WPS8313 Policy Research Working Paper 8313 The Impact of Social Mobilization on Health Service Delivery and Health Outcomes Evidence from Rural Pakistan Xavier Giné Salma Khalid Ghazala Mansuri Development Research Group Finance and Private Sector Development Team & Poverty and Equity Global Practice Group January 2018 Policy Research Working Paper 8313 Abstract This paper uses a randomized community development health providers. The study detects economically large program in rural Pakistan to assess the impact of citizen improvements in pregnancy and well-baby visits by lady engagement on the quality of public health services. The health workers, as well as increased utilization of pre- and program had a strong emphasis on organizing women, post-natal care by pregnant women. In contrast, the who also identified health services as a development pri- quality of supra-village health services did not improve, ority at baseline. Assessing the program at midline, the underscoring the importance of community enforcement paper finds that the mobilization effort alone had a sig- and monitoring capacity for improving service delivery. nificant impact on the performance of village-based This paper is a product of the Finance and Private Sector Development Team, Development Research Group and the Poverty and Equity Global Practice Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at xgine@worldbank.org and gmansuri@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team The Impact of Social Mobilization on Health Service Delivery and Health Outcomes: Evidence from Rural Pakistan e, Salma Khalid, and Ghazala Mansuri∗ Xavier Gin´ JEL Classification: C93, D7, H75, I11, O12 Keywords: public service delivery, health services, citizen engagement, commu- nity driven development, maternal and child health, Pakistan, Lady Health Worker (LHW) ∗ e: Development Research Group, The World Bank, xgine@worldbank.org. Khalid: Univer- Gin´ sity of Michigan, salmak@umich.edu. Mansuri: The World Bank, gmansuri@worldbank.org. This project was jointly funded by the Development Research Group and SAFANSI at the World Bank, and the Pakistan Poverty Alleviation Fund. The views expressed herein are those of the authors and should not be attributed to the World Bank, its executive directors, or the countries they represent. 1 Introduction Community-driven development (CDD) is viewed as an important vehicle for im- proving public sector accountability and the quality of public service delivery by both governments and donors. Despite this, evidence on the effectiveness of CDD programs remains mixed. This is due, in part, to the inherent difficulty of evaluat- ing interventions that aim to change the nature of the interaction between citizens and the state. Such interventions usually have complex and unpredictable trajecto- ries of change (Mansuri and Rao, 2013). However, it is also due to a fundamental characteristic of the CDD approach. Communities are offered a bundle of distinct interventions, usually simultaneously, making it difficult to identify what aspects of a program worked or did not work in a specific context. A case in point, and one which is of some policy interest, is the investment that CDD programs make on the social mobilization of poor and disenfranchised groups. Virtually all CDD programs invest considerable resources in supporting community organizations. These organizations are meant to provide a platform for disadvantaged groups to engage in collective action around development priorities and interact with and influence institutions of the state, at the local level. Assessing the impact of these investments is difficult, however, since social mobilization is invariably combined with resource injections for community infrastructure, asset transfers for the poor, skills training or microcredit, all of which can have an impact on the demand for improved public services or influence over the policy process through other channels. In this paper, we provide evidence on the impact of social mobilization on the quality of public service delivery in a context where other simultaneous inputs are absent. The Social Mobilization for Empowerment (MORE) program was implemented as a large- scale randomized intervention in rural Pakistan in 2010. It is a typical CDD program in design, however, in the first 3 years of the program, treatment villages were only provided support for social mobilization. By assessing program impact at this three- year mark, in mid-2013, we can decouple the impact of social mobilization from the 2 injection of resources or other inputs. The social mobilization effort focused on encouraging self-help and collective action within the community as well as better linkages with government. In treatment villages, citizens were organized into grassroots organizations which appointed rep- resentatives to a village-level institution that had the authority to decide on village development priorities and to eventually allocate resources from a village develop- ment fund (VDF), which was assessed and provided after the midline of the program was