Policy Research Working Paper 9220 Strengthening Public Health Systems Policy Ideas from a Governance Perspective Stuti Khemani Sarang Chaudhary Thiago Scot Development Economics Development Research Group April 2020 Policy Research Working Paper 9220 Abstract Public health systems that are capable of disease surveil- state, that supports the insights of economic theory and lance and action to prevent and manage outbreaks require taken together yields ideas that can immediately be put to trustworthy community-embedded public health workers work in policy responses to the COVID-19 crisis. These who are empowered to undertake their tasks as profession- ideas address problems of governance and trust that have als. Economic theory on incentives and norms of agents bedeviled health policymakers. Managing the current and tasked with performing activities that society cares about preventing future pandemics requires going beyond tech- yield direct implications for how to recruit and manage nical health policies to the political institutions that shape frontline health workers to promote trustworthiness and incentives and norms of health workers tasked with imple- professionalism. This paper provides novel evidence from menting those policies. a survey of public health workers in Bihar, India’s poorest This paper is a product of the Development Research Group, Development Economics. 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://www.worldbank.org/prwp. The authors may be contacted at skhemani@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 Strengthening Public Health Systems: Policy Ideas from a Governance Perspective ∗ Stuti Khemani † Sarang Chaudhary ‡ Thiago Scot§ Keywords: politicians, bureaucrats, local government, health services JEL Codes: H1, H4, H75 ∗ We are grateful to James Habyarimana and Irfan Nooruddin for undertaking the survey through Georgetown Uni- versity, and their collaboration in designing it. We thank Shanta Devarajan, Quy-Toan Do, Irfan Nooruddin, Eva Paus, Sergio Schmukler and Adam Wagsta for comments on this draft. We thank Hemant Pawar and the World Bank’s India Country Team for their insights and suggestions. We thank Dr. Sanjay Kumar in the Government of Bihar, and the ADRI team of Dr. Shaibal Gupta, Dr. Prabhat Ghosh, and Dr. Sudhanshu Kumar for comments during the course of this work. We thank the Cicero team for implementing the survey, and especially Dhananjai Joshi and Ankita Barthwal. We thank Nikhil Utture for excellent assistance with all aspects of the survey. We thank CARE’s team in Bihar, especially Dr. Tanmay Mohapatra, for facilitating preliminary eld tests of the survey instruments. We thank the Bill & Melinda Gates Foundation for funding the survey, and Saachi Bhalla, Debarshi Bhattacharya, Santhosh Mathew, Nachiket Mor, Usha Tarigopula and Sandhya Venkateswaran for their previous comments during the survey work. Disclaimer: The views expressed are those of the authors and not necessarily of the World Bank, its Executive Directors, or the gov- ernments they represent. Comments are welcome and can be sent to the following email of the corresponding author: skhemani@worldbank.org. † World Bank, Development Research Group. Corresponding author: skhemani@worldbank.org ‡ World Bank, Development Research Group. schaudhary3@worldbank.org § UC Berkeley Haas School of Business. thiago_scot@berkeley.edu 1 Introduction The global spread of a coronavirus has led to renewed calls for greater investments in public health systems around the world. Policymakers are being advised to increase public spending on health to combat the disease as well as provide a scal stimulus to the economy (Wolf, 2020). A previous disease outbreak, of Ebola, had led researchers to conclude then that public health systems in poor countries need Community Health Workers (CHWs) as the frontline health workers to prevent, detect and manage disease outbreaks (Miller et al, 2018; Lo et al, 2017; Moon et al, 2015). This is the workforce that can ultimately implement any public health policy strategy advised by senior scientists and health professionals. Without the institution of such a workforce the world remains at risk of not being able to e ectively manage and prevent disease outbreaks, especially those that originate in and spread from low capacity, poor countries. This paper provides new ideas for policymakers to invest in frontline public health work- ers using economic theory of "principal-agent" relationships, in which one type of actor, the agent, takes actions on behalf of another, the principal. Insights from economic theory are supported by survey evidence from Bihar, the poorest state of India, a place which is repre- sentative of the challenges involved in building capacity of public health systems in poor en- vironments. These ideas are: to combine steady wage structures, as opposed to high-powered incentives where salaries are conditioned on speci c inputs or outputs of health workers, with management to strengthen peer-to-peer monitoring and professional norms; and political en- gagement of local leaders around public health. These ideas can immediately be put to work in policy responses to the COVID-19 crisis and strengthen public health systems to prevent future pandemics. Policy-makers in poor countries, such as India, have typically relied on voluntary or quasi- voluntary CHWs, rather than contracting CHWs to perform speci ed tasks under a wage structure. Any remuneration that is provided is typically incentivized on indicators of service delivery, such as in the case of the Accredited Social Health Activists (ASHAs) in India who receive payments when they bring pregnant women