there is some evidence that they can work in certain contexts (see, for example, Kremer et al. (2009); Cappelen et al. (2016); and Jayachandran et al. (2017)), the evidence for using nonfinancial rewards is also appealing (see, for example, Ashraf, Bandiera, and Jack (2014a); Ashraf, Bandiera, and Lee (2014b); Ariely, Bracha, and Meier (2009); Deserranno (2019); Frey and Oberholzer-Gee (1997); and Gneezy, Meier, and Rey-Biel (2011)). This paper builds on the findings of this literature by combining different aspects of each of the ap- proaches described above into the program design. A key aspect of CDD is harnessing local knowledge that is specific to the context. This is done by working with existing groups within the community and giving them significant decision-making power about the type of activities to engage in. This paper’s main contribution to this literature lies in combining these aspects of CDD with providing direct incentives to these groups in the form of social recognition if they successfully achieve provision of the public good. Another branch of the literature on the community provision of public goods focuses on the role of community-based approaches in addressing the maintenance of the shared environment. While the lit- erature is scarce, there are some notable examples, including Arunachalam et al. (2012), who evaluate the impact of a program aimed at reducing the spread of dengue in communities in Chennai, India. Part of this program involved improving the cleanliness of the community through engaging with existing women’s self-help groups. They find significant effects of the program on households’ knowledge and awareness, and also very significant reductions in factors responsible for spreading dengue in treatment communities. Support for grassroots community-level engagement is also provided by Sheely (2013) for the case of antilittering in Kenya. A number of other studies discuss the importance of including commu- nity groups in environmental-improvement programs such as improving the street environment (Adams et al. (2017), maintaining sustainable drainage systems (Everett and Lamond 2016), and the provision of urban infrastructure (Ibem 2009). These papers do not, however, carry out a test of the impact of these interventions. The present study contributes to this literature by providing new evidence, using an RCT, on the effectiveness of incentivizing community-based organizations to maintain local public goods in an urban setting. The second strand of literature to which the study contributes is the experimental literature related to threshold public goods. The relationship between cleanliness of the community and flooding represents a threshold public goods game, as it is necessary to reach a minimum level of cleanliness for drainage infrastructure to function properly. This study’s intervention creates an additional threshold public goods The World Bank Economic Review 827 problem for the CBOs, as they must achieve a minimum level of cleanliness in order to receive the reward of public recognition. Most of the studies examining the effective provision of threshold public goods have been based on lab or lab-in-the-field experiments. The present paper contributes to this literature by providing new evidence, based on a field experiment, on how to overcome the threshold public goods problem in settings where local community groups can play a role in reaching the threshold. Moreover, most field experiments examining public goods problems such as these have largely been studied in rural Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 areas (Carlsson, Johansson-Stenman, and Nam 2015; Bakhtiar et al. 2021; Aflagah, Bernard, and Viceisza 2022). Little is known about achieving this kind of coordinated action in a densely populated urban community. The design of this study’s program draws on a number of findings in this literature. The factors influenc- ing contributions to threshold public goods have been extensively tested in lab settings. A key component in such games is communication and coordination. A large literature exists on the role of group size in coordination games. In a meta-analysis of public-goods games, Croson and Marks (2000) find that suc- cess rates decrease as the group grows larger. This is further supported by Feltovich and Grossman (2015) who find a negative relationship between group size and communication and cooperation. Communi- cation has also been shown to improve through increasing interlinkages between units (Ostrom 2009). Moreover, the coordination of beliefs about how others will contribute to the public good is related to increased contributions to the public good (Turiansky 2021). In this study’s intervention, targeting CBOs both reduces the size of the group that needs to work together and focuses on individuals with existing forms of social capital that are already interconnected.2 The value of the reward and the way in which participation is incentivized have also been shown to be important components (Cadsby and Maynes 1999; Croson and Marks 2000). Brandts and Cooper (2006), for example, use a series of controlled lab experiments to investigate the impact of increasing the financial incentives to contribute to the public good and find experimental evidence to show that increasing the bonus that the groups receive breaks the coordination failure and allows the groups to coordinate at the highest equilibrium level.3 Even once the bonus is removed, the groups still cooperate to achieve the higher equilibrium level. Bripi and Grieco (2023) also find that contributions to the public good increase when additional participatory incentives are provided, in their case in the form of an additional prize awarded to the group. In this article’s intervention, the social recognition provided by the program increases the value of the reward, providing an additional incentive to contribute, and also creates a focal point that could shift beliefs about the contributions of others. Most significantly, this experiment shows that while the incentivized program is targeted at a particular subgroup within the community, the impact of the program is to reduce the experience of flooding for the whole community: community cleanliness is a public good not just for the CBOs but also for the entire community. This insight may be particularly important for understanding how to achieve public-good contributions in an urban setting. The third related area to which this study contributes is a small but growing literature that provides experimental evidence of the effectiveness of policies aimed at improving cleanliness and reducing flood- ing in urban communities. Nepal et al. (2023) investigate the impact on cleanliness of an information campaign coupled with the provision of street waste bins in Nepal. They find that the intervention had a positive impact on self-reported neighborhood cleanliness and appropriate methods of waste disposal. Information interventions related to flood risk and possible mitigation strategies were found to have posi- 2 The importance of communication and information sharing is also highlighted by Mitra, Buisson, and Bastakoti (2017) in a study that focuses on a setting with a similar public-goods problem to this study’s. In their case, they investigate contributions in a lab public-goods game designed to mimic contributions toward canal maintenance in Bangladesh and find that the presence of leaders and increased communication both result in increased contributions. 3 Brandts and Cooper (2006) consider a weak-link game, which shares some similar properties with threshold public good games. 828 Newman et al. tive impacts on households’ mitigation behavior and reduced the presence of solid waste around canals in Mozambique (Leeffers 2023), and a positive effect on the uptake of flood insurance in Bangkok (Allaire 2016). In addition, a number of studies have found that comprehensive risk-reduction programs can have a positive impact on the information that households have about flood risk, their level of preparedness, and their ability to manage this risk (Bottazzi et al. 2018; Sarabia et al. 2020). None of these studies mea- sured the impact of these interventions on households’ actual experience of flooding. The present study Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 contributes to this literature by providing evidence from an RCT that an incentivized community engage- ment program led to a significant improvement in the perceived level of cleanliness of the neighborhood and, possibly, a reduction in the experience of flooding. The rest of the paper is structured as follows. Section 2 provides more background on the context of the study and describes the intervention in more detail. The experimental design is presented in section 3. Section 4 describes the data and the empirical approach. Section 5 presents the results, and section 6 concludes. 2. Context and OQP Intervention Senegal’s vulnerability to natural disasters and climate change is due to its 700 km coastline, latitudinal position, and major river systems. The country ranks ninth in the world for the largest share of its urban population living in low-elevation coastal zones (McGranahan, Balk, and Anderson 2007). Nearly 40 percent of new residents in peri-urban Dakar settle in areas with high hazard potential, particularly inland flooding (GFDRR 2011), and lack of urban planning, coupled with rapid, dense population growth, has had disastrous environmental consequences, such as desertification, trash accumulation, and blockage of natural drainage systems. The lack of drainage systems and maintenance, as well as obstruction of natural drains due to urbanization, prevent stormwater evacuation, and increase groundwater levels, exacerbating vulnerability. In 2012, a large infrastructure investment project, the World Bank–assisted Senegal Stormwater Man- agement and Climate Change Adaptation Project (Projet de Gestion des Eaux Pluviales et d’Adaptation au Changement Climatique, or PROGEP) was launched in Pikine and Guédiawaye, where this study took place. PROGEP involved the construction and rehabilitation of drainage infrastructure in flood-prone areas as well as flooding sensitization campaigns. The study tests the effectiveness of an intervention OQP that was designed as a complement to infrastructure investments.4 The project areas lacked a well- functioning solid-waste management system and, consequently, suffered from widespread dumping of trash in public spaces. This clogged drainage infrastructure, reducing its effectiveness and ultimately con- tributed to flooding in the area. The objective of OQP was therefore to encourage CBOs to work within their local communities to improve and maintain the cleanliness of public spaces and drainage infras- tructure with the overall aim of reducing flooding in these communities. Through the promise of public recognition, cleaning materials, and small prizes such as T-shirts, CBOs were incentivized to try to achieve a certain level of cleanliness within their neighborhood. More specifically, OQP functioned as follows: 1. A focal CBO was identified in each neighborhood. Selection criteria were broad and included the stipu- lation that the CBO’s activities should be mainly confined to a single neighborhood (to maximize local knowledge and minimize spillovers), that it should have some minimum capacity needed to participate, and that it should have some influence within the community.5 4 OQP was designed and implemented in partnership with the Senegalese Municipal Development Agency (Agence de Développement Municipale). 5 If a flood management CBO was present in the community, it was automatically selected as the focal CBO. This was the case for 3 percent of the CBO sample (5 CBOs). The World Bank Economic Review 829 Table 1. Communes, Neighborhoods, and Groups in the Study Area Communes Neighborhoods Subgroups Pikine 6 349 42 Guédiawaye 1 47 5 Dakar 1 2 1 Total 8 398 48 Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: This table illustrates the total number of communes, neighborhoods, and subgroups in each of the study areas. 2. CBOs were supported in the development of action plans, giving them guidance on how to innovate and implement cleaning activities in their local neighborhood. 3. CBOs received an initial endowment package, consisting of cleaning tools and materials.6 4. CBOs signed letters of engagement with the commune’s mayor, affirming their intent to participate in OQP. This was considered the start of the intervention. 