done in mid-2013. The social mobilization effort had a strong focus on increasing the participation of women in the village level decision making bodies. Since women identified access to primary health care as critical to their own needs and those of their children, at baseline, we look at the impact of mobilizing women on public health provision. It is important to note that the social mobilization effort did not focus on health-related issues and no information was provided to community members on the performance of local public health providers. The context we study is characterized by relatively high levels of maternal and child mortality, and malnutrition among infants and young children. Women have low decision-making power within the household and social mores restrict female mobility and autonomy. Education levels among adult women also remain extremely low, limiting their ability to access information or engage effectively with service providers. This context allows us to examine whether social mobilization targeted at women can lead to an improvement in the performance of public health providers even in a context of low female literacy and mobility. Rural villages in Pakistan have access to two types of health care providers. The first is a female community health provider known as the Lady Health Worker (LHW). LHWs deliver a range of services related to maternal and child health including pre- and post-natal care, well-baby visits, child growth monitoring, childhood immuniza- tion, family planning, and health education. Because LHWs are field workers who make home visits, particularly targeting households with young children or women 3 of child-bearing age, they are the first line of direct access to public health care. In addition, the house of each LHW is declared as a Health House where residents can go in case of emergency to receive basic treatment or advice. Due to this, LHWs are typically selected from and reside within the villages that they serve. While each village is entitled to a LHW, their presence was not universal at project start in 2010. In fact, only 62% of villages reported having a LHW assigned to them. The second type of health care provider is the Basic Health Unit (BHU), a primary care health facility that typically serves multiple villages within a catchment area. LHWs are responsible for making referrals of all pregnant women to the BHU which delivers additional pre- and post-natal care services and deals with minor illness of all types. While all villages in the study had a BHU within their catchment area, BHUs varied substantially in both quality and the availability of trained medical staff. Since expo- sure to the MORE program was randomized at the village-level, and the catchment area of a BHU typically includes both mobilized and non-mobilized villages, we ex- pect community mobilization to be less effective at influencing BHU-level outcomes, as compared to effects on village-based LHWs. We examine effects of community mobilization on two sets of health-related out- comes. The first focuses on women’s interactions with service providers (health care utilization, access to and quality of care from LHWs), the second looks at improve- ments in health outcomes for women and young children such as the incidence of illness, ante- and post-natal care, well-baby checkups and child immunization. We find no significant improvements overall in the utilization of BHUs. However, there is a substantial reduction in reported wait-times at these facilities and an improvement in the odds of a woman’s pregnancy being registered at the BHU as well as in the odds of receiving post-natal care. Since LHWs connect women to BHUs and are the first providers of post-natal care, we cannot disentangle whether the improvements in registration or post-natal care are due to improvements in service delivery at the LHW or BHU level. However, women are significantly more likely to report having been visited by an LHW. They also report significant improvements 4 in antenatal and postnatal care provided by the LHW, as well as significantly higher LHW well-baby visits, including a visit to check child height. In contrast, outcomes that are not driven by the type or quality of care provided by LHWs, such as the incidence of diarrhea or stunting, which depends far more on community-level factors such as water quality and sanitation conditions and house- hold health behaviors, such as the use of soap or barefoot walking among children and adults, registered no improvement. These findings suggest that community collective action can improve the performance of service providers only if the provider is accessible and can be held accountable by the village. This chain of accountability is most effective if the purview of the service provider is at the level of the mobilized community, as is the case with LHWs, and less effective, as in the case of BHUs, for providers who are located at the supra- community level and are therefore accountable to multiple stakeholders. The rest of the chapter is organized as follows. The next section describes the literature on social mobilization and health. Section 3 describes the data we use for the analysis, Sections 4 and 5 provide the econometric framework and results, and Section 6 concludes. 