of their village to deliver babies in public 1 health centers (rather than at home, where institutional assistance for delivery is unavailable). Incentive payments are also supported by a growing body of research which uses randomized control trials (RCTs) to assess the impact of performance-based nancing in the health sector (Ahmed et al, 2019, provide a review). These RCTs have found consistent evidence of positive impact on health facility quality indicators such as cleanliness and availability of medicines and equipment, but mixed evidence on the medical content of care, utilization and health out- comes. However, most RCTs are unable to disentangle how much of the impact is owing to greater nancing becoming available, in poor resource contexts, versus the incentives-based design of providing that nancing. Kandpal et al (2019) provide some of the rst evidence on this question through the evaluation of a World Bank project in Nigeria that enabled lessons to be drawn about how to design the nancing provided to health facilities. The ndings suggest no substantial di erence in impact between the performance-based arm, which had speci c features of high-powered incentives, such as bonus payments to sta , and the gen- eral provision of nancing to be used at the discretion of autonomous, decentralized health facilities. The push in economics research to examine incentives in public service delivery came from the 2004 World Development Report, Making Services Work for Poor People (World Bank, 2004), which documented widespread absenteeism, a clear and measurable indicator of poor performance, among public sector health workers and teachers in public schools, along with more qualitative evidence which suggests that public service providers in poor countries are not motivated to do their jobs well even when they are present at their facilities. Given this sta- tus quo situation of poor performance and weak incentives (for example, in one study, Baner- jee, Du o and Glennerster, 2008, found that health workers face no sanctions when they are absent from work), it is perhaps not surprising that when incentives are credibly strength- ened, such as within an RCT, outcomes are better than the status quo of weak incentives. These research studies also show that scaling-up what "worked" within the context of the study into actual policy (such as, for example, routine application of sanctions for unexcused absences) is di cult and often runs into political problems (Muralidharan and Sundararaman, 2011; Banerjee, Du o and Glennerster, 2008; Dhaliwal and Hanna, 2014). 2 This paper uses a synthesis of economic theory to challenge the prevailing policy regimes that focus on high-powered incentives to deliver basic health services and o ers new ideas that have the potential of working more e ectively to improve public health. First, the paper draws upon Khemani (2019) and World Bank (2016) to summarize the insights from a synthesis of economic theory of principal-agent relationships. Economic theory argues that optimal contracts in complex organizations that are tasked with providing public goods could use low-powered incentives (steady wages, job security), and rely on recruitment of intrinsically motivated workers, as well as professional norms among peers to achieve high performance or productivity. Second, the paper provides evidence from a rich survey in Bihar on current health policies, management practices, and incentives, beliefs and norms of health workers in the public sec- tor. It shows how current management practices and the environment in which health work- ers perform their tasks are the opposite of what economic theory would recommend. Third, the paper discusses how this di erence between economic theory and policy practice can be rationalized and explained. We use available research on institutions to argue that low trust and perverse expectations of behavior in the public sector, for rent-seeking rather than service delivery, characterize institutions in poor places like Bihar. Because of these institutions, reform leaders are reluctant to provide steady wages to new cadres of public health workers because they do not trust that the additional scal burden will yield signi cant results that are valued by the citizens who vote them into, or remove them from, o ce. Fourth, the paper shows how local politics in Bihar (as in other poor places of the world, examined in World Bank, 2016) exhibit potential for change that policymakers could harness to transition to institutions of greater trust, and stronger norms of performance in the public health sector. The paper concludes with outlining a set of policy ideas for reform leaders and international partners to immediately put to use as they respond to the COVID-19 pandemic. Over time, these ideas can contribute to addressing weaknesses in public health systems that has bedev- iled national and global policymakers. Fixing di cult problems like public health requires a 3 complex mix of policy instruments, going beyond health to institutions of governance and trust. This paper aims to bring these complex ideas to the fore at a time when governments are forced to contend with health systems, and international nancing is being mobilized to assist them. Projects to combat the new coronavirus outbreak should use the arguments pro- vided in this paper to inform the design of those projects, and help developing countries move beyond the crisis to establish the public institutions needed for health. 