5. CBOs were first inspected at mid-term, six months after the start of the intervention.7 CBOs that passed the threshold were recognized for their achievement through a public ceremony and received in-kind low-value goods such as plastic chairs, cooking utensils, and T-shirts. 6. CBOs were inspected a second time 12 months after the start of the intervention. Rewards for passing the threshold score included a similar ceremony and token prizes such as T-shirts and cooking utensils. All selected CBOs were aware of the prizes and the cleanliness criteria required to meet the threshold grade; 64 percent of neighborhoods received the end-line prize. 3. Experimental Design The study uses an RCT to evaluate the impact of OQP on cleanliness and other flood-related outcomes. The unit of randomization is the quartier, or neighborhood. Pikine and Guédiawaye are comprised of 16 and 5 communes, respectively, which are formal political structures with an elected local government. Neighborhoods are primarily informal geographic groupings within those communes; though there are often local governance structures and community leaders, these are informal institutions that vary in nature across neighborhoods. In total, there are 398 neighborhoods in the PROGEP area. For political reasons, to ensure even representation across the area, neighborhoods were stratified into 48 subgroups on the basis of geographic and social ties. The total number of communes, subgroups, and neighborhoods in each area is provided in table 1. Of the 398 neighborhoods, 160 were randomly selected for inclusion in the study with proportional representation of neighborhoods in each subgroup. This number was primarily informed by the available budget for OQP, which was sufficient to cover 80 treatment neighborhoods. Randomization of neighborhoods was done within these subgroups with the number of treatment and control neighborhoods proportional to the size of the group. A total of 80 neighborhoods were assigned to the OQP treatment group and 80 to the control group. CBOs were selected in all 160 study neighborhoods, prior to random assignment. Random assignment into treatment and control neighborhoods was carried out through a public lot- tery held with representatives of all CBOs selected for the study, stratified by grouping, in which the representatives themselves drew their CBO’s treatment status. A public lottery was selected for maxi- 6 The contents of the endowment are detailed in table S1.1 of the supplementary online appendix. 7 The evaluation criteria are described in table S1.2 of the supplementary online appendix. A neighborhood had to achieve a grade of 60 or more in order to be awarded the mid-term or the final prize with the same grading scale applied to both evaluations; 68 percent of neighborhoods received the mid-term prize. 830 Newman et al. Figure 1. Map of Intervention Area. Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Source: PROGEP (Senegal Stormwater Management and Climate Change Adaptation Project) baseline data. Note: This map was produced using data from PROGEP and presents the treatment and control quartiers in 2014. mum transparency and to avoid allegations of corruption or clientelism to the maximum extent possible. Operationally, this was considered essential, especially as the proximity of neighborhoods and the pop- ulation density made it highly unlikely that control neighborhoods would not eventually find out about the intervention. As such, the study pays close attention to spillovers in its empirical analysis. Finally, it should be noted that PROGEP’s infrastructural component (rehabilitation of old water basins, implementation of new drainage pipelines, etc.) was rolled out in two phases.8 Infrastructure was implemented in Phase 1 areas between 2014 and 2016 and in Phase 2 areas between 2017 and 2019. The randomization of OQP was stratified across these two phases of infrastructure construction and was implemented simultaneously in these areas. This means that new infrastructure had already been con- structed in Phase 1 areas but had not yet been constructed in Phase 2 areas when OQP was implemented. This makes it possible to investigate the heterogeneous impact of OQP, depending on the existing level of infrastructure in the area. A map of the intervention area and the selected neighborhoods in Phase 1 and Phase 2 of the larger infrastructure project, PROGEP, is provided in fig. 1. The full timeline of the project is described in fig. S1.1 of the supplementary online appendix. 8 Phase 1 covers Dalifort-Thiouorour (communes of Wakhinane Nimzatt, Djeddah Thiaroye Kao, Dalifort-Foirail, Hann Bel-Air, and the western halves of Yeumbeul Nord and Yeumbeul Sud), and Phase 2 covers Yeumbeul-Mbeubeuss (Keur Massar, Malika, and the eastern halves of Yeumbeul Nord and Yeumbeul Sud). The World Bank Economic Review 831 Table 2. Baseline Household Characteristics n Mean total Mean control Mean treatment p-value Characteristics of household head Male 2, 285 0.69 0.70 0.68 0.29 Muslim 2, 285 0.95 0.96 0.95 0.23 Married 2, 285 0.79 0.80 0.78 0.10 Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Age 2, 285 55.41 55.45 55.36 0.86 Salary (FCFA per day) 1, 062 10884.97 12, 656.05 9, 113.88 0.12 Household size 2, 285 10.17 10.14 10.19 0.82 Cleaning and flooding–related characteristics Cleanliness of 2, 285 0.30 0.29 0.30 0.51 neighborhood Flood victim in last year 2, 285 0.22 0.21 0.23 0.36 Flood prevention training 2, 285 0.15 0.16 0.13 0.04 Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: p-value based on a test of the statistical significance of the difference between the mean of the control group and the mean of the treatment group at baseline; n indicates the number of observations at baseline. 4. Data Collection and Empirical Approach Data were collected at the household and CBO levels. In each neighborhood, 15 households were ran- domly chosen to be surveyed, giving a total sample size at baseline of 2,400 households. The household questionnaire collected information on household demographics, livelihoods and income sources, socioe- conomic characteristics, health outcomes, exposure to flooding, knowledge of risk-mitigation methods, and attitudes toward community participation and one’s general responsibilities vis-à-vis the community (and vice-versa). Within each household, data relating to certain variables (education, health, etc.) were collected for all individual household members, giving a sample of 28,010 individuals. The study focuses most of its analysis on the household sample. The CBO level questionnaire was administered as a group survey and focused on basic group character- istics, motivations for participating, attitudes toward civic participation, and the nature of CBO activities. The end-line survey differed slightly for treatment and control CBOs, as it aimed to collect data to gauge the awareness of control CBOs about OQP and whether this had influenced their activities, in order for the study to analyze spillovers. The baseline survey was carried out in July 2015 prior to the randomization to avoid any potential anticipation effects. The end-line survey was carried out in November 2016. The attrition rate was low. Out of the 2,400 households surveyed at baseline, only 115 could not be included in the end-line survey (an attrition rate of less than 5 percent). Table 2 presents summary statistics and balance checks for a range of baseline household character- istics. The first set of variables relates to the characteristics of the household head. Approximately 70 percent of household heads are male, and the vast majority (96 percent) are Muslim. Most are married (around 80 percent), either in monogamous or polygamous marriages. The average age is 55, and the av- erage household size is 10 people. The average daily salary of household heads (where available) in FCFA is 10,885 (approximately US$18). The data are perfectly balanced on all household-head characteristics at baseline. The second set of variables, the outcome variables, include the perceptions of households in terms of the cleanliness of the neighborhood, whether they were victims of flooding, and whether they received any training in relation to flood prevention by local CBOs. To measure cleanliness, households were asked to rate the cleanliness of their neighborhood on a 5-point scale ranging from “Very Clean” to “Very Dirty.” A “Clean” dummy was created, which is equal to 1 if the household rated their neighborhood as “Clean” or 832 Newman et al. “Very Clean.” To measure flooding, households were asked questions about their experience of flooding in the most recent rainy season and in the previous season. Both of these seasons occurred between the study’s baseline and end-line surveys while OQP was running. At baseline, around 30 percent of households rate their neighborhood as clean, and 20 percent were victims of flooding in the previous year. There are no statistically significant differences between the mean of these variables across treatment and control households at baseline. Households were also asked, at Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 baseline, whether any members received training in relation to flood prevention. It turns out that 16 percent of households in the control group and 13 percent of households in the treatment group received training and that this difference is statistically significant. If households in the control group are more informed about flood-prevention measures, then this would likely work against the study’s finding an effect. The analysis nonetheless ensures that all the results are robust to the inclusion of baseline values for training in the empirical specification. Table 3 presents descriptive statistics and balance checks on the characteristics of the CBOs at baseline. The first set of characteristics relates to the flood-prevention and cleaning activities of the CBO. A number of CBOs (approximately 70 percent) undertook some action in the fight against flooding in the 12 months prior to the baseline survey. Many were also engaged in activities to counteract flooding, including raising awareness and providing financial assistance for victims of flooding. Around 19 percent of CBOs have as their objective raising awareness on flood-related issues or organizing cleaning events.9 The study achieves balance on almost all of these measures with the exception of whether the CBO engaged in cleaning activities, where it is found that control CBOs were more likely to engage. The differ- ence is, however, only marginally statistically significant. This difference would likely work against finding an effect of the treatment. All of the results presented in later sections are robust to the inclusion of these baseline CBO controls. Table 3 also includes baseline statistics for other characteristics of the CBOs. The average number of members is around 177 with around 39 members present at the previous meeting. On average, around 94 of their members engage in the activities of the CBO for an average of 12 hours per week. This suggests that individuals in the study area are indeed very actively engaged with CBOs. There are some differences across the treatment and control groups in these characteristics at base- line. In particular, CBOs in the control group report more active engagement of their members. CBOs in treatment neighborhoods are less likely to vote in their leaders by election but are more likely to be the head office of the CBO in cases where the CBO is part of a network. They are also more likely to be a flood-prevention committee. All of the results presented in subsequent sections are robust to the inclusion of baseline controls. The final set of CBO characteristics that are considered relates to the engagement of CBOs with CBOs in other neighborhoods. This is important given the close physical distance between neighborhoods in the sample and the likelihood that the activities of CBOs in the treatment area might affect the activities of CBOs in the control areas, potentially contaminating the experiment. It turns out that while there is some degree of collaboration between CBOs within the neighborhood and with CBOs in neighboring neighborhoods, there is no statistically significant difference between the treatment and the control groups on these characteristics. However, given that 63 percent of CBOs in the treatment group collaborate with outside CBOs, spillovers are possible and are explored further in the empirical analysis. Because the study’s treatment was randomly assigned, a direct comparison of outcomes between