2 Community-Driven Development and Health The existing literature on community-based health service interventions suggests po- tentially positive impacts of CDD activities on health outcomes, particularly in the domain of maternal and child health. However, since these interventions bundle several activities together, one cannot isolate the impact of community mobilization alone. Community-based health service programs encompass a range of activities that focus on maternal and child care and household health behaviors. These inter- ventions can be roughly divided into two categories: (1) projects where communities are encouraged to take an active role in resource allocation, and (2) interventions where community volunteers or community-based health workers are mobilized to 5 deliver health services or information. For example, Chase and Sherburne-Benz (2001) examine the impact of community organization and resource allocation via the Zambia Social Fund on health and ed- ucation outcomes. They find that communities using social investment funds to construct a health facility see higher utilization of primary care services and lower utilization of hospital services compared to control communities, but they find no overall difference in total health care utilization between treated and control com- munities. Other community-based health projects mobilize communities to improve health through direct engagement with formal service providers. Binka et al. (2007) im- plement a randomized intervention in Ghana to compare the efficacy of providing trained nurses to communities versus community volunteers. While in this study vol- unteers on their own do not improve child survival significantly, the combination of volunteers working together with trained nurses outperforms nurses working on their own. This suggests a strong role for community organization in improving health orkman and Svensson outcomes when used as a supplement to formal provision. Bj¨ (2009) evaluate the impact of citizen report cards on quality of health care deliv- ery. They find that improvements in outcomes and service provision vectors can be best explained through the degree of community engagement with the program as opposed to supply-driven factors such as the engagement of the staff. Interventions in India (Tripathy et al., 2010) and Nepal (Manandhar et al., 2004) use community facilitators to organize women’s groups that tackle, among other subjects, health behaviors and health entitlements. Both randomized trials find improvements along a range of outcomes, with large reductions in neonatal mortality. On balance, the literature on CDD and health suggests that communities can play a significant role in improving community health through various mechanisms includ- ing resource allocation, health service delivery, dissemination of information, and monitoring of service providers. However, the literature to date, while suggestive of a positive role for social mobilization alone, has not identified it cleanly. 6 3 The MORE Program The goal of the MORE program is to foster social mobilization and strengthen community development through the creation of community- and village-level or- ganizations and the provision of village-level development funds. The program was implemented in partnership with the Pakistan Poverty Alleviation Fund (PPAF). Social mobilization activities in the study areas were supported by a key partner of the PPAF, the National Rural Support Program (NRSP). NRSP is the largest community-based development NGO in Pakistan in terms of outreach and coverage, and currently operates in 51 districts spread across all four provinces of Pakistan. NRSP identified 158 villages drawn from 5 districts where it currently has presence.1 The identified villages had no prior history of social mobilization by either NRSP or any other organization. 108 study villages were randomly assigned to treatment status with the remaining being held as controls. In treatment villages, representatives from NRSP helped organize villagers into grass- roots organizations of 15 to 20 members called Community Organizations (COs). The aim of the COs is to provide a platform for collective efforts and allow members to pool their resources for common development goals. COs hold regular meetings where members can discuss local issues, prioritize community needs, and resolve any conflicts at the local level. The procedure followed by NRSP for social mobilization was standardized in all the villages and districts to allow comparability. In the treatment villages, a social mobilization team (SMT) approached a few people in the village to help organize a meeting of the community with the Social Organizer (SO). In that meeting, the SO introduced the concept of Community Organization (CO) and how villagers can pool their resources to create a platform for collective efforts. The SO shared ex- amples of other areas where people formed COs and were able to achieve significant improvement in their lives through this platform. 