2 Insights from a synthesis of economic theory of principal- agent relationships Public health services–disease surveillance, actions and outreach to prevent and manage outbreaks– involve signi cant positive externalities, in the classic economic sense (Claesen et al, 2003). Delivery of services in one local area or to one population group has bene ts for all, while service failures can lead to global spread of disease, as the world is currently experiencing with a new coronavirus. These positive externalities constitute a case for public spending on public health to address the theoretical under-investment by private actors and markets alone. Furthermore, a case can be made for direct government provision of public health services, rather than through subsidies to private providers, because of the need for systems that stay vigilant in normal times, in the absence of any crisis or visible health need, and for the legiti- macy to regulate citizen behaviors in the broader public interest. The power of regulation of individual behavior in the public interest resides with the state. All countries thus have some form of a state agency tasked with maintaining public health goals and enforcing health laws. Economic theory lends itself to understanding how such state public health agencies should be organized, how health workers should be recruited and managed to e ectively deliver the services needed for detecting, preventing, and mitigating disease outbreaks. The health pro- fession has weighed-in on these questions in the aftermath of the Ebola outbreak in West Africa, arguing for public investments in community health workers (CHWs) who are mem- bers of the communities they serve. From the economic perspective, the value of community- embedded public health workers can be traced to their informational advantage. As mem- 4 bers of the community they are likely to have greater information about who is falling sick, whether there are suspicious clusters of disease, whether citizens are complying with health regulations, etcetera, than temporary visits by health supervisors from outside the commu- nity. As members of the community, CHWs may also have greater communication capacity to win trust and compliance with public health regulations. The trade-o with these intuitive advantages of greater local information and communication capacity is the scal burden of keeping such health workers on the public payroll. For example, critics of bureaucracies in the United States argue that federal workers are overpaid and underworked (Johnson and Libecap, 1994). Countries have not invested in keeping a regular cadre of CHWs on the public payroll probably because of competing needs for public funding and lower public valuation of a task focused on prevention–the bene ts of surveillance and public health regulations typ- ically do not make headlines, and are not visible (Mani and Mukand, 2007), while the costs of complying with regulations are all too palpable. Furthermore, in poor countries where workers on the public payroll are viewed as having weak incentives and motivation to per- form the tasks assigned to them, even reform oriented politicians are reluctant to spend scarce public resources on hiring them. Evidence of widespread absenteeism among public sector doctors and teachers, for example, provides clear indicators of weak incentives and motivation (Chaudhury et al, 2006). Weak incentives and motivation are arguably the bigger problem (Filmer and Wagsta , 2020)1 The scal burden can be reduced, or the "value for money" by keeping CHWs on the pub- lic payroll can be increased by tasking CHWs with additional health functions, such as ma- ternal and child health, and expanding coverage of primary health to those who cannot af- 1 Filmer and Wagsta (2020) provide a recent review of the problem of quality in the delivery of health and education services. They also report emerging insights from ongoing meta-analysis of the impact of di erent pay-for-performance schemes on a variety of health indicators. They conclude that much more research is needed on the impact of di erent "blended" schemes that combine steady budgets to health or education facilities, or steady salaries to individual providers, with elements of pay-for-performance. In this paper we focus on the theory and evidence for remunerating community health workers who get tasked with di erent public health goals by senior policymakers. As we show below, the "blend" we argue for is to combine steady wages to CHWs tasked with public health services with management practices in health bureaucracies and engagement of local political leaders. We discuss how economic and game theory provide insights for how communication can be targeted at bureaucratic and political institutions to strengthen peer-to-peer professional norms, and local polictical support for, rather than hindrance of health workers. We too call for more research to evaluate impact, and experiment with di erent policy designs, as policymakers take-up the ideas proposed here. 5 ford private providers. For example, medical research has found that neonatal care can be improved and childhood disease mortality can be reduced in poor resource environments through community-based interventions (Bang et al, 1990 and 1999). These health bene ts can improve economic outcomes, and thus scal space in the future, through the channel of greater human capital (Flabbi and Gatti, 2018). But, the health, human capital and eco- nomic bene ts of community-based interventions can only materialize if the frontline health workers are motivated to perform their tasks well. Economic theory provides policy ideas to improve the performance of frontline health workers, which can reduce the trade-o s and increase the payo s from greater public spending on public health workers. The performance of