the treatment and control groups will provide a causal estimate of the impact of the program on those out- comes. The basic specification that is used is given in equation (1). Yi j1 = α + βT T j + δYi j0 + θ Xi j + φS S j + εi js , (1) 9 If a flood-prevention CBO was present in a community, then this was automatically chosen as the focal CBO. This applied to 3 percent of the CBOs in the sample. The World Bank Economic Review 833 Table 3. Baseline CBO Characteristics Mean total Mean control Mean treat p-value CBO flood-prevention and cleaning activities CBO took action in relation to flooding in last year 0.69 0.72 0.66 0.42 Flood activities: Raising awareness 0.30 0.30 0.30 1.00 Flood activities: Financial assistance for victims 0.24 0.28 0.20 0.27 Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Flood activities: Cleaning of canals and lakes 0.11 0.11 0.11 1.00 Flood activities: Surveillance of canals and lakes 0.04 0.04 0.05 0.70 Flood activities: Cooperation other stakeholders 0.26 0.25 0.28 0.72 Flood activities: Small works to avoid flooding 0.14 0.15 0.14 0.82 Objective: Raise awareness on flood-related issues/organize 0.19 0.19 0.19 1.00 cleaning activities CBO engaged in flood-communication campaign 0.22 0.24 0.20 0.57 CBO engaged in cleaning activities 0.57 0.64 0.50 0.08 Number of days spent on flood-communication or cleaning 12.31 10.32 14.30 0.38 activities Number of members engaged in flood-communication or 49.07 59.38 38.76 0.17 cleaning activities CBO Characteristics Number of members 177.62 200.01 155.22 0.54 Number of members present at the last meeting 39.08 38.48 39.67 0.83 Number of members engaged in activities of CBO 94.46 117.55 71.36 0.07 Average hours per week members engage in CBO activities 12.24 11.93 12.56 0.83 Proportion of female members 0.60 0.59 0.62 0.63 Proportion of youth members 0.05 0.06 0.04 0.46 CBO represents a head office 0.90 0.85 0.95 0.04 CBO votes in leaders by election 0.61 0.69 0.53 0.04 Most important benefit of membership of CBO are benefits for 0.11 0.14 0.07 0.20 the community CBO type: Sports and culture association 0.13 0.16 0.09 0.15 CBO type: Economic interest group 0.13 0.10 0.16 0.24 CBO type: Women’s involvement group 0.18 0.20 0.15 0.41 CBO type: Development association 0.46 0.49 0.44 0.53 CBO type: Flood prevention committee 0.03 0.00 0.06 0.02 CBO type: Other association 0.08 0.05 0.10 0.23 Collaboration between different CBOs CBO only intervenes in this neighborhood 0.65 0.65 0.65 1.00 CBO collaborates with other CBOs in the neighborhood 0.68 0.65 0.71 0.40 CBO collaborates with other CBOs in other neighborhoods 0.66 0.69 0.63 0.41 Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: Number of observations is 160 representing the number of CBOs in the sample; p-value based on a test of the statistical significance of the difference in the mean of the control group and mean of the treatment group at baseline. Yi j1 represents the outcome of interest for household i, in neighborhood j, at endline. T j is a dummy variable indicating treatment at the neighborhood level. Yi j0 is a measure of the outcome variable of interest at baseline, which is included when available. Xi j is a set of baseline control variables. S j are stratification dummy variables. 5. Results The analysis first investigates the impact of the OQP intervention on cleanliness, flooding, and CBO efforts to mobilize the community. Second, the study explores heterogeneity in the impact of the intervention 834 Newman et al. Table 4. Impact of OQP on Perceptions of Cleanliness and Flood-Related Outcomes (1) (2) (3) (4) (5) (6) Cleanliness Flood victim this year Flood victim last year Treatment 0.075∗ 0.123∗∗∗ −0.014 −0.021∗∗ −0.015 −0.016 (0.038) (0.023) (0.019) (0.010) (0.025) (0.015) {0.063} {0.001} {0.330} {0.054} {0.356} {0.220} Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Baseline No Yes No Yes No Yes outcome Baseline controls No Yes No Yes No Yes Group FE No Yes No Yes No Yes Enumerator FE No Yes No Yes No Yes Mean control 0.359 0.095 0.194 Observations 2, 285 2, 271 2, 285 2, 271 2, 285 2, 271 R-squared 0.006 0.226 0.001 0.123 0.000 0.165 Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: Robust standard errors clustered at the neighborhood level in parentheses. Anderson’s (2008) sharpened False Discovery Rate q-values for all outcomes included in the main analysis are presented in braces. Baseline mean control is the average of the outcome at baseline for the control group. ∗∗∗ p < 0.01, ∗ ∗ p < 0.05, ∗ p < 0.1 based on robust clustered standard errors. across the size distribution of CBOs and the geographic characteristics of the neighborhood (i.e., whether they are in Phase 1 or Phase 2 of the infrastructure development). The analysis also considers secondary outcomes and examines whether there is evidence of spillovers of the OQP into bordering neighborhoods. The section concludes with a discussion of alternative mechanisms. For each outcome of interest, two different specifications are used: (1) a simple bivariate regression and (2) a regression including the baseline value of the outcome variable (where available), group (random- ization strata) fixed effects, baseline household controls (reported level of cleanliness of the community and whether or not any household member is a member of the CBO), and baseline CBO characteristics (size of the CBO, the number of hours a week members on average engage in CBO activities, type of CBO, an indicator for whether an objective of the CBO is to reduce flooding in the community, and a dummy indicator for whether the CBO engaged in flood cleaning activities at baseline). The latter is the preferred specification.10 Impact of OQP on Cleanliness, Flooding, and CBO Effort The first set of results examines the impact of treatment on cleanliness.11 The results are presented in columns 1 and 2 of table 4. Once all controls are included, it is found that households in treatment areas are 12.3 percentage points more likely to give their neighborhood one of the higher cleanliness ratings than households in control areas and this difference is statistically significant (column 2).12 This corresponds to a 34 percent increase relative to the control sample mean. 10 All of the results, with the exception of being a flood victim this year, are robust to using Lin’s (2013) approach to ensuring the asymptotic precision of the estimator when OLS is used in randomized experiments. All of the results hold when using a double LASSO approach to variable selection. The results are provided in table S1.4 of the supplementary online appendix. 11 This measure is self-reported and so experimenter demand effects are potentially a concern. It is probable that they may not be a major concern for two reasons. First, the treatment took place at the level of the CBO and not the household; half of the households in the treatment group had not heard of the OQP at end line. Second, the analysis finds a positive and statistically significant correlation of 0.41 between the household reported level of cleanliness of the neighborhood and the community-level score. 12 The results are robust to using an ordered logit model with the full (1–5) cleanliness rating provided by the respondents as the dependent variable. The World Bank Economic Review 835 Table 5. Impact of OQP on CBO Effort (1) (2) (3) (4) (5) (6) CBO listed cleaning events among Household received training on top three main activities Awareness of OQP flood prevention Treatment 0.188∗∗ 0.269∗∗∗ 0.168∗∗∗ 0.199∗∗∗ 0.041∗∗ 0.053∗∗∗ (0.078) (0.095) (0.040) (0.028) (0.019) (0.014) Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 {0.031} {0.014} {0.001} {0.001} {0.048} {0.048} Baseline outcome No Yes No No No Yes Baseline controls No Yes No Yes No Yes Group FE No No No Yes No Yes Enumerator FE No No No Yes No Yes Mean control 0.362 - 0.100 Observations 160 160 2, 400 2, 271 2, 285 2, 271 R-squared 0.035 0.564 0.028 0.338 0.004 0.182 Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: In columns (1) and (2), robust standard errors in parentheses. In columns (3) to (6), robust standard errors clustered at the neighborhood level in parentheses. Anderson’s (2008) sharpened False Discovery Rate q-values for all outcomes included in the main analysis are presented in braces. Baseline mean control is the average of the outcome at baseline for the control group where available. ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1 based on robust clustered standard errors. The analysis hypothesizes that increased cleanliness through the OQP should lead to a reduction in flooding in either the most recent (this year) or the previous rainy season (last year).13 The results are presented in columns 3 and 4 and columns 5 and 6 of table 4, respectively. The coefficients on the treatment indicator are negative in all specifications but are only statistically significant in the case of flooding last year. Column 4 shows that households in treatment neighborhoods are 2 percentage points less likely to have experienced flooding than those in control neighborhoods corresponding to a 22 percent decrease relative to the control sample mean. Next, the study considers the impact the treatment had on whether the CBO carried out cleaning events. The results are presented in table 5. In the preferred specification (column 2), CBOs in treatment neighborhoods are 26.9 percentage points more likely to include cleaning events in their activities (a 74 percent increase relative to the control sample mean). This provides evidence that CBOs in treatment neighborhoods were more likely to engage in keeping the neighborhood clean.14 Next, the study investigates whether households in OQP-treated areas were more likely to have heard of OQP. This provides a measure of whether treatment CBOs were visibly active within the community when executing OQP. Moreover, given that the neighborhoods were selected via public lottery, and so it is likely that some households in control areas heard of the program, examining the impact on OQP- awareness serves as a check on the extent of contamination of the control group and the potential for spillovers. The results are presented in columns 3 and 4 of table 5. It is found that households in treatment neighborhoods were almost 20 percentage points more likely to have heard of OQP than households in control neighborhoods (column 4). This suggests that CBOs did engage households in treatment neigh- borhoods. 13 At baseline, it is found that households that report having cleaner neighborhoods are 5 percent less likely to have experienced flooding in the previous winter season. The results are presented in table S1.3 of the supplementary online appendix. While this evidence is not causal, it does provide assurance of a positive relationship between increased cleanliness and decreased flooding. 14 Given that some CBOs were Flood Prevention Committees at baseline, and there were more of these in the treatment group, the analysis also re-estimates all specifications excluding these CBOs and the results hold. The study also con- sidered the extent to which the engagement of CBOs in cleaning activities crowded out other CBO activities but does not find any evidence for such an effect. It should be noted that the analysis is not powered to detect small effects at the CBO level. 836 Newman et al. Finally, the analysis examines the impact of OQP on whether households report that they received interventions related to cleanliness and flood reduction, which would provide further evidence that CBOs did in fact engage with households in the community. The results are presented in columns 5 and 6 of table 5. It is found that treated households are over 5 percentage points more likely to have received training than control households once all relevant control variables have been included. This is a large effect relative to a baseline mean for the control group of 10 percent. Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 It is possible that the results are being driven by CBO members only. To rule this out, the study ex- amines whether the impact of the treatment is different in households with CBO members (52 percent of the sample) by including an interaction term between the treatment and whether or not there are CBO members in the household at baseline. It is important to note that this measure is membership of any CBO, not just the focal CBO that was the target of the intervention. Table S1.5 of the supplementary online appendix presents the results for each of the household-level outcomes considered in tables 4 and 5. The analysis does not find any evidence that households with CBO members are impacted differently by the treatment. This suggests that nonmembers were affected in the same way as CBO members. Cleanliness and experience of flooding are public goods and, therefore it would not be expected that the impact of OQP on these variables would be different for CBO or non-CBO members. It is reassuring that this is what is found, as this suggests that the results are not being driven by biased reporting by CBO members. In addition, the fact that there is no difference between CBO and non-CBO members in the impact of the treatment on OQP awareness or having received training on flood prevention suggests that non-CBO members are being engaged by the treatment. Heterogeneous Effects It is possible that the effects of the intervention will be heterogeneous across different types of CBOs. Of particular importance is the size of the CBO. Larger CBOs may be more likely to reach the cleanliness threshold for receipt of the reward by themselves and so will have less of an incentive to engage community members. It is therefore expected that smaller CBOs will be more likely to be active in mobilizing the community and so the impact of the intervention on awareness of community and cleaning events will be smaller for larger CBOs. The study explores this empirically by introducing heterogeneity along the size distribution of CBOs. The analysis includes the number of members of the CBO at baseline (scaled by 100 for easier interpretation of coefficients) and its interaction with the treatment indicator. The study also trims the sample to exclude six neighborhoods with very large CBOs with over 500 members (three treatment neighborhoods and three control neighborhoods). The results are presented in table 6. Column 1 shows that the level of awareness of OQP among households in treatment neighborhoods is decreasing in the size of the focal CBO. This suggests that larger CBOs may be less dependent on nonmembers for achieving the cleanliness thresholds and may be less likely to engage with them. There is no clear prediction about the relationship between the size of the CBO and the impact of the intervention on cleanliness or flooding. While larger CBOs could achieve improved cleanliness by themselves, smaller CBOs may exert more effort to mobilize the community. Indeed, columns 2 and 3 do not display any differential effect of the size of the CBO on the perceived cleanliness of the neighborhood or the probability of experiencing flooding. This suggests that larger CBOs may be more likely to reach the cleanliness threshold for receipt of the nonmonetary reward and reduced flooding through their own actions rather than through engaging with other members of the community. The study also explores geographical heterogeneity, specifically considering differential effects of the intervention in Phase 1 and Phase 2 areas of the larger infrastructure project.15 Given that the social recognition reward for improving cleanliness should be the same in both areas, the analysis should not 15 Balance tests for treatment and control groups within each phase are presented in tables S1.6 to S1.9 of the supple- mentary online appendix. Reasonably good balance is achieved on most variables with some exceptions. In Phase 2, the study finds that households in the treatment group are more likely to have been flood victims in the previous year and The World Bank Economic Review 837 Table 6. Heterogeneity by Size of CBO (1) (2) (3) Awareness of OQP Cleanliness Flood victim this year Treatment 0.299∗∗∗ 0.092∗∗ −0.017 (0.051) (0.039) (0.020) {0.001} {0.035} {0.279} Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Treatment × CBO Size −0.107∗∗ 0.030 −0.004 (0.051) (0.038) (0.021) {0.051} {0.297} {0.511} CBO Size 0.071∗∗ −0.048∗ −0.021 (0.033) (0.028) (0.015) {0.048} {0.093} {0.144} Baseline outcome No Yes Yes Observations 2, 126 2, 126 2, 126 R-squared 0.349 0.230 0.130 Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: All specifications include baseline controls, group fixed effects, and enumerator fixed effects. Robust standard errors clustered at the neighborhood level in parentheses. Anderson’s (2008) sharpened False Discovery Rate q-values for all outcomes included in the main analysis are presented in braces. Baseline mean control is the average of the outcome at baseline for the control group where available. ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1 based on robust clustered standard errors. see a difference in cleanliness across the two phases. However, Phase 1 areas are both more prone to flooding and receive the flood-prevention infrastructure earlier. As such one would expect the intervention to be less effective at reducing flooding in Phase 1 areas. To explore this empirically the study includes a dummy indicator for whether the community is in Phase 1 or Phase 2 and its interaction with the treatment indicator. The results for cleanliness and flooding outcomes, with and without the interaction term, are presented in panel A of table 7. As predicted, the analysis finds no differential impact of the treatment across Phase 1 and Phase 2 neighborhoods on the cleanliness of the community (column 2), but it does find evidence to suggest that the effect of the treatment on reducing flooding is larger in Phase 2 neighborhoods. The probability of experiencing flooding in the most recent year declines by 3.9 percentage points in Phase 2 (column 4); given that the average probability of experiencing flooding in control households in Phase 2 areas in this time period is 8.5 percent, the magnitude of this effect is meaningful. The coefficient on the interaction term is positive for all flooding outcomes and is statistically significant in the case of flooding in the most recent year. This suggests that, as predicted, OQP may not have been as effective at reducing flooding in Phase 1 areas. It is worth noting that while there are some differences between the treatment groups in Phase 1 and Phase 2 (see tables S1.10 and S1.11 of the supplementary online appendix), the relevant counterfactual in each phase is the control group where there is balance within phases. The baseline differences across phases are however potentially important in explaining why the intervention was more effective in Phase 2 than in Phase 1. Treatment households in Phase 2 were more likely to report that their neighborhood was clean at baseline, and Phase 1 CBOs were more engaged in flood-prevention activities. This is consistent with the higher flood risk in Phase 1 areas before the infrastructure works. This lends support to the interpretation of the fact that the OQP is more effective in Phase 2 as the program plays a role in making flood prevention more salient and facilitating community engagement in addressing flood-related issues. are less likely to have received training about flood prevention. Both differences would work against finding an effect. Controls at baseline are included in the empirical specification to correct for these baseline differences. 838 Newman et al. Table 7. Heterogeneity by Infrastructure Phase and Flood Risk (1) (2) (3) (4) (5) (6) Panel A Cleanliness Flood victim this year Flood victim last year Treatment 0.123∗∗∗ 0.120∗∗∗ −0.021∗∗ −0.039∗∗∗ −0.016 −0.021 (0.023) (0.033) (0.010) (0.013) (0.015) (0.018) {0.001} {0.005} {0.054} {0.014} {0.220} {0.213} Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Phase 1 −0.101 −0.102 0.175∗∗∗ 0.174∗∗∗ 0.368∗∗∗ 0.368∗∗∗ (0.063) (0.063) (0.047) (0.047) (0.066) (0.065) {0.105} {0.105} {0.005} {0.005} {0.001} {0.001} Phase 1 × Treatment 0.008 0.045∗∗ 0.012 (0.049) (0.018) (0.031) {0.511} {0.031} {0.436} Observations 2, 271 2, 271 2, 271 2, 271 2, 271 2, 271 R-squared 0.226 0.226 0.123 0.124 0.165 0.165 Panel B Treatment 0.124∗∗∗ 0.141∗∗∗ −0.022∗∗ −0.002 −0.017 −0.008 (0.023) (0.030) (0.010) (0.015) (0.015) (0.021) {0.001} {0.001} {0.048} {0.519} {0.198} {0.436} Flood risk −0.017 0.004 0.054∗∗∗ 0.078∗∗∗ 0.131∗∗∗ 0.144∗∗∗ (0.016) (0.033) (0.014) (0.018) (0.037) (0.045) {0.223} {0.519} {0.001} {0.001} {0.005} {0.008} Flood risk × Treatment −0.041 −0.049∗ −0.024 (0.049) (0.028) (0.038) {0.284} {0.093} {0.356} Observations 2, 271 2, 271 2, 271 2, 271 2, 271 2, 271 R-squared 0.226 0.227 0.128 0.130 0.169 0.169 Panel C: Phase 1 Treatment 0.109∗∗∗ 0.105∗∗∗ −0.009 −0.001 −0.010 −0.042 (0.016) (0.034) (0.016) (0.026) (0.021) (0.031) {0.001} {0.015} {0.385} {0.552} {0.396} {0.170} Flood risk 0.009 0.004 0.082∗∗∗ 0.091∗∗∗ 0.149∗∗ 0.115∗ (0.027) (0.048) (0.027) (0.029) (0.056) (0.064) {0.438} {0.521} {0.018} {0.015} {0.031} {0.093} Flood risk × Treatment 0.010 −0.019 0.077 (0.071) (0.051) (0.065) {0.514} {0.436} {0.205} Observations 922 922 922 922 922 922 R-squared 0.304 0.304 0.149 0.150 0.183 0.185 Panel D: Phase 2 Treatment 0.106∗∗∗ 0.128∗∗∗ −0.037∗∗ −0.005 −0.027 0.013 (0.037) (0.039) (0.015) (0.020) (0.019) (0.027) {0.015} {0.008} {0.031} {0.463} {0.147} {0.396} Flood risk −0.040 −0.012 0.035∗∗ 0.073∗∗∗ 0.144∗∗∗ 0.201∗∗∗ (0.024) (0.046) (0.017) (0.023) (0.047) (0.054) {0.098} {0.463} {0.053} {0.008} {0.012} {0.005} Flood risk × Treatment −0.053 −0.078∗∗ −0.099∗∗ (0.062) (0.031) (0.040) {0.279} {0.031} {0.031} Observations 1, 349 1, 349 1, 349 1, 349 1, 349 1, 349 R-squared 0.236 0.237 0.118 0.124 0.149 0.236 Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: Panels A and B report the results for the full sample while panels C and D report the results for the Phase 1 and Phase 2 subsamples, respectively. Robust standard errors clustered at the neighborhood level in parentheses. Anderson’s (2008) sharpened False Discovery Rate q-values for all outcomes included in the main analysis are presented in braces. ∗∗∗ p < 0.01, ∗ ∗ p < 0.05, ∗ p < 0.1 based on robust clustered standard errors. The World Bank Economic Review 839 To probe further how flood risk is related to the effectiveness of the intervention, the article also con- siders the extent to which there is heterogeneity in the impact of the intervention depending on previous experience of flooding at baseline.16 The results are presented in panel B of table 7 for the full sample and panels C and D for Phases 1 and 2, respectively. The analysis finds no effect of baseline flood risk on the impact of OQP on the cleanliness of the neighborhood, which suggests that the level of engagement of the community with the intervention did not depend on risk exposure.17 It is found, however, that the Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 impact on the probability of being a flood victim is increasing in exposure to flood risk. This suggests that the program is most effective where flooding is more likely. When comparing this result across Phase 1 and Phase 2, it is found that the impact of the program on reducing flooding is greater in the areas more exposed to flood risk in Phase 2 only. This suggests that the availability of infrastructure in Phase 1 lessens the impact of the intervention in these communities.18 Secondary Outcomes In addition to the primary outcomes of interest, the study also considers whether there is any evidence for positive impacts on health, education, and work-related outcomes for individuals living in OQP neigh- borhoods.19 Given the short time frame of the experiment, the analysis does not expect there to be very large impacts on such secondary outcomes, but it is possible that they are affected through both improved cleanliness and reduced flooding.20 Table 8 shows that individuals in treated neighborhoods are around 2 percentage points less likely to report that they have been sick in the previous 30 days (column 1), are less likely to have been ill in the most recent rainy season (column 2), and, in particular, are less likely to have been ill due to malaria in the most recent rainy season (column 3). Given that 24.2 percent of households in the control group reported that they were sick in the previous 30 days, the magnitude of the effect is relatively large. This suggests that the OQP, through its impact on cleanliness and reduced flooding, also impacts the health of individuals living in the neighborhood. The study also explores the extent to which there are effects on the number of missed days at work and school and the number of days that schools were closed (the rainy season can last until September when schools open, and so flooding could cause schools to remain closed and impact school days). While the coefficients are not statistically significant the signs in all cases are negative (columns 4 to 7). 