1 The districts are Nowshera, Mianwali, Bahawalpur, Hyderabad and Tando Muhammed Khan. 7 The SO also informed the community that the basis on which they would get funds for developmental activities is the number of households organized in a village, where a household is considered organized if at least one member (male or female) is an active member of a CO and has attended more than one CO meeting. A minimum of 40% of the village population needed to be organized in order to be eligible for the village development grant, with the size of the grant increasing with the number of households organized past the 40% cut-off, thus providing a strong incentive for broad mobilization. In addition, the inclusion of women and poor households in the mobilization and CO formation process was actively encouraged. Once 40% of village households had at least one CO member, the village formed a Village Support Organization (VSO). This village institution comprises two elected members from each CO in the village. One of the main tasks of the VSO was the design and implementation of the Village Development Plan (VDP), a document that prioritized village development projects to be funded by the grant. The grant could be used for any productive purpose for the general benefit of the entire community including physical infrastructure, health, education, training, asset transfers and other livelihood activities. The amount of the grant varied from village to village depending on the total number of households and percentage of households that were organized in that village. On average, villages received a grant totaling 2,897,883 Pakistan rupees (PKR), or 10,482 PKR per household in the village. The VSO was also charged with the management of the grant and the active involvement of community members in monitoring and promoting transparency. The MORE intervention was successful at encouraging broad participation from the community. On average, 59% of households in treatment villages were organized. Women were well-represented in the community mobilization activities, comprising 51% of CO members and 41% of VSO members per village, on average. The timing of the intervention and data collection allows us to isolate the impact of community mobilization from the direct impacts of the village-level grants. In each treatment and control village, households were surveyed at baseline, after the 8 formation of the first COs in treatment villages. Households were surveyed again 3 years later at midline. In treatment villages, the midline survey occurred after approval of the VDP but before the disbursement of grant funds. This study focuses on impacts at midline between treatment and control villages, which isolates the impact of community mobilization. 3.1 Data The baseline and midline surveys were administered to a random sample of 40 house- holds drawn from each of the treatment and control villages and included detailed modules on health facility utilization, health outcomes and household health behav- iors. All adult women in the household were separately surveyed for specific sub-modules related to maternal health, antenatal and postnatal care, and child birth and health outcomes. Respondents were asked about their most recent pregnancy in the past 3 years to cover relevant health care utilization for pregnancies occurring between the baseline and midline data collection. Finally, all women in the household were asked about their interaction with the LHW assigned to their village. As discussed above, the quality of service provided by the LHW should be responsive to changes in local accountability, given that the LHW is recruited from within the community that she serves. 3.2 Sample Characteristics Table 1 presents the mean of village level characteristics and checks for balance between treatment and control villages. Villages have about 279 households living in 7 to 8 settlements on average. Villages are relatively poor, with about 52% of households below the poverty line and landless households comprising about 67% of all households. Households have between 6 and 7 members on average. Most household heads (63%) do not have formal education. 9 When we compare treatment and control villages we find no relationship between treatment assignment and any of these variables, giving us confidence in the success of our randomization. In fact, when we run a regression of treatment status as the dependent variable against all of these variables, the p-value for the F-test that all the variables are jointly significant is 0.98. Table 2 provides a description of the variables used in the analysis that follows. Our outcomes of interest are broadly classified as incidence of illness, utilization of health services, quality of care provided by the BHU, maternal health and child health outcomes, and perceptions regarding the quality of care provided by the LHW. 4 Econometric Framework Given random assignment to treatment and control villages, we can