frontline public health workers can be examined within a series of in- terdependent “principal-agent” problems in which one type of actor, the agent, takes actions on behalf of, or at the behest of another, the principal. Public health policies are selected and implemented by government agencies within the following principal-agent relationships illustrated in Figure 1: (i) between citizens and political leaders, (ii) between political lead- ers and public o cials who lead government agencies, and (iii) between public o cials and frontline providers. This gure also shows how popular development initiatives of citizen engagement, to monitor frontline providers and participate in service delivery, t into this framework. To illustrate in the health policy context of India, there are at least three cadres of workers who are expected to live and work in the communities assigned to them to serve: the Auxiliary Nurse Midwives (ANMs) who are supposed to provide primary and maternal health services at the lowest tier of public clinics, the health sub-center; the Accredited Social Health Ac- tivists (ASHAs) who could provide a range of services and are currently incentivized to shift home-based child deliveries to medical institutions; and the Aanganwadi Workers (AWWs), who are responsible for delivering early childhood development and nutrition programs that have direct implications for child health. The ANMs, ASHAs and AWWs together organize immunization drives and vaccinate children in rural India, for example. ANMs and ASHAs are hired and managed by the Department of Health of state governments in India, and im- 6 mediately supervised by the Medical O cer in Charge of Primary Health Centers at the level of blocks, the lowest tier of state administration above the village or community. AWWs are hired and managed by the Department of Women and Child Development. Health policy in India mandates the constitution of a committee, the Village Health Sanitation and Nutrition Committees (VHSNC) through which citizens can participate in monitoring and contributing to health and nutrition services. The VHSNC is presided by the directly-elected head of local government. Local politicians can thus play a role in monitoring and managing health work- ers, even though the formal powers reside within the bureaucracy of state governments who ultimately report to the state-level executive, the Chief Minister and her cabinet of ministers under a parliamentary form of government. Elections to gain power over the state govern- ment are vigorously contested in India, with frequent turnover of political parties in o ce (Nooruddin and Chhibber, 2008; Nooruddin and Simmons, 2015), thus completing the chain to citizens and society as the ultimate principals illustrated in Figure 1.2 Frontline health workers are thus tasked with multiple and complex functions that could be expanded as policymakers strive to prepare for more global outbreaks of new diseases. In fact, India is mobilizing these workers to deploy non-pharmaceutical interventions (NPIs) in its rst line of defense against the new coronavirus (The Hindu, March 31, 2020)3 . While other complex organizations share the problem of motivating agents to perform multiple tasks that are di cult to monitor, agents in the public sector have more unique roles of serving the public interest. Public health tasks and functions, like disease surveillance, testing, enforc- ing compliance with NPI policies, have precisely these characteristics of serving the public good. Furthermore, in normal times, these frontline health workers serve the children and women whose families’ capacity is limited by budget and credit constraints, in addition to any behavioral constraints imposed by poverty, lack of education, and social discrimination. In addition to the "public good" nature of tasks that health workers are required to undertake, 2 The governance of CHWs in Africa can similarly be framed in this interdependent principal-agent frame- work. A detailed analysis of these relationships in Uganda is provided in Habyarimana, Khemani and Scot (2018). The case of Brazil, examined in Rocha and Soares (2010) is described further below. 3 India’s Ministry of Health announced online training of ANMs, ASHAs and AWWs, along with nurses and doctors, for the implementation of activities to combat the pandemic. https://www.thehindu.com/news/national/coronavirus-panel-formed-for-co-ordinated-research-and- development-activity/article31221310.ece 7 the structure of principal-agent relationships in the public sector is also quite di erent from those in other complex organizations outside government. Namely, the presence of multiple principals with potentially con icting interests, such as citizens belonging to di erent socio- economic groups with di erent preferences and attitudes to the public sector, local politicians, medical o cers, district and state bureaucrats, state politicians. This is a particular feature of public sector organizations, or government bureaucracies, that is examined in economic principal-agent theory (Dixit, 2002). Two insights emerge from the literature about how to structure principal agent relationships in government bureaucracies when tasks are multiple and complex, and involve serving the public interest4 : 1. Reduced role of high-powered incentives and greater role for recruiting in- trinsically motivated agents 2. Reduced role for top-down hierarchical monitoring and greater role for autonomy and peer-to-peer professional norms In practice, bureaucracies across the world tend to use at and above-market wages, presum- ably to attract public service motivated and talented workers (Finan, Olken