16 The past realization of flooding is used as a proxy for flood risk. It is possible that households are exposed to flood risk without ever having experienced flooding. This should be borne in mind when interpreting the results. 17 The analysis also finds no differential effect across phase or flood risk on awareness of OQP or training. These results are presented in table S1.12 of the supplementary online appendix. This suggests that the fact that the intervention did not have as great an impact in Phase 1 is not related to the level or extent of engagement of the CBOs in Phase 1. Indeed, the take-up rate was 100 percent in both areas (all CBOs signed the social contract and engaged with the intervention). Moreover, at midline 94 percent of communities in Phase 1 received the prize and only 50 percent in Phase 2. At endline 71 percent of communities in Phase 1 received the prize and 59 percent of communities in Phase 2. 18 It is also possible that the wealth of an area impacts the extent to which the intervention is effective. To explore this possibility, a household-level wealth index is constructed using information on the type of housing and ownership of assets. The analysis considers the extent to which there is heterogeneity in the impact of treatment across different quartiles of this index but finds no effect for the whole sample or for the Phase 1 and Phase 2 samples. The study also considers whether there is an interactive effect between flood risk and wealth, but no clear pattern emerges. This suggests that the results are not being driven by wealth or its interaction with flood risk. The results for these specifications are presented in table S1.13 of the supplementary online appendix. 19 Baseline balance tests for whether the household was ill in the last 30 days and the number of workdays missed are presented in table S1.14 of the supplementary online appendix. There is some baseline imbalance, and so in the preferred specification the study includes the baseline value of the outcome variable where available and the baseline level of reported illness in the previous 30 days as an additional control. 20 At baseline the study finds a statistically significant correlation between flooding and illness in the previous 30 days with households that experienced flooding 5 percent more likely to report having been sick. There is also a small positive correlation between cleanliness and health, but this is not statistically significant. 840 Newman et al. Table 8. Impact of OQP on Individual-Level Health, Work, and Education Outcomes (1) (2) (3) (4) (5) (6) (7) Ill in the last 30 Ill in recent Ill due to malaria in Workdays School days School year days rainy season recent rainy season Working missed missed began on time Treatment −0.020∗∗∗ −0.025∗∗∗ −0.013∗∗ −0.004 −0.436 −0.013 −0.003 (0.007) (0.008) (0.005) (0.005) (0.284) (0.082) (0.005) Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 {0.010} {0.010} {0.037} {0.330} {0.116} {0.511} {0.389} Baseline Yes No No Yes Yes Yes No outcome Mean control 0.242 - - 0.262 2.604 1.932 - Observations 20, 389 20, 399 20, 399 23, 083 3, 039 4, 188 5, 990 R-squared 0.119 0.121 0.041 0.283 0.148 0.091 0.054 Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: All specifications include baseline controls, group fixed effects and enumerator fixed effects. The baseline level of reported illness in the previous 30 days is included in columns (2) and (3) as an additional control. Robust standard errors clustered at the neighborhood level in parentheses. Anderson’s (2008) sharpened False Discovery Rate q-values for all outcomes included in the main analysis are presented in braces. ∗ ∗ ∗ p < 0.01, ∗ ∗ p < 0.05, ∗ p < 0.1 based on robust clustered standard errors. Spillovers The densely populated urban setting coupled with the selection into treatment by public lottery makes spillovers likely. It is possible that households in untreated neighborhoods are also informed about OQP and change their behavior in response. On the one hand, they could observe the behavior of households in treated neighborhoods and copy this behavior in their own neighborhood. This would work against finding an effect of the treatment. On the other hand, if households in untreated neighborhoods know that CBOs in treated neighborhoods are being incentivized to engage in cleaning behavior (albeit through nonmonetary incentives), they may be discouraged from engaging in any cleaning activities. This would lead to an over-estimation of the impact of OQP on behavior. It is also possible that the proximity of neighborhoods means that poorly maintained public infrastructure in one neighborhood leads to flooding in a neighboring neighborhood, even if they perform well at keeping their community clean. If the latter are treated, then this will lead to a downward bias on the estimate of the impact of OQP on flooding. It might also impact on the behavioral response in treated neighborhoods if they feel that their efforts to keep the neighborhood clean are pointless. This article explores the extent to which there are spillovers of this kind by taking into account the treatment status of bordering neighborhoods.21 It first considers whether the level of awareness of OQP and the behavior of households in control neighborhoods is impacted by whether they share a border with a treated neighborhood. The regression specified in equation (2) is estimated for the control group for the primary household-level outcomes: Yi j = α + βT border_T j + βB nr_border j + δYi j0 + θ Xi j + φS S j + εi j , (2) where Yi j , Yi j0 , Xi j , S j and εi j are as for equation (1), border_T j is a dummy indicator for whether neighborhood j shares a border with a treated neighborhood and nr_border j is a control for the total number of bordering neighborhoods. Around 75 percent of control neighborhoods share a border with a treatment neighborhood. Where relevant to the outcome of interest the analysis also includes the same set of baseline control variables as included in the main regressions. The results are presented in columns 1 to 3 of table 9. 21 No differences of note are found in the extent of spillovers in Phase 1 compared with Phase 2, despite the fact that CBOs in Phase 1 report collaborating more with CBOs in other neighborhoods (see tables S1.6 and S1.7 of the supplementary online appendix). The World Bank Economic Review 841 Table 9. Spillovers (1) (2) (3) (4) (5) Cleanliness of Cleanliness of Awareness of OQP neighborhood Flood victim this year neighborhood Flood victim this year Treat border 0.067 0.053 0.038 (0.048) (0.075) (0.026) Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Control border 0.306∗∗∗ 0.032 (0.058) (0.026) Nr border −0.003 0.009 −0.006 −0.010 −0.004 (0.007) (0.011) (0.004) (0.007) (0.003) Constant −0.733∗∗∗ 1.164∗∗∗ −0.114 0.492∗∗∗ 0.242∗∗∗ (0.172) (0.407) (0.081) (0.142) (0.070) Baseline outcome No Yes Yes Yes Yes Observations 1, 060 1, 060 1, 060 1, 098 1, 098 R-squared 0.393 0.278 0.176 0.295 0.155 Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: All specifications include baseline controls, group fixed effects, and enumerator fixed effects. Robust standard errors clustered at the neighborhood level in parentheses. ∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1. Bordering a treated neighborhood has no impact on the level of awareness of OQP (column 1). In addition, no evidence is found that the cleanliness perceptions of households in control neighborhoods that share a border with treated neighborhoods are different to those that do not (column 2), nor is there any evidence to suggest that they are more likely to be flood victims (column 3). It is also possible that treated neighborhoods experience flooding as a result of poorly maintained public infrastructure in neighboring control neighborhoods. This could have two effects: (1) it could demotivate households in treated neighborhoods from keeping their neighborhood clean; and (2) it could lead to a downward bias in the estimate of OQP on flooding. To explore this, the analysis estimates the specification given in equation (3) for the treatment group for the primary household-level outcomes: Yi j = α + βT border_C j + βB nr_border j + δYi j0 + θ Xi j + φS S j + εi j , (3) where border_C j is a dummy indicator for whether neighborhood j shares a border with a control neigh- borhood and all other variables are as described for equation (2). The results are also presented in table 9. There is some evidence in column 4 to suggest that households in treatment neighborhoods that share a border with a control neighborhood are more likely to report their area as clean. This suggests that households are not demotivated by being in close proximity to control neighborhoods, although it may be that they perceive their neighborhoods as being relatively cleaner. Flood-related outcomes in treated neighborhoods, however, are not affected by bordering control neighborhoods (column 5) suggesting that there are no spillover effects from flooding in either direction. Alternative Mechanisms The most likely explanation for the results is that the provision of social rewards to CBO members en- couraged them to increase their efforts to keep the neighborhood clean and also to engage non-CBO members in this activity. There are, however, alternative mechanisms that could also explain the results that should be considered. First, the study has not explored the possible role of social preferences or social norms in this context. Although social norms can play a very important role in facilitating contributions to public goods, it is likely that the time frame for the experiment is too short to significantly change social norms. However, 842 Newman et al. any change in social norms or the existence of reciprocal preferences of community members would reinforce the effect of increased contributions by CBO members. Second, it is possible that community members do not fully understand the link between cleanliness and flooding or underestimate the benefit that they would gain from reduced flooding. If they also believe that CBOs have access to better information about the value of the public good, then it is possible that the contribution of CBO members could convey some information to non-CBO members about this value and Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 this in turn could influence their decision to contribute. It is unlikely that this is the primary mechanism at work given that as part of the larger PROGEP project an information campaign was carried out in both control and treatment areas communicating the importance of keeping the neighborhood clean in order to reduce flooding. It is likely that any additional information effect of the OQP treatment would be marginal. Finally, it is possible that community members, on seeing the involvement of NGOs and officials, and receiving some initial support, simply believed that increasing their efforts to improve cleanliness was what they were supposed to do. While it is possible that this influenced the behavior of CBO members and was part of their motivation to act, it is not likely it would have had a significant effect on non-CBO members, as they were not directly targeted by the treatment. While it is not possible to completely rule out these different possible underlying mechanisms, the key policy-relevant result from the experiment remains the same: targeting an intervention at some key groups within the community can motivate them to make increased contributions to the public good to improve the functioning of drainage infrastructure, thereby reducing the effects of flooding for the community as a whole. 