compare midline outcomes between experimental groups in order to establish the causal impact of the treatment on the variables of interest. We estimate the following specification: Yivb = α + βTvb + γb + ivb (1) where Yivb is the outcome of interest for household i in village v mapped to SMT b, Tvb is an indicator for whether village v was assigned to the treatment group, and γb are SMT fixed effects. The coefficient β measures the impact of social mobilization by capturing the difference in the outcome between treatment and control villages. We cluster standard errors at the village level since treatment is assigned at the village level. To allay concerns related to multiple hypothesis testing, we also create composite indices of related variables (see Kling et al., 2007) and assess treatment effects relative to these indices in addition to their individual components. 10 5 Results The first set of outcomes in Table 3 relate to the overall incidence of illness in the past month and health services sought for these episodes of illness. Column 1 indicates that at midline, self-reported incidence of illness is significantly lower among households in treatment villages relative to those in control villages where no mobilization had occurred. Households in treatment villages also appear to consult a larger number of providers when a household member is ill, as shown in Column 2. While over 90% of households report seeking some form of consultation during episodes of illness, only 20% of households seek health services from government providers. Column 3 reports no increase in the likelihood of utilizing government health service providers following social mobilization. Table 4 analyzes whether assignment to treatment results in an improvement in the experiences of households using the BHU. Column 1 reports a statistically significant reduction of roughly 6 minutes in wait times reported at the BHU among households in treated villages. However, no other indicator shows significant improvement. Con- sequently, our BHU index which combines all measures of BHU performance has a positive but insignificant coefficient. Note that the sample is smaller because only households that visited the BHU facility provided information on their performance. Tables 5 and 6 look at maternal outcomes pre- and post-delivery, as well as child outcomes immediately following birth. Since LHWs are the first point of contact for pregnant women and BHUs provide the secondary level of care, this set of outcomes could plausibly have been influenced by better performance of BHUs or LHWs. Table 5 reports pre-delivery outcomes while Table 6 focuses on post-delivery mother and child outcomes. Since these data cover pre-delivery outcomes for completed preg- nancies, and there was only a 3 year period between the start of social mobilization and midline data collection, we expect weaker effects on pre-natal outcomes due to the lower exposure to treatment. Column 1 of Table 5 shows that at midline there had been no increase in the odds of 11 pregnancy from the base of 44% at baseline. There is a statistically significant 20% increase in the odds of a pregnancy being registered at the BHU, over a base of 26%, but no significant change in the odds of receiving antenatal care. Overall, the effect of social mobilization on improvements on pre-pregnancy maternal health, captured in in Column 4 of Table 5, is not significant. Examining the post-delivery outcomes (Table 6), we find a significant and sizeable increase in utilization of postnatal care in treatment villages. The likelihood of seeking postnatal care increased by 26% in treatment villages, relative to a base of 27%. In contrast, there is no change in the odds of child mortality at birth, birth registration or the recording of weight at birth. Overall, the coefficient for the post-pregnancy index is significant at conventional levels, driven in large part by the substantial increase in post-natal care. Tables 7 and 8 focus on the performance of LHWs. Table 7 looks at the incidence of specific services provided by the LHW while Table 8 looks at household perceptions of satisfaction with the LHW service provision. Column 1 of Table 7 indicates that the likelihood of the LHW visiting pregnant women in treatment villages rose by 19% from a base of 35% in control villages. Treatment villages also report a 37% higher probability that pregnant women received antenatal care from LHWs, from a base of 14% and a near doubling of LHW provided postnatal care, though from a very low base of 3% (Columns 2 and 3). Given that there was an insignificant increase in the level of antenatal care in treatment villages overall, the increase in care provided by the LHW implies a substitution away from other providers to the LHW. Turning to child outcomes, we again see a significant and large increase in the prob- ability of receiving a well-baby visit by the LHW to check infant height and weight. The odds of the LHW making a well-baby visit more than doubled in the treatment sample, though again the probability in control villages was only 4% (Column 