and Pande, 2015). In many striking cases, these arrangements also “work”. For example, the success of one of the highest performing education systems in the world, the Finnish public education system, has been attributed to the meritocratic recruitment of highly trained teachers, imbued with strong professional norms, and autonomy in their classrooms (World Bank, 2018). Incentives are also strong in that teacher salaries are high to be able to attract highly competent indi- viduals into the profession, and teachers can be let go by school administrators (who also exercise autonomy in how they manage schools). But incentives are not high powered in that salary structures are at rather than consisting of bonus components contingent on test scores of students. The Republic of Korea’s high-performing education system shares with Finland these characteristics of the management of public school teachers (World Bank, 2018)5 Greater autonomy in public agencies is also found to be robustly associated with better outcomes in the delivery of public investment projects in developing countries (Rasul and Rogger, 2017).6 4 Following are some of the pioneering contributions: Tirole, 1994; Dewatripont et al., 1999; Francois, 2000; Dixit, 2002; Besley and Ghatak, 2005; Acemoglu et al., 2008; Alesina and Tabellini, 2007, 2008. 5 Even though the Korean and Finnish systems diverge in their pedagogical approach, they are strikingly similar when it comes to management of teachers in public education bureaucracies. 6 Peer-to-peer monitoring and social interaction also matters in private rms (Ashraf and Bandiera, 2018, 8 Yet, a body of research on public service delivery in poor countries cautions against trusting the intrinsic motivation of workers in the public sector, providing evidence of poor perfor- mance, and perverse incentives and motivation (for rent-seeking rather than service delivery). For example, one study found that doctors in India systematically under-perform in the public sector compared to in their own private practice (Das et al., 2016). Other qualitative research suggests that the problem of poor performance is likely to be widespread because bribery and corruption are the norm in human resource management in the public health sector, weak- ening incentives for good performance (La Forgia et al., 2015). Furthermore, the available research suggests that powerful leaders at upper levels of the government hierarchy, who wield formal power over the humble workers on the frontlines of the state, can be thwarted in their attempts to exact accountability and performance from them. Banerjee, Du o and Glennerster (2008) and Dhaliwal and Hanna (2014) provide evidence that reformers who tried to use new technology to monitor frontline health workers and strengthen their incentives ultimately failed to implement or sustain these reforms. The available research has focused on documenting evidence of weak incentives and low ac- countability for service delivery in the public sector, and thence on evaluating interventions targeted at strengthening incentives, such as making some part of pay conditional on perfor- mance indicators (for example, Singh and Masters, 2017, for CHWs). But what is available is barely scratching the surface of knowledge needed to help reform leaders think about how to structure government bureaucracies and assign tasks to leverage intrinsic motivation and to reduce reliance on high-powered incentives. Even when increasing the power of incentives has been shown to “work”, the authors of those ndings concede that implementing optimal incentive contracts at scale can place signi cant demands on state capacity (Muralidharan and Sundararaman, 2011). Finally, the logic of economic theory, and growing international evidence in support of it, further suggests that politics casts a long shadow on culture in the bureaucracy and at the frontlines (World Bank, 2004; Devarajan et al, 2015; World Bank, 2016; Khemani, 2019). For provide a review). For example, Kandel and Lazear (1992) showed how di erences in the observed management policies of American versus Japanese rms can be interpreted in economic theory as arising from di erential degrees of team-work norms and social pressure among peers. 9 example, research has found that e ective enfranchisement of poor voters has direct conse- quences for the delivery of health services and improved health outcomes, while poor voters’ disenfranchisement, through vote buying, has the opposite e ects (Fujiwara, 2015 and Khe- mani, 2015). But these ideas about the role of politics and exploration of its policy implications has not happened through RCTs for the simple reason that the suggested interventions are more complicated than what is feasible in most RCTs. The interventions involve not only changes in contract design (level of wages, whether high powered or not) which other RCTs have undertaken (Dal Bo et al, 2013; Ashraf et al, 2014; Ashraf et al, 2015), but crucially, these contract changes need to be combined with political and bureaucratic reforms if the insights of economic theory are to be fully tested and exploited. Collaboration with policy-makers who are willing to experiment with policy is hard enough for most RCTs, and compounded in this case by the complexity and potential sensitivity of targeting bureaucratic management culture and political contestation. In the absence of opportunity to directly undertake a policy experiment suggested by eco- nomic theory, and evaluate its impact to the standards of econometric rigor of other RCTs, the research in Bihar we report on in this paper gathered descriptive data, through rich sur- veys using innovative modules to measure the variables that economic theory suggests are important (such as, intrinsic motivation, and peer norms). This survey provides evidence of the mis-match between actual policy practice and the insights from economic theory and o ers ideas to policy makers to adopt a di erent approach to strengthening health systems using the current crisis as an opportunity. 