6. Conclusions Climate-resilient infrastructure can play a key role in climate-change adaptation and mitigation (Hallegatte, Rentschler, and Rozenberg 2019). The need is especially acute in poor communities in low- income countries that have historically contributed least to climate change but that are disproportionately at risk (IPCC 2023). The returns to infrastructure investments depend critically on operations, mainte- nance, and the interaction of users with the infrastructure. In poor communities lacking basic services and enforcement mechanisms, infrastructure is especially vulnerable to a public-goods problem. This reduces the welfare gains from infrastructure, decreases the economic and social rates of return, and discourages investment. Understanding how individuals in large, urban communities can coordinate to overcome public-goods problems is thus extremely important. Using an RCT, this paper tests the effectiveness of an intervention targeting a key subset of the community to motivate them to keep their local area clean in order to increase the effectiveness of drainage infrastructure and therefore reduce the risk of flooding in that area. This was a bottom-up intervention: OQP engaged existing community-based organizations and empowered them to use their local knowledge and networks to work toward improved community cleanliness, while providing minimal guidance, basic materials, and relatively inexpensive incentives. It is found that the program had a significant positive effect on households’ perception of the cleanliness of their neighborhood and reduced their vulnerability to flooding. The results demonstrate that an intervention targeted at some key groups within the community can motivate them to make increased contributions to the public good, thereby improving outcomes for the community as a whole. Evidence was also found that OQP, through its impact on cleanliness and reduced flooding, had positive impacts on the health of individuals living in the treated neighborhoods, suggesting the potential for further welfare-enhancing benefits from this type of intervention. This intervention took place in the context of a larger infrastructure program that was being rolled out in this area in two phases. In addition to the overall impact of OQP, the study also finds differences in the The World Bank Economic Review 843 results for Phase 1 and Phase 2 areas: Phase 1 areas had received upgraded infrastructure by the start of the OQP intervention, whereas the Phase 2 areas had not. The study found that, while the improved level of cleanliness was no different across phases, the intervention was more effective in reducing flooding in Phase 2 areas. Heterogeneity analysis along the extent of exposure to flood risk at baseline reveals that the impact of the program on reducing flooding in Phase 2 is increasing in the extent of flood risk. This suggests that the availability of infrastructure in Phase 1 lessens the impact of the intervention in these Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 communities. This type of intervention could therefore act as a temporary substitute or stop-gap measure for bulky investment in infrastructure. It is important to note that this study only evaluated the impact of this program after a 12-month period. It would be interesting to see if this CBO activity can be sustained over a longer time period and, if so, what impact this has on the sustainability of these infrastructure investments in these areas. The results also suggest that the impact of the intervention differed depending on the size of the CBO that was engaged. Larger CBOs, who were able to reach the cleaning threshold by themselves, had less of an incentive to mobilize the community as a whole. While in this case, this had no differential impact on flooding, it is something important to bear in mind for the success of this type of intervention in other settings. The realized returns to infrastructure investments are often low, not because the infrastructure lacks inherent value but because of insufficient attention to complementary factors (World Bank 2022). This study demonstrates that incentivizing existing local groups (CBOs) provides one channel to increase the returns to flood-prevention infrastructure. As highlighted by Mansuri and Rao (2004), one of the argu- ments for community involvement in development projects is that they have local information. OQP was designed to leverage existing local knowledge by giving CBOs near full autonomy to focus on achieving the end goal of community cleanliness, rather than on performing prescribed tasks. This design has a rela- tively low cost and has potential application to a range of public investments in poor urban communities. In particular, it is a model that can be used to better engage local communities in building their resilience against flooding and other climate-related shocks. Data Availability Statement The baseline data underlying this article are available in the World Bank Microdata Catalog, at https: //doi.org/10.48529/dkh4-wy42. The endline data will also be made available shortly after publication. Funding This work was supported by the World Bank-administered Impact Evaluation to Development Impact Trust Fund, financed by the UK [grant number TF018690]; World Bank-administered Bank-Netherlands Partnership Program, financed by the Netherlands [grant number TF099415]; and the Nordic Develop- ment Fund [grant number C33]. References Abramovsky, L., B. Augsburg, M. Lührmann, F. Oteiza, and J.P. Rud. 2023. “Community Matters: Heterogeneous Impacts of a Sanitation Intervention.” World Development 165: 106197. Adams, E.J., N. Cavill, and L.B. Sherar. 2017. “Evaluation of the Implementation of an Intervention to Improve the Street Environment and Promote Walking for Transport in Deprived Neighbourhoods.” BMC Public Health [Electronic Resource] 17: 655. Aflagah, K., T. Bernard, and A. Viceisza. 2022. “Cheap Talk and Coordination in the Lab and in the Field: Collective Commercialization in Senegal.” Journal of Development Economics 154: 102751. Allaire, M.C. 2016. “Using Practical and Social Information to Influence Flood Adaptation Behavior.” Water Resources Research 52: 6078–93. 844 Newman et al. Anderson, M. 2008. “Multiple Inference and Gender Differences in the Effects of Early Intervention: A Reevaluation of the Abecedarian, Perry Preschool and Early Training Projects.” Journal of the American Statistical Association 103(484): 1481–95. Ariely, D., A. Bracha, and S. Meier. 2009. “Doing Good or Doing Well? Image Motivation and Monetary Incentives in Behaving Prosocially.” American Economic Review 99(1): 544–55. Arunachalam, N., B.K. Tyagi, M. Samuel, R. Krishnamoorthi, R. Manavalan, S.C. Tewari, V. Ashokkumar, A. Kroeger, J. Sommerfeld, and M. Petzold. 2012. “Community-Based Control of Aedes aegypti by Adoption of Eco-Health Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Methods in Chennai City, India.” Pathogens and Global Health 106(8): 488–96. Ashraf, N., O. Bandiera, and J. Jack. 2014a. “No Margin, No Mission? A Field Experiment on Incentives for Public Service Delivery.” Journal of Public Economics 120: 1–17. Ashraf, N., O. Bandiera, and S. Lee. 2014b. “Awards Unbundled: Evidence from a Natural Field Experiment.” Journal of Economic Behavior and Organization 100: 44–63. Bakhtiar, M.M., R. Guiteras, J.A. Levinsohn, and A.M. Mobarak. 2021. “Social and Financial Incentives for Over- coming a Collective Action Problem.” Working Paper Number 29294. National Bureau of Economic Research. Cambridge, MA, USA. Beath, A., F. Christia, and R. Enikolopov. 2015. “The National Solidarity Programme: Assessing the Effects of Community-driven Development in Afghanistan.” International Peacekeeping 22(4): 302–20. Bottazzi, P., M. Winkler, S. Boillat, A. Diagne, M. Maman Chabi Sika, A. Kpangon, S. Faye, and C. Speranza. 2018. “Measuring Subjective Flood Resilience in Suburban Dakar: A Before-After Evaluation of the “Live with Water” Project.” Sustainability 10:2135. Brandts, J., and D.J. Cooper. 2006. “A Change Would Do You Good: An Experimental Study on How to Overcome Coordination Failure in Organizations.” American Economic Review 96(3): 669–93. Bripi, F., and D. Grieco. 2023. “Participatory Incentives.” Experimental Economics 26(4): 813–849. https://doi.org/ 10.1007/s10683- 023- 09798- 7. Cadsby, C.B., and E. Maynes. 1999. “Voluntary Provision of Threshold Public Goods with Continuous Contributions: Experimental Evidence.” Journal of Public Economics 71: 53–73. Cameron, L., S. Olivia, and M. Shah. 2019. “Scaling up Sanitation: Evidence from an RCT in Indonesia.” Journal of Development Economics 138: 1–16. Cappelen, A., B.A. Reme, E.O. Sorenson, and B. Tungodden. 2016. “Leadership and Incentives.” Management Science 62(7): 1944–53. Carlsson, F., O. Johansson-Stenman, and P. Nam. 2015. “Funding a New Bridge in Rural Vietnam: A Field Experiment on Social Influence and Default Contributions.” Oxford Economic Papers 67(4): 984–1014. Casey, C. 2018. “Radical Decentralization: Does Community-Driven Development Work?” Annual Review of Eco- nomics 10: 139–63. Croson, R., and M. Marks. 2000. “Step Returns in Threshold Public Good: A Meta- and Experimental Analysis.” Experimental Economics 2(3): 239–59. Deserranno, E. 2019. “Financial Incentives as Signals: Experimental Evidence from the Recruitment of Village Pro- moters in Uganda.” American Economic Journal: Applied Economics 11(1): 277–317. Everett, G., and J. Lamond. 2016. “SuDS and Human Behaviour: Co-Developing Solutions to Encourage Sustainable Behaviour.” E3S Web of Conferences 7: 15004. Feltovich, N., and P. Grossman. 2015. “How Does the Effect of Pre-Play Suggestions Vary with Group Size? Experi- mental Evidence from a Threshold Public-Good Game.” European Economic Review 79: 263–80. Frey, B., and F. Oberholzer-Gee. 1997. “The Cost of Price Incentives: An Empirical Analysis of Motivation Crowding- Out.” American Economic Review 87(4): 746–55. GFDRR. 2011. Climate Risk and Adaptation to Climate Change: Country Profile Senegal. Washington: World Bank. Gneezy, U., S. Meier, and P. Rey-Biel. 2011. “When and Why Incentives (Don’t) Work to Modify Behavior.” Journal of Economic Perspectives 25(4): 191–210. Guiteras, R., J. Levinsohn, and A. Mobarak. 2015. “Encouraging Sanitation Investment in the Developing World: A Cluster-Randomized Trial.” Science 348(6237): 903–6. Hallegatte, S., J. Rentschler, and J. Rozenberg. 2019. Lifelines: The Resilient Infrastructure Opportunity: Sustainable Infrastructure. Washington, DC: World Bank. Ibem, E.O. 2009. “Community-Led Infrastructure Provision in Low-Income Urban Communities in Developing Coun- tries: A Study on Ohafia, Nigeria.” Cities 26(3): 125–32. The World Bank Economic Review 845 IPCC. 2023. Summary for Policymakers. In Climate Change 2023: Synthesis Report. Contribution of Working Groups I, II and III to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change, edited by H. Lee, and J. Romero, 1–34 Geneva: IPCC. Jayachandran, S., J. de Laat, E.F. Lambin, C.Y. Stanton, R. Audy, and N.E. Thomas. 2017. “Cash for Carbon: A Randomized Trial of Payments for Ecosystem Services to Reduce Deforestation.” Science 357(6348): 267–73. Kremer, M., E. Miguel, S. Mullainathan, C. Null, and A. Peterson Zwane. 2009. “Making Water Safe: Price, Persuasion, Peers, Promoters, or Product Design.” Working Paper, August 2009, Cambridge, MA, USA: Harvard University. Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Leeffers, S. 2023. “It Will Rain: The Effect of Flood Information on Flood Preparedness in Mozambique.” https: //www.stefanleeffers.com/jobmarketpaper. Lin, W. 2013. “Agnostic Notes on Regression Adjustments to Experimental Data: Reexamining Friedman’s Critique.” Annals of Applied Statistics 7(1): 295–318. Madajewicz, M., A. Tompsett, and M.A. Habib. 2021. “How Does Delegating Decisions to Communities Affect the Provision and Use of Public Services? Evidence from a Field Experiment in Bangladesh.” Journal of Development Economics 150: 102609. Mansuri, G., and V. Rao. 2004. “Community-Based and -Driven Development: A Critical Review.” World Bank Research Observer 19(1): 1–39. McGranahan, G., D. Balk, and B. Anderson. 2007. “The Rising Tide: Assessing the Risks of Climate Change and Human Settlements in Low-Elevation Coastal Zones.” Environment and Urbanization 19(1): 17–37. Mitra, A., M.-C. Buisson, and R. Bastakoti. 2017. “Enhancing Coordination in Water Management through Commu- nication Tools: Results from Experimental Games in Coastal Bangladesh.” International Journal of the Commons 11(2): 774–98 Moser, C., and D. Satterthwaite. 2010. “Toward Pro-Poor Adaptation to Climate Change in the Urban Centers of Low- and Middle-Income Countries.” In Social Dimensions of Climate Change: Equity and Vulnerability in a Warming World, edited by R. Mearns, and A. Norton, Washington, DC: The World Bank. 231-258. Nepal, M., A. Karki Nepal, M.S. Khadayat, R.K. Rai, P. Shyamsundar, and E. Somanathan. 2023. “Low-Cost Strategies to Improve Municipal Solid Waste Management in Developing Countries: Experimental Evidence from Nepal.” Environmental and Resource Economics 84: 729–52. Ostrom, E. 2009. “A Polycentric Approach for Coping with Climate change.” Policy Research Working Paper No. 5095. Washington DC: The World Bank. Pervin, I.A., S.M.M. Rahman, M. Nepal, A.K.E. Haque, H. Karim, and G. Dhakal. 2020. “Adapting to Urban Flooding: A Case of Two Cities in South Asia.” Water Policy 22: 162–88. Quatrocchi, J.P., A. Coville, E. Mvukiyehe, C.J. Dohou, F. Esu, B. Cohen, Y.L. Bokasola, and C. Croke. 