4). The impact of social mobilization on our LHW index, which combines standardized measures from Columns 1-4, is positive and statistically significant at the 1% level. Table 8 captures household perceptions of LHW performance in two samples. Columns 12 1-3 include all women of reproductive age while Columns 4-6 restrict the sample to women who had a completed pregnancy in the past 3 years. In both samples we see a large and significant increase in the odds of households reporting that a LHW was assigned to their village. The size of the effect ranges from 20% to 25%, depending on the sample, from a base of 60% in control villages. This result may be explained by a greater presence of an already assigned LHW in the community or the assignment of new LHWs to a previously unserved village. Interestingly, conditional on being in a village with an assigned LHW, treated households do not report an increase in the frequency of LHW visits or a higher satisfaction with LHW visits relative to households in control villages. Tables 9 and 10 turn to health outcomes for infants and young children up to 3 years of age. For this sample of children, Column 2 of Table 9 reports that the odds of having an immunization card are substantially higher in treatment communities (39% increase from a base of 11% in control communities). However, there is no statistically significant impact on the completeness of the immunization record in Column 1. It is worth nothing that conditional on having an immunization record available, completeness rates for immunization were at 65% among controls. This complements the results on LHW service provision in Table 7, since LHWs typically identify children eligible for immunization and work together with field workers to provide immunizations. Table 10 finds no significant impact of social mobilization on the incidence of diarrhea (Column 1) or child stunting (Column 2) in children aged 3 and under.2 Finally, Table 11 shows no change in the use of soap or the incidence of walking barefoot in the home or around the village. Overall, this suggests weak evidence for improvement in water and sanitation outcomes or other household health behaviors. Again, this may not be surprising as no particular investments aimed at either water 2 Column 2 has fewer observations because the measure for stunting requires the age in months that was only collected at endline. There was some attrition in the sample, including an entire district (Nowshera) which could not be surveyed due to security concerns. As a check, we examine immunization outcomes and incidence of diarrhea for this restricted sample and find similar results compared to the full sample. 13 and sanitation or preventive health information were made by the time the midline data were collected. 6 Conclusion In this paper, we assess whether social mobilization aimed at strengthening women’s participation in collective action can improve the performance of public health providers even in the absence of ancillary health inputs or financing. We find little overall im- provement in the quality of services provided by supra-village public providers like Basic Health Units (BHU). In contrast, we see a substantial increase in the quality of service provision by village-based skilled female health workers under the Lady Health Worker program. BHUs cater to multiple villages in a catchment area, not all of which were organized, limiting the capacity of any one village to influence BHU- level performance through any collective action measures. In comparison, the LHW’s catchment area is limited to the village in which she typically resides, allowing for a more effective exercise of collective action on the part of the community in ensuring her presence and monitoring her performance. Specifically, we find that a range of health services which fall under the purview of the LHW show a significant improvement in villages that were mobilized. This includes access to antenatal care, post-natal care and well-baby visits. Households in mobilized villages are also far more likely to report receiving visits from LHWs during pregnancy or reporting that they have an LHW assigned to their village. The improvement occurs in a context where there was no treatment effect on the odds of pregnant women receiving any antenatal care, suggesting a substitution away from other public and private providers towards LHWs. This is not the case for post-natal care, where we find a sizeable increase in access to care among women in mobilized villages. Given that LHW provision of postnatal care is low at baseline, even the doubling of care by LHWs that we observe in treated villages cannot explain the overall increase in access to post-natal care. This implies a greater use of private 14 facilities by women in mobilized villages given that there is no increase of BHU utilization. Our results suggest that while community collective action is not a panacea for im- proving all levels of public service delivery, it can be quite effective in improving aspects of service delivery where community members have enforcement and moni- toring capacity. The results also show that the active engagement of women in efforts to improve community collective action can have important payoffs in improved ser- vice provision targeting to the needs of women and young children. 