3 Description of the survey in Bihar, India The survey was undertaken in the state of Bihar in India between November 2018 and March 2019.7 Data were collected across di erent layers of government jurisdictions within the state–districts, blocks and village governments, known as Gram Panchayats (GPs)–and across di erent types of respondents–politicians, bureaucrats or public o cials, frontline service 7 Data were collected in two phases– one at the village level, and one above the village at blocks and districts 10 providers–who share interdependent relationships while undertaking their tasks of deliver- ing public health and nutrition services. In the rst phase, data were gathered from village- level respondents–Gram Panchayat (GP) politicians, frontline health and nutrition workers (ASHAs, AWWs, and ANMs at health sub-centers), citizens and leaders of women’s Self Help Groups (SHGs), an important social institution in this context. In the second phase, the survey was implemented to block and district-level respondents. 3.1 Geography of the survey Budget and implementation constraints required us to select a sample of districts rather than covering all 38 districts of Bihar. At the same time, we needed a large sample to be represen- tative of the diversity within the state, and allow us to capture some variation across district- level institutional characteristics. These constraints led us to determine 16 as the number of districts in which to undertake the survey. The purposive selection of which 16 study districts, from among the 38 of Bihar is described in the sampling materials posted at the following web- site, along with survey questionnaires and all results currently available at this stage of our data analysis: https://sites.google.com/view/stutikhemani/public-health-systems-survey Within each of the 16 districts, 4 blocks were selected using a random number generator, after stratifying by proximity to the main railway line. Within each block, 4 Gram Panchayats (GPs) were selected using a random number generator. However, in one block each in the districts of Lakhisarai and Buxar, 3 GPs instead of 4 were selected because the sampling protocol required a su cient number of replacement respondents to be available, and these districts only had 3 GPs ful lling the replacement requirement (more details in section on Respondents below). This yields a sample of respondents drawn from 16 districts, 64 blocks from within those districts, and 254 Gram Panchayats (GPs) from within those blocks. The sampled districts and blocks are shown in Figure 2. 11 3.2 Respondents of the survey Following the framework in Figure 1, data were gathered from the following types of respon- dents: 1. Citizens: The citizen survey was aimed at respondents from 16 households residing in each GP area. The survey rm was provided with a list of respondents (with replacements) drawn randomly from the electoral rolls available of all voting-age adults in Bihar’s population. The target sample size is thus 4064 citizens (16 each from 254 GPs). Within the category of citizens, the survey additionally targeted o ce-bearing members of women’s Self Help Groups (SHG) under a rural livelihoods program in Bihar known as Jeevika. However, we had no lists available with names of SHG leaders of the village- level organizations across GPs. In the absence of these lists, we relied on the survey rm to ensure that enumerator teams would identify SHG leaders during their eld-work. The data from SHG leaders that have been provided to us is thus subject to a greater than usual caveat: the risk of whether the enumerator teams accurately identi ed and obtained interviews with the targeted SHG respondents. The instructions provided to the survey teams was to ask the GP Mukhiya and other GP-level respondents (such as the ANM, ASHA and AWW) about the GP-level federated organization of all the SHGs across the GP’s communities to identify its President, Secretary and Treasurer. That is, 3 SHG leaders were targeted for each GP, for a total sample of 762 (3 each from 254 GPs) SHG leaders. 2. Politicians: • Village level: – GP Mukhiya (or head of the elected village council; 1 per GP) – Elected village councilors or Ward members: 3 per GP – Candidates who contested the Mukhiya position in the last election of 2016, 12 but lost: 3 contenders per GP Lists were provided to the survey teams of all incumbent Mukhiyas to be inter- viewed, and a random selection (with replacement) of 3 Ward members and 3 can- didates from among those who contested the previous GP elections of 2016.8 The targeted sample size of GP politicians is thus 1778 (7 each from 254 GPs) respon- dents. • Block level: – Elected head of block-level government (Panchayat Samiti Chairperson) – Elected member of Panchayat Samiti: targeted to be the member who is on a committee related to public health – Member of Legislative Assembly (MLA, Bihar state) elected from the MLA constituencies in our study area: approximately 1 per block (57 MLAs across the 64 blocks of the study area) The survey rm was responsible for identifying the block-level politicians targeted to be interviewed. The targeted sample size of Block-Panchayat (Panchayat Samiti) elected members’ is 128 respondents (2 each from 64 blocks). The 57 MLAs across the 64 blocks of the study area were also identi ed by the survey rm. However, because of problems of reaching politicians at a time that was close to the 2019 elections in India, the survey rm was able to complete interviews with only 39 MLAs (of the targeted 57) , and with 119 Panchayat Samiti members (of the targeted 128). • District level: – Member of Parliament (MP, national-level) elected from the MP constituencies in our study area: approximately 1 per district – Elected head of district-level government (Zilla Parishad Chairperson) 8 Across Bihar, on average, 12 candidates per GP contested the Mukhiya elections in 2016. 