2021. “Effects of a Community-Driven Water, Sanitation, and Hygiene Intervention on Water and Sanitation Infrastructure, Access, Behaviour, and Governance: A Cluster-Randomized Controlled Trial in Rural Democratic Republic of Congo.” BMJ Global Health 6: 5. Rentschler, J., M. Salhab, and B.A. Jafino. 2022. “Flood Exposure and Poverty in 188 Countries.” Nature Communi- cations 13: 3527. Rozenberg, J., and M. Fay. 2019. Beyond the Gap: How Countries Can Afford the Infrastructure They Need While Protecting the Planet. Washington, DC: World Bank. Sarabia, M.M., A. Kägi, A.C. Davison, N. Banwell, C. Montes, C. Aebischer, and S. Hostettler. 2020. “The Challenges of Impact Evaluation: Attempting to Measure the Effectiveness of Community-Based Disaster Risk Management.” International Journal of Disaster Risk Reduction 49: 101732. Sheely, R. 2013. “Maintaining Local Public Goods: Evidence from Rural Kenya.” Faculty Research Working Paper Series, RWP13-051. Harvard Kennedy School. Camridge, MA, USA. Turiansky, A. 2021. “Collective Action in Games as in Life: Experimental Evidence from Canal Cleaning in Haiti.” Journal of Development Economics 153: 102722. Winter, A.K., and A. Karvonen. 2022. “Climate Governance at the Fringes: Peri-Urban Flooding Drivers and Re- sponses.” Land Use Policy 117: 106124. World Bank. 2022. Driving Revolutionary Ideas into Practice. Washington, DC: World Bank. Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Supplementary Online Appendix Group Incentives for the Public Good: A Field Experiment on Improving the Urban Environment Carol Newman , Tara Mitchell, Marcus Holmlund, and Chloë Fernandez S1: Additional Figure and Tables Figure S1.1, Table S1.1 to S1.14. Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Figure S1.1. Study Timeline. Source: Author’s compilation. Note: This figure shows the timeline of the Operation Clean Neighborhood (OQP) study. Table S1.1. Initial Endowment Package Materials Amount per neighborhood 1 Boots 15 2 Shovel 15 3 Wheelbarrow 05 4 Pitchfork 10 5 Broom 25 6 Professional 15 gloves 7 Sifter 05 8 Rake 10 9 Machete 05 10 Shears 05 Source: Implementation Manual of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: This table lists the components of the initial endowment package given to treatment Community-Based Orga- nizations. Table S1.2. Assessment Criteria Criteria Rationale Weight 1. Nonobstruction of natural Neighborhoods are often flooded due to 30% drainage structures and waterways waterway obstruction from solid waste. 2. Lack of waste dumping in the Households often dump their waste in public 20% streets spaces. Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 3. Cleanliness of abandoned Many houses are abandoned in the flood zones 20% houses/land. and they are often transformed into dumping grounds. 4. Cleanliness of public spaces The cleanliness of social spaces is an indicator of 10% the level of community awareness of sanitation. 5. Lack of water on streets and Households often discharge domestic wastewater 10% house fronts into the streets. 6. Innovative initiatives taken by the OQP aimed to encourage CBOs to be creative in 10% CBO addressing community cleanliness. Source: Implementation Manual of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: This table describes the criteria used to assess the cleanliness of treated neighborhoods by the evaluators at mid-line and end-line. Table S1.3. Baseline Correlation between Flooding and Cleanliness Outcome: Experienced flooding in the previous rainy season (1) (2) Cleanliness −0.054∗∗ −0.053∗∗∗ (0.020) (0.020) Constant 0.234∗∗∗ −0.237 (0.017) (0.165) Baseline controls No Yes Group fixed effects No Yes Enumerator fixed effects No Yes Observations 2,285 2,271 R-squared 0.004 0.129 Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: Robust standard errors clustered at the neighborhood level in parentheses. ∗∗∗ p < 0.01, ∗ ∗ p < 0.05, ∗ p < 0.1. Table S1.4. Robustness Check: Inclusion of Interactions between Treatment and Covariates (Panel A) and Variable Selection by Double Lasso (Panel B) (1) (2) (2) (4) (5) (6) Household received CBO listed Flood victim Flood victim Awareness of training on flood cleaning events in Panel A Cleanliness this year last year OQP prevention top three activities Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Treatment 0.124∗∗∗ −0.017 −0.013 0.203∗∗∗ 0.054∗∗∗ 0.229∗∗∗ (0.022) (0.010) (0.014) (0.028) (0.014) (0.085) Baseline outcome Yes Yes Yes No Yes Yes Observations 2,271 2,271 2,271 2,271 2,271 160 R-squared 0.230 0.132 0.170 0.354 0.184 0.133 Panel B Treatment 0.125∗∗∗ −0.021∗ −0.016 0.203∗∗∗ 0.052∗∗∗ 0.200∗∗ (0.024) (0.011) (0.013) (0.029) (0.013) (0.078) Baseline outcome Yes Yes Yes No Yes Yes Observations 2,271 2,271 2,271 2,271 2,271 160 Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: All specifications include group fixed effects and enumerator fixed effects. In panel A, interaction terms between the treatment indicator and each of the (de- meaned) covariates are included (see Lin (2013). In panel B, covariates are included in each column selected using double LASSO. In columns (1) to (5), robust standard errors are clustered at the neighborhood level in parentheses. In column (6) robust standard errors are presented in parentheses. ∗∗∗ p < 0.01, ∗ ∗ p < 0.05, ∗ p < 0.1. Table S1.5. Impact of OQP on Main Outcomes: Heterogeneity by CBO Membership (1) (2) (3) (4) (5) Household received Cleanliness of Flood victim Flood victim Household training on flood neighborhood this year last year awareness of OQP prevention Treatment 0.114∗∗∗ −0.019 −0.026 0.200∗∗∗ 0.055∗∗∗ (0.029) (0.018) (0.023) (0.037) (0.016) CBO Member in HH × 0.016 −0.003 0.019 −0.002 −0.003 Treatment (0.035) (0.026) (0.028) (0.036) (0.028) CBO member in HH 0.004 0.006 −0.007 0.070∗∗ 0.057∗∗∗ (0.028) (0.016) (0.016) (0.031) (0.014) Constant 1.076∗∗∗ 0.001 −0.020 0.206 0.151 (0.093) (0.066) (0.084) (0.139) (0.153) Baseline outcome Yes Yes Yes No Yes Baseline controls Yes Yes Yes Yes Yes Group fixed effects Yes Yes Yes Yes Yes Enumerator fixed effects Yes Yes Yes Yes Yes Observations 2,271 2,271 2,271 2,271 2,271 R-squared 0.226 0.123 0.165 0.338 0.182 Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: Robust standard errors clustered at the neighborhood level in parentheses. ∗∗∗ p < 0.01, ∗ ∗ p < 0.05, ∗ p < 0.1. Table S1.6. Baseline CBO Characteristics (Phase 1) Mean Total Mean control Mean treatment p-value CBO flood prevention and cleaning activities CBO has taken action in relation to flooding in last year 0.77 0.75 0.78 0.77 Flood activities: Raising awareness 0.36 0.34 0.38 0.80 Flood activities: Financial assistance for victims 0.20 0.28 0.13 0.12 Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Flood activities: Cleaning of canals and lakes 0.14 0.13 0.16 0.72 Flood activities: Surveillance of canals and lakes 0.09 0.06 0.13 0.40 Flood activities: Cooperation with other stakeholders 0.33 0.28 0.38 0.43 Flood activities: Small works to avoid flood 0.14 0.09 0.19 0.29 Objective of CBO: Raise awareness on flood-related 0.23 0.22 0.25 0.77 issues / organize cleaning activities CBO engaged in flood communication campaign 0.30 0.38 0.22 0.18 CBO engaged in cleaning activities 0.61 0.66 0.56 0.45 Number of days spent on flood communication or 13.30 13.44 13.16 0.97 cleaning activities Number of members engaged in flood communication 52.48 64.78 40.19 0.09 or cleaning activities CBO characteristics Number of members 185.22 177.53 192.91 0.90 Number of members present at the last meeting 33.06 35.72 30.41 0.46 Number of members engaged in activities of CBO 110.34 143.66 77.03 0.08 Average hours per week members engage in CBO 13.02 11.50 14.53 0.53 activities Proportion of members that are women 0.54 0.55 0.53 0.81 Proportion of members that are young 0.07 0.05 0.08 0.30 CBO represents a head office 0.91 0.84 0.97 0.09 CBO votes in leaders by election 0.66 0.69 0.63 0.61 Most important benefit of membership of CBO are 0.11 0.13 0.09 0.69 benefits for the community CBO type: Sports and culture association 0.14 0.13 0.16 0.72 CBO type: Economic interest group 0.11 0.09 0.13 0.69 CBO type: Women’s involvement group 0.11 0.19 0.03 0.05 CBO type: Development association 0.47 0.53 0.41 0.32 CBO type: Flood prevention committee 0.02 0.00 0.03 0.32 CBO type: Other association 0.16 0.06 0.25 0.04 Collaboration between different CBOs CBO only intervenes in this neighborhood 0.56 0.59 0.53 0.62 CBO collaborates with other CBOs in the 0.66 0.56 0.75 0.12 neighborhood CBO collaborates with other CBOs in other 0.73 0.69 0.78 0.40 neighborhoods Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: Number of observations is 64 representing the number of CBOs in the sample in Phase 1; p-value based on a test of the statistical significance of the difference in the mean of the control group and mean of the treatment group at baseline. Table S1.7. Baseline CBO Characteristics (Phase 2) Mean total Mean control Mean treatment p-value CBO flood prevention and cleaning activities CBO has taken action in relation to flooding in last year 0.64 0.70 0.58 0.23 Flood activities: Raising awareness 0.26 0.27 0.25 0.82 Flood activities: Financial assistance for victims 0.26 0.27 0.25 0.82 Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Flood activities: Cleaning Of canals and lakes 0.09 0.10 0.08 0.73 Flood activities: Surveillance of canals and lakes 0.01 0.02 0.00 0.32 Flood activities: Cooperation with other stakeholders 0.22 0.23 0.21 0.81 Flood activities: Small works to avoid flood 0.15 0.19 0.10 0.25 Objective of CBO: Raise awareness on flood-related 0.16 0.17 0.15 0.78 issues / organize cleaning activities CBO engaged in flood communication campaign 0.17 0.15 0.19 0.59 CBO engaged in cleaning activities 0.54 0.63 0.46 0.10 Number of days spent on flood communication or 11.66 8.25 15.06 0.26 cleaning activities Number of members engaged in flood communication 46.79 55.77 37.81 0.45 or cleaning activities CBO characteristics Number of members 172.55 215.00 130.10 0.34 Number of members present at the last meeting 43.08 40.31 45.85 0.49 Number of members engaged in activities of CBO 83.87 100.15 67.58 0.33 Average hours per week members engage in CBO 11.73 12.21 11.25 0.80 activities Proportion of members that are women 0.65 0.62 0.67 0.45 Proportion of members that are young 0.04 0.06 0.02 0.02 CBO represents a head office 0.90 0.85 0.94 0.19 CBO votes in leaders by election 0.57 0.69 0.46 0.02 Most important benefit of membership of CBO are 0.10 0.15 0.06 0.19 benefits for the community CBO type: Sports and culture association 0.12 0.19 0.04 0.02 CBO type: Economic interest group 0.15 0.10 0.19 0.25 CBO type: Women’s involvement group 0.22 0.21 0.23 0.81 CBO type: Development association 0.46 0.46 0.46 1.00 CBO type: Flood-prevention committee 0.04 0.00 0.08 0.04 CBO type: Other association 0.02 0.04 0.00 0.16 Collaboration between different CBOs CBO only intervenes in this neighborhood 0.71 0.69 0.73 0.66 CBO collaborates with other CBOs in the 0.70 0.71 0.69 0.83 neighborhood CBO collaborates with other CBOs in other 0.60 0.69 0.52 0.10 neighborhoods Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: Number of observations is 96 representing the number of CBOs in the sample in Phase 2; p-value based on a test of the statistical significance of the difference in the mean of the control group and mean of the treatment group at baseline. Table S1.8. Baseline Household Characteristics (Phase 1) n Mean total Mean control Mean treatment p-value Characteristics of household head Male 922 0.67 0.69 0.64 0.13 Muslim 922 0.96 0.97 0.95 0.22 Married 922 0.75 0.77 0.73 0.07 Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Age 922 56.64 57.14 56.14 0.26 Salary (FCFA per day) 405 8880 9,535 8,222 0.61 Household size 922 10.99 11.10 10.88 0.59 Cleaning and flooding–related characteristics Cleanliness of 922 0.25 0.26 0.25 0.68 neighborhood Flood victim in last year 922 0.26 0.28 0.25 0.28 Training about flood 922 0.14 0.15 0.13 0.47 prevention Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: p-value based on a test of the statistical significance of the difference in the mean of the control group and mean of the treatment group at baseline; n indicates the number of observations. Table S1.9. Baseline Household Characteristics (Phase 2) n Mean total Mean control Mean treatment p-value Characteristics of household head Male 1,363 0.71 0.71 0.71 0.92 Muslim 1,363 0.95 0.95 0.94 0.55 Married 1,363 0.82 0.82 0.81 0.54 Age 1,363 54.57 54.31 54.83 0.44 Salary (FCFA per day) 657 