15 References Binka, Fred N., et al. “Rapid achievement of the child survival millennium develop- ment goal: evidence from the Navrongo experiment in Northern Ghana.” Tropical Medicine & International Health 12.5 (2007): 578-593. orkman, Martina, and Jakob Svensson. “Power to the people: evidence from a Bj¨ randomized field experiment on community-based monitoring in Uganda.” The Quarterly Journal of Economics 124.2 (2009): 735-769. Chase, Robert S., and Lynn Sherburne Benz. “Household effects of community edu- cation and health initiatives: evaluating the impact of the Zambia Social Fund.” World Bank, Washington, DC Processed (2001). Manandhar, Dharma S., et al. “Effect of a participatory intervention with women’s groups on birth outcomes in Nepal: cluster-randomised controlled trial.” The Lancet 364.9438 (2004): 970-979. Mansuri, Ghazala, and Vijayendra Rao. Localizing development: does participation work? World Bank Publications, 2012. Tripathy, Prasanta, et al. “Effect of a participatory intervention with women’s groups on birth outcomes and maternal depression in Jharkhand and Orissa, India: a cluster-randomised controlled trial.” The Lancet 375.9721 (2010): 1182-1192. 16 Table 1: Descriptive Statistics Descriptive Statistics All Treatment Control P-Value (T=C) Number of Villages 158 108 50 Number of Households (HHs) 5828 3990 1838 Ever Married Women 15-40 years 6109 4169 1940 Women w/ pregnancies in past 3 yrs 2762 1907 855 Children <= 3 years of age at midline 4509 3060 1449 Village Population (No. of HHs) 278.98 267.32 284.38 0.499 (18.5) (14.9) No. of Settlements 7.50 7.82 7.35 0.720 (1.1) (0.7) No. of villages in Union Council 10.93 10.74 11.03 0.685 (0.6) (0.4) Proportion of Landless HHs in Village 0.67 0.67 0.66 0.820 (0.030) (0.020) Proportion of Poor HHs in Village 0.52 0.52 0.53 0.690 (0.006) (0.005) Number of HH members 6.34 6.33 6.34 0.968 (0.163) (0.127) HH heads with primary education 0.15 0.15 0.14 0.587 (0.012) (0.006) HH heads with middle education 0.18 0.17 0.19 0.365 (0.017) (0.013) 17 Table 2: Description of Variables Variable Name Description Illness Characteristics Incidence of Illness Fell ill in past month (1=Yes) N. of Consultations (If sick) Number of health care providers consulted Govt. Provider Consulted (If sick) Govt. health care provider was consulted (1=Yes) BHU Utilization Wait Time (If used BHU) Wait time at BHU Consult Fee (If used BHU) Amount of consultation fee paid at BHU Convey Concerns (If used BHU) Able to convey concerns to service provider (1=Yes) Treated Well (If used BHU) Treated well by the service provider (1=Yes) BHU Index Index combining Wait Time, Consult Fee, Convey Concerns, Treated Well Pregancy Pregnancy Pregnant in the past 3 years (1=Yes) Registered (If pregnant) Registered with the BHU (1=Yes) Antenatal Care (If pregnant) Received antenatal care during this pregnancy (1=Yes) Pre-Preg Index Index combining Pregnancy registered and Antenatal Care Postnatal Care Received postnatal care following delivery (1=Yes) Birth Registered Child was registered at BHU after delivery (1=Yes) Weight Recorded Child was weighed at birth (1=Yes) Post-Preg Index Index combining Postnatal Care, Birth Registered, Weight Recorded Lady Health Worker Performace and Satisfaction Visit LHW visited during last pregnancy (1=Yes) Antenatal Care Antenatal care received from the LHW (1=Yes) Postnatal Care Postnatal care received from the LHW (1=Yes) Height Visit Received well-baby visits for checking height/weight of baby (1=Yes) Vaccination Visit Received well-baby visits for vaccination/immunization help (1=Yes) LHW Index Index of LHW Visit, Antenatal Care, Postnatal Care, Height Visit Assigned to Village HH reported that an LHW is assigned to their village (1=Yes) Frequency of Visits (If LHW assigned) Freq. of visits in a month (Recall period: Last 3 months) Satisfaction (If LHW assigned) Satisfied with services/advice provided by LHW (1=Yes) Immunization and Health Outcomes for Children (0-3 years) Incomplete Immunization Child not fully immunized against Polio, BCG, Measles or DPT (1=Yes) Immunization Card Child has an immunization card (1=Yes) Diarrhea Incidence Child had diarrhea in the last 6 months (1=Yes) Stunting Incidence Height of the child indicates stunted linear growth (1=Yes) WASH Outcomes Use Soap Self-report of whether soap is used for washing hands (1=Yes) Saw Soap Enumerator could verify presence of soap in household (1=Yes) Adults Barefoot Adults in HH walk barefoot in the settlement (1=Yes) Children Barefoot Children in the HH walk barefoot in the settlement (1=Yes) 18 Table 3: Illness Incidence (1) (2) (3) Incidence of Illness N. of Consultations Govt. Provider Consulted Treated Village -0.043*** 0.053** 0.026 (0.016) (0.023) (0.023) N 44265 12505 11494 R-squared .0377 .0297 .015 Mean of Dep Var in 0.316 1.040 0.199 Control Villages Note: The symbols *, **, *** represent significance at the 10, 5 and 1 percent respectively. Standard errors are reported in parentheses below the coefficient and are clustered at the vil- lage level. All specifications include social mobilization team effects. Variables are defined in Table 2. Table 4: Utilization of Basic Health Unit (BHU) (1) (2) (3) (4) (5) Wait Time Consultation Fee Convey Concerns Treated Well BHU Index Treated Village -5.821*** 9.407 0.007 -0.010 0.060 (1.756) (20.395) (0.022) (0.025) (0.051) N 1003 1003 1003 1003 1003 R-squared .175 .075 .026 .092 .129 Mean of Dep Var in 20.7 37.9 .960 .934 Control Villages Note: The symbols *, **, *** represent significance at the 10, 5 and 1 percent respectively. Standard errors are reported in parentheses below the coefficient and are clustered at the village level. All speci- fications include social mobilization team effects. Variables are defined in Table 2. 19 Table 5: Maternal Health - Pre Delivery (1) (2) (3) (4) Pregnancy Registered Antenatal Care Pre-Preg Index Treated Village 0.013 0.052* 0.013 0.072 (0.015) (0.030) (0.023) (0.048) N 6109 2762 2762 2762 R-squared .009 .225 .270 .329 Mean of Dep Var in .441 .256 .553 Control Villages Note: The symbols *, **, *** represent significance at the 10, 5 and 1 percent re- spectively. Standard errors are reported in parentheses below the coefficient and are clustered at the village level. All specifications include social mobilization team effects. Variables are defined in Table 2. Table 6: Maternal Health - Post Delivery (1) (2) (3) (4) (5) Postnatal Child Died Birth Weight Post-Preg Care at Birth Registered Recorded Index Treated Village 0.065** 0.008 0.011 0.021 0.097** (0.027) (0.007) (0.019) (0.014) (0.043) N 2762 2762 2626 2626 2626 R-squared .194 .006 .575 .008 .282 Mean of Dep Var in .269 .021 .421 .038 Control Villages Note: The symbols *, **, *** represent significance at the 10, 5 and 1 percent re- spectively. Standard errors are reported in parentheses below the coefficient and are clustered at the village level. All specifications include social mobilization team ef- fects. Variables are defined in Table 2. 20 Table 7: Lady Health Worker (LHW) Health Service Provision (1) (2) (3) (4) (5) Visit Antenatal Care Postnatal Care Height Visit LHW Index Treated Village 0.068* 0.053*** 0.028** 0.046*** 0.139*** (0.038) (0.020) (0.012) (0.014) (0.042) N 2762 2762 2762 2626 2626 R-squared .355 .204 .099 .012 .286 Mean of Dep Var in .353 .142 .034 .041 Control Villages Note: The symbols *, **, *** represent significance at the 10, 5 and 1 percent respectively. Standard errors are reported in parentheses below the coefficient and are clustered at the village level. All specifications include social mobilization team effects. Variables are defined in Table 2. Table 8: LHW Performance and Satisfaction (1) (2) (3) (4) (5) (6) All Women Pregnant in Past 3 Years Assigned Frequency Satisfaction Assigned Frequency Satisfaction Treated Village 0.133** -0.029 0.017 0.150** 0.059 0.029 (0.055) (0.037) (0.033) (0.064) (0.059) (0.043) N 5828 4160 4220 1466 1034 1041 R-squared .142 .012 .065 .173 0.017 0.083 Mean of Dep Var in .621 1.050 .686 .603 1.07 .692 Control Villages Note: The symbols *, **, *** represent significance at the 10, 5 and 1 percent respectively. Stan- dard errors are reported in parentheses below the coefficient and are clustered at the village level. All specifications include social mobilization team effects. Variables are defined in Table 2. 21 Table 9: Immunization Outcomes (Children 3 years and Under) (1) (2) Incomplete Immunization Immunization Card Treated Village -0.038 0.041** (0.030) (0.019) N 4372 4372 R-squared .195 .096 Mean of Dep Var in .354 .106 Control Villages Note: The symbols *, **, *** represent significance at the 10, 5 and 1 percent respectively. Standard errors are reported in parentheses below the coefficient and are clustered at the village level. All specifications in- clude social mobilization team effects. Variables are defined in Table 2. Table 10: Incidence Diarrhea and Nutritional Outcomes (1) (2) Diarrhea Stunting Treated Village -0.015 0.001 (0.026) (0.026) N 4372 1915 R-squared 0.035 .005 Mean of Dep Var in .370 .535 Control Villages Note: The symbols *, **, *** represent significance at the 10, 5 and 1 percent re- spectively. Standard errors are reported in parentheses below the coefficient and are clustered at the village level. All specifi- cations include social mobilization team ef- fects. Variables are defined in Table 2. 22 Table 11: WASH Outcomes (1) (2) (3) (4) Soap Barefoot Walking Use Saw Adults Children Treated Village 0.009 0.033 -0.042 -0.039 (0.014) (0.025) (0.026) (0.030) N 5823 5823 5823 4764 R-squared .041 .111 .091 .054 Mean of Dep Var in .926 .637 .385 .624 Control Villages Note: The symbols *, **, *** represent significance at the 10, 5 and 1 percent respectively. Standard errors are reported in parentheses below the coefficient and are clustered at the vil- lage level. All specifications include social mobilization team effects. Variables are defined in Table 2. 23