13 – Elected member of Zilla Parishad: targeted to be the member who is on a committee related to public health The survey rm was responsible for identifying the MPs from constituencies within the 16 study districts, and the 32 respondents of the District-Panchayat (Zilla Parishad). Again, because of problems reaching political leaders at election time, the survey rm was able to interview only 9 MPs, and 28 Zilla Parishad members. 3. Bureaucrats: The following district and block-level positions were identi ed which have supervision and management powers over frontline health service providers. • Block level – Block Medical O cer (typically known as MOIC– Medical O cer in Charge of the Block-level Primary Health Center) – Block Programme Manager of the National Health Mission (NHM) – Block Programme O cer of Reproductive and Child Health (RCH) and Im- munization – Block Community Mobilizer (Block-level supervisor of ASHAs) • District level – Civil Surgeon or Chief Medical O cer (CMO) – Additional (like a deputy) Chief Medical O cer (ACMO) – District Programme Manager of NHM – District RCH and Immunization In-Charge – District Community Mobilizer (District-level supervisor of ASHAs) The survey rm was responsible for identifying and interviewing the respondents hold- ing these positions. The nal data submitted by the survey rm contains 293 respon- 14 dents in supervisory or management positions, including: 13 Civil Surgeons, 11 Chief Medical O cers (including 4 who were in Acting capacity), 23 Superintendents (in- cluding 13 in Deputy or Acting capacity), 9 District Programme O cers-NHM, 4 Dis- trict RCH and Immunization In-charge, 7 District Community Mobilizers, 58 MOICs, 58 Acting Facility Incharge, 43 Block Program Managers-NHM, 29 Block RCH Programme o cers, and 35 Block Community Mobilizers. 4. Public providers of health services: • Village level: – ANM at village health sub-center – ASHA – AWW The survey team was provided a list (with replacements) of 3 AWW workers to interveiw per GP, for a targeted sample of 762 AWW respondents. We did not have population lists of ASHAs and ANMs. Hence, the survey team was instructed to ask the AWW respondents to identify the ASHAs and ANMs in their communities within the GP. The survey teams were to pick 3 ASHAs and all available ANMs to interview in a GP. We describe the shortfalls from these targeted numbers in the data description further below. • Block level: – Doctors at Primary Health Centers (PHC) – Nurses at PHCs – ANMs at PHC • District-level: – Doctors at District Hospitals or equivalent 15 – Nurses at District Hospitals or equivalent The survey team was provided with a list of randomly selected candidates for the above categories of respondents for all the PHCs and higher-level health facilities (such as District Hospitals) across the 64 blocks of the study area. However, the survey team reports substantial di culty in adhering to this list because the personnel were not found at the health facilities. We describe the numbers of respondents that the survey team was able to reach in the sections below, and highlight the fact that we were not able to reach a random sample of providers appointed at these positions. 3.3 Description of the questionnaire The questionnaires were designed to understand whether the characteristics or traits (such as, public service motivation and integrity) of frontline health workers in the public health system, and how they are managed by their supervisors within the bureaucracy, correspond to the insights of economic theory discussed above. All of the questionnaires are posted at https://sites.google.com/view/stutikhemani/public-health-systems-survey. Below we brie y summarize some of the novel aspects of the data. The concepts of intrinsic motivation and social pressure among peers are at the heart of eco- nomic theory of managing complex organizations tasked with delivering public goods, but are very di cult to measure. The di culties arise not only because they are intangible and subjective concepts which makes it hard for researchers to design the questions that would go into a survey, but also because it is hard for respondents to understand and answer the questions. In our extensive eld-testing of these survey modules in Bihar we found that the abstract nature of the questions made it di cult for respondents to grasp what we were ask- ing. For example, the module to measure "integrity" comes from literature in psychology and aims to assess the extent of moral disengagement (Moore et al, 2012). In eld testing these questions in rural Bihar, we found that respondents struggled to understand what we were asking. The very rst question of this module, for example, asks respondents to indicate the extent to which they agree or disagree with the statement "It is okay to spread rumors to de- 16 fend those you care about." In Bihar, respondents repeatedly asked for concrete examples to understand this question, leading us down a path where the module might di er signi cantly from how it is administered in other countries and contexts. Furthermore, in any country or cultural context in which similar questions have been tried, respondents tend to exhibit what is termed as "social desirability bias"– that is, to answer some of the questions in ways which they think would win social approbation even if their true responses would be otherwise. For example, a question like "are you very careful or attentive when undertaking a task?" has been used to measure the personality trait of con- scientiousness, but getting sincere responses to it is a problem, as evidenced by 100 percent of some respondents in our survey saying they are always careful. As we analyzed the data, we assessed whether/which questions worked as intended, to capture meaningful variation across respondents (even though the "levels" are subject to desirability bias). We report all our ndings so that readers can reach their own interpretation and argue against the ones we o er. Part of the contribution of this work is methodological– nding better ways to measure these concepts, that may be crucial for strengthening public health systems, through trial and error. We follow an established practice in the growing literature measuring public service mo- tivation and integrity of grouping di erent sets of questions into indices. The underlying questions in each of the indices created are summarized below.9 In this paper we focus on the following traits: • All Surveys: 1. Integrity Index (8 questions): Questions on agreement or disagreement with morally disengaged behavior (e.g. it’s OK to spread rumors to defend those you care about or it’s no big deal to pass someone else’s work as your own). 9 All indices are created using Inverse Covariance Weighting (ICW) as described in Michael Anderson. "Mul- tiple Inference and Gender Di erences in the E ects of Early Intervention: A Reevaluation of the Abecedarian, Perry Preschool and Early Training Projects. Journal of the American Statistical Association. December 2008, Vol. 103, No. 484. This is similar to Principal Component Analysis (PCA), but overweights variables that pro- vide "new information" (i.e. have lower covariance with other variables). All underlying data is coded such that higher value of indices are "more positive" outcomes. 17 2. Public Sector Motivation index (8 questions): Questions on commitment to serve the public, capacity to resolve con ict among people, faith in government’s role in improving society. We developed a new module to understand the attitudes of health workers towards their work, and group these questions into an index of professional identity and e cacy: • Public providers’ surveys: Professional identity and e cacy (5 items): whether respondent would have pre- ferred another profession; whether feels a sense of inner pride/ful llment when doing their work; whether gets social status from being in the profession; whether feel they can improve outcomes through their e ort; whether they have to take permission for every little thing. Finally, we asked frontline health workers and their supervisors within the bureaucracy about management culture. During the eld work to develop the survey instruments, we learnt that frontline health workers attend regular meetings with their supervisors at the block level. We thus targeted the questions to understand management culture by speci cally referencing the types of issues that come-up in these regular meetings with supervisors. We start by reporting these results. 4 Management and professional culture Health workers in Bihar attend regular meetings with their supervisors, which provides an opportunity to measure management and professional culture through survey questions tar- geted at the content of these meetings. Among the lowest tier health workers–the ASHAs–94 percent report meeting with the ANM of the village health center, their immediate supervisor, and 85 percent report meetings organized at the next level, by the Medical O cer In-Charge (MOIC) of the Block Primary Health Center, at least monthly. Similarly, 95 percent of ANMs report regular meetings (at least monthly) convened by their supervisor (block-level MOIC). In the survey of supervisors (Medical O cers in Charge, Civil Surgeons, various programme 18 o cers at the district and block levels, including those who supervise the AWWs under the Women and Child Welfare department), 98 percent respond that they hold meetings at least monthly with the sta who work under their supervision. Figure 3 shows that high rates of respondents across health cadres report that these meetings involve "scoldings" and discussion of bad performance. Among the lowest tier of workers, the ASHAs and AWWs, more than 77 percent report that meetings with supervisors involve scold- ings. These numbers are even higher for the ANMs, more than 90 percent of whom report that meetings consist of scoldings. In response to a module of questions asked about meet- ings with their peers and supervisors, the ANMs report that issues about bad performance are often raised– 48 percent respond "always", and 44 percent "sometimes"; only 8 percent say "rarely" or "never". In contrast, when it comes to the question of how often good work is praised or recognized, 47 percent of the ANMs respond "rarely" or "never". As many as 30 percent of ANMs say that their supervisors always scold people at meetings for bad perfor- mance. Similar patterns are reported by higher-level cadres of health personnel interviewed at block PHCs and district hospitals– there is bunching of respondents at "always" when asked about discussion of bad performance and "scoldings" at meetings, and at "never" when asked whether good work gets recognized