12,121 14,588 9,661 0.13 Household size 1,363 9.61 9.48 9.73 0.37 Cleaning and flooding–related characteristics Cleanliness of 1,363 0.33 0.32 0.34 0.26 neighborhood Flood victim in last year 1,363 0.19 0.16 0.21 0.02 Training about flood 1,363 0.16 0.17 0.13 0.04 prevention Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: p-value based on a test of the statistical significance of the difference in the mean of the control group and mean of the treatment group at baseline; n indicates the number of observations. Table S1.10. Baseline Household Characteristics: Comparability of Phase 1 and Phase 2 Treatment Groups Mean treatment phase Mean treatment phase 1 2 p-value Characteristics of household head Male 0.641 0.710 0.015∗∗ Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Muslim 0.954 0.943 0.392 Married 0.726 0.811 0.001∗∗∗ Age 56.143 54.830 0.089∗ Salary (FCFA per day) 8222 9661 0.503 Household size 10.880 9.727 0.000∗∗∗ Cleaning and flooding–related characteristics Cleanliness of neighborhood 0.246 0.345 0.000∗∗∗ Flood victim in last year 0.246 0.213 0.191 Training about flood prevention 0.133 0.135 0.911 Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: p-value based on a test of the statistical significance of the difference in the mean of the treatment group in Phase1 and the treatment group in Phase 2 at baseline. Table S1.11. Baseline CBO Characteristics: Comparability of Phase 1 and Phase 2 Treatment Groups Mean treatment Mean treatment phase 1 phase 2 p-value CBO flood-prevention and cleaning activities CBO has taken action in relation to flooding in 0.781 0.583 0.068∗ last year Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Flood activities: Raising awareness 0.375 0.250 0.237 Flood activities: Financial assistance for victims 0.125 0.250 0.175 Flood activities: Cleaning of canals and lakes 0.156 0.083 0.318 Flood activities: Surveillance of canals and lakes 0.125 0.000 0.012∗∗ Flood activities: Cooperation with other 0.375 0.208 0.104 stakeholders Flood activities: Small works to avoid flooding 0.188 0.104 0.295 Objective of CBO: Raise awareness on 0.250 0.146 0.248 flood-related issues/organize cleaning activities CBO engaged in flood-communication campaign 0.219 0.188 0.736 CBO engaged in cleaning activities 0.563 0.458 0.368 Number of days spent on flood-communication 13.156 15.063 0.817 or cleaning activities Number of members engaged in 40.188 37.813 0.868 flood-communication or cleaning activities CBO characteristics Number of members 30.406 45.854 0.062∗ Number of members present at the last meeting 77.031 67.583 0.659 Number of members engaged in activities of CBO 14.531 11.250 0.455 Average hours per week members engage in CBO 0.528 0.674 0.042∗∗ activities Proportion of female members 0.085 0.017 0.003∗∗∗ Proportion of youth members 0.969 0.938 0.536 CBO represents a head office 0.625 0.458 0.147 CBO votes in leaders by election 0.094 0.063 0.609 Most important benefit of membership of CBO 0.156 0.042 0.077∗ are benefits for the community CBO type: Sports and culture association 0.125 0.188 0.464 CBO type: Economic interest group 0.031 0.229 0.015∗∗ CBO type: Women’s involvement group 0.406 0.458 0.650 CBO type: Development association 0.031 0.083 0.352 CBO type: Flood-prevention committee 0.250 0.000 0.000∗∗∗ CBO type: Other association 30.406 45.854 0.062∗ Collaboration between different CBOs CBO only intervenes in this neighborhood 0.531 0.729 0.071∗ CBO collaborates with other CBOs in the 0.750 0.688 0.551 neighborhood CBO collaborates with other CBOs in other 0.781 0.521 0.018∗∗ neighborhoods Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: p-value based on a test of the statistical significance of the difference in the mean of the treatment group in Phase1 and the treatment group in phase 2 at baseline. Table S1.12. Heterogeneity by Flood Risk and Infrastructure Phase: Additional Results (1) (2) (3) (4) Panel A Awareness of OQP Household received training on flood prevention Treatment 0.199∗∗∗ 0.252∗∗∗ 0.053∗∗∗ 0.042∗∗ (0.028) (0.036) (0.014) (0.017) Phase 1 0.465∗∗∗ 0.466∗∗∗ 0.079 0.079 Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 (0.084) (0.085) (0.050) (0.050) Phase 1 × Treatment −0.132∗∗∗ 0.027 (0.042) (0.032) 2,271 2,271 2,271 2,271 R-squared 0.338 0.341 0.182 0.182 Panel B Treatment 0.198∗∗∗ 0.183∗∗∗ 0.054∗∗∗ 0.045∗∗ (0.028) (0.033) (0.014) (0.019) Flood risk 0.016 −0.001 −0.005 −0.015 (0.020) (0.028) (0.016) (0.020) Flood risk × Treatment 0.035 0.019 (0.035) (0.021) 2,271 2,271 2,271 2,271 R-squared 0.338 0.338 0.182 0.182 Panel C: Phase 1 Treatment 0.124∗∗∗ 0.128∗∗∗ 0.065∗ 0.045 (0.036) (0.041) (0.033) (0.039) Flood risk −0.014 −0.008 0.005 −0.019 (0.023) (0.035) (0.032) (0.043) Flood risk × Treatment −0.011 0.048 (0.054) (0.045) 922 922 922 922 R-squared 0.418 0.418 0.253 0.254 Panel D: Phase 2 Treatment 0.240∗∗∗ 0.216∗∗∗ 0.041∗∗ 0.039∗ (0.039) (0.046) (0.018) (0.021) Flood risk 0.026 −0.004 −0.002 −0.005 (0.025) (0.033) (0.018) (0.020) Flood risk × Treatment 0.059 0.006 (0.045) (0.025) 1,349 1,349 1,349 1,349 R-squared 0.336 0.337 0.156 0.156 Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: All specifications include the value of the outcome variable at baseline, baseline controls, group fixed effects and enumerator fixed effects. Robust standard errors clustered at the neighborhood level in parentheses. ∗∗∗ p < 0.01, ∗ ∗ p < 0.05, ∗ p < 0.1 based on robust clustered standard errors. Table S1.13. Heterogeneity by Wealth and Flood Risk (1) (2) (2) (4) (5) Flood victim this Flood victim last Received training Panel A: Full Sample Cleanliness year year Awareness of OQP on flood prevention Treatment 0.129∗∗∗ −0.041 −0.024 0.227∗∗∗ 0.042 (0.037) (0.029) (0.033) (0.046) (0.026) Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 Treatment × Wealth 2 −0.006 0.019 0.019 −0.028 0.007 (0.047) (0.032) (0.040) (0.043) (0.041) Treatment × Wealth 3 −0.030 0.043 −0.023 −0.103 0.027 (0.055) (0.052) (0.062) (0.063) (0.032) Treatment × Wealth 4 0.012 0.028 0.020 0.002 0.027 (0.050) (0.046) (0.055) (0.060) (0.034) Wealth 2 −0.014 −0.036 −0.046 −0.001 −0.006 (0.035) (0.027) (0.031) (0.029) (0.024) Wealth 3 0.016 −0.033 −0.059 0.069 −0.009 (0.034) (0.035) (0.044) (0.043) (0.023) Wealth 4 −0.003 −0.075∗∗ −0.118∗∗∗ −0.007 0.000 (0.030) (0.031) (0.035) (0.044) (0.026) Observations 2,271 2,271 2,271 2,271 2,271 R-squared 0.227 0.129 0.172 0.339 0.183 Panel B: Phase 1 Treatment 0.197∗∗∗ −0.057 0.002 0.188∗∗ −0.001 (0.038) (0.063) (0.072) (0.066) (0.059) Treatment × Wealth 2 −0.127∗ 0.086 0.058 −0.105 0.069 (0.063) (0.069) (0.082) (0.063) (0.074) Treatment × Wealth 3 −0.130∗∗ 0.058 −0.138 −0.138∗ 0.107∗ (0.061) (0.086) (0.103) (0.079) (0.052) Treatment × Wealth 4 −0.070 0.066 −0.012 0.018 0.107 (0.085) (0.091) (0.102) (0.110) (0.068) Wealth 2 0.041 −0.075 −0.081 0.039 −0.021 (0.042) (0.051) (0.049) (0.041) (0.036) Wealth 3 0.020 −0.065 −0.064 0.127∗∗ −0.071∗ (0.060) (0.062) (0.070) (0.048) (0.036) Wealth 4 −0.038 −0.109∗ −0.151∗∗ −0.040 −0.061 (0.055) (0.061) (0.058) (0.073) (0.038) Observations 922 922 922 922 922 R-squared 0.309 0.149 0.196 0.425 0.257 Panel C: Phase 2 Treatment 0.076 −0.033 −0.030 0.228∗∗∗ 0.062∗ (0.047) (0.031) (0.033) (0.057) (0.031) Treatment × Wealth 2 0.037 −0.024 −0.018 0.039 −0.037 (0.059) (0.038) (0.051) (0.061) (0.036) Treatment × Wealth 3 0.032 0.031 0.048 −0.019 −0.015 (0.069) (0.059) (0.062) (0.091) (0.050) Treatment × Wealth 4 0.067 −0.001 0.031 0.022 −0.015 (0.065) (0.049) (0.057) (0.075) (0.051) Wealth 2 −0.027 −0.012 −0.022 −0.019 0.011 (0.045) (0.033) (0.036) (0.045) (0.027) Wealth 3 0.012 −0.009 −0.053 −0.009 0.034 (0.053) (0.040) (0.049) (0.064) (0.040) Wealth 4 0.027 −0.047 −0.088∗∗ 0.009 0.048 (0.045) (0.033) (0.042) (0.057) (0.041) Observations 1,349 1,349 1,349 1,349 1,349 R-squared 0.237 0.122 0.149 0.337 0.160 Panel D: Full Sample Treatment 0.140∗∗∗ −0.015 −0.001 0.193∗∗∗ 0.051∗∗ Table S1.13. Continued (1) (2) (2) (4) (5) Flood victim this Flood victim last Received training Panel A: Full Sample Cleanliness year year Awareness of OQP on flood prevention (0.033) (0.015) (0.024) (0.038) (0.025) Treatment × Wealthy 0.001 0.031 −0.014 −0.024 −0.013 Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 (0.049) (0.030) (0.035) (0.047) (0.033) Treatment × Flood risk −0.030 −0.035 −0.029 0.032 −0.010 (0.046) (0.031) (0.044) (0.037) (0.024) Treatment × Wealthy −0.037 −0.032 −0.003 −0.001 0.113∗∗ × Flood risk (0.086) (0.066) (0.082) (0.076) (0.044) Wealthy × Flood risk 0.059 0.039 −0.026 0.017 −0.059∗ (0.049) (0.045) (0.055) (0.063) (0.033) Wealthy −0.007 −0.038∗∗ −0.044∗ 0.025 0.019 (0.033) (0.017) (0.024) (0.039) (0.021) Flood risk −0.012 0.061∗∗∗ 0.143∗∗∗ −0.001 0.004 (0.036) (0.021) (0.044) (0.030) (0.021) Observations 2,271 2,271 2,271 2,271 2,271 R-squared 0.227 0.129 0.172 0.339 0.183 Panel E: Phase 1 Treatment 0.112∗∗ −0.003 0.010 0.112∗∗∗ 0.057 (0.039) (0.028) (0.046) (0.036) (0.052) Treatment × Wealthy −0.013 0.007 −0.110∗ 0.032 −0.017 (0.068) (0.048) (0.056) (0.053) (0.067) Treatment × Flood risk 0.009 −0.017 0.052 0.015 −0.024 (0.060) (0.066) (0.093) (0.072) (0.054) Treatment × Wealthy −0.017 −0.000 0.000 −0.064 0.209∗∗ x Flood risk (0.099) (0.121) (0.143) (0.135) (0.086) Wealthy × Flood risk −0.025 0.052 −0.032 0.027 −0.174∗∗∗ (0.057) (0.077) (0.095) (0.097) (0.037) Wealthy −0.021 −0.055∗ −0.035 0.010 0.018 (0.049) (0.031) (0.045) (0.046) (0.038) Flood risk 0.011 0.066 0.131∗ −0.017 0.041 (0.050) (0.045) (0.074) (0.044) (0.040) Observations 922 922 922 922 922 R-squared 0.309 0.149 0.196 0.425 0.257 Panel F: Phase 2 Treatment 0.103∗∗ −0.018 −0.005 0.230∗∗∗ 0.038 (0.045) (0.022) (0.028) (0.057) (0.031) Treatment × Wealthy 0.059 0.033 0.042 −0.036 −0.001 (0.060) (0.036) (0.039) (0.071) (0.040) Treatment × Flood risk −0.010 −0.060∗ −0.087∗∗ 0.035 0.003 (0.067) (0.031) (0.043) (0.047) (0.028) Treatment × Wealthy −0.133 −0.057 −0.014 0.083 0.028 × Flood risk (0.116) (0.064) (0.097) (0.098) (0.063) Wealthy 0.217∗∗∗ 0.021 0.000 −0.025 0.038 × Flood risk (0.066) (0.057) (0.065) (0.089) (0.049) Wealthy −0.035 −0.015 −0.045 0.020 0.024 (0.045) (0.022) (0.027) (0.057) (0.031) Flood risk −0.064 0.066∗∗∗ 0.193∗∗∗ 0.006 −0.008 Table S1.13. Continued (1) (2) (2) (4) (5) Flood victim this Flood victim last Received training Panel A: Full Sample Cleanliness year year Awareness of OQP on flood prevention (0.054) (0.023) (0.056) (0.034) (0.025) Observations 1,349 1,349 1,349 1,349 1,349 Downloaded from https://academic.oup.com/wber/article/38/4/824/7626364 by LEGVP Law Library user on 29 October 2024 R-squared 0.237 0.122 0.149 0.337 0.160 Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: The household-level wealth index is constructed using information on housing (source of water, type of floor, type of energy used for cooking, heating and light, number of rooms per person) and ownership of assets (radio, fridge, sewing machine, bicycle, TV, computer, and vehicle). The index is a simple sum of each of these assets and ranges from 0 to 13. The analysis uses quartiles of this index and creates an interaction term between the treatment indicator and the top three quartiles with the lowest quartile in the base category. The analysis also considers whether there is an interactive effect between flood risk and wealth. The study combines the top two wealth categories and the bottom two wealth categories to create a 0–1 indicator variable which takes a value of 1 if the household is in the top two categories and 0 otherwise. The analysis interacts it with the treatment and with the flood risk at baseline. All specifications include baseline controls, group fixed effects, and enumerator fixed effects. Robust standard errors clustered at the neighborhood level in parentheses. ∗ ∗ ∗ p < 0.01, ∗ ∗ p < 0.05, ∗ p < 0.1. Table S1.14. Baseline Individual Characteristics n Mean total Mean control Mean treatment p-value Ill in the last 30 days 22,749 0.22 0.224 0.214 0.07 Number of workdays missed 5,476 1.22 1.115 1.333 0.09 Source: Authors’ analysis based on data collected as part of the World Bank–assisted Senegal Stormwater Management and Climate Change Adaptation Project (PROGEP). Note: p-value based on a test of the statistical significance of the difference in the mean of the control group and the mean of the treatment group at baseline; n indicates the number of observations. C The World Bank 2024. Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. This is an Open Access article distributed under the terms of the Creative Commons Attribution 3.0 IGO License (https://creativecommons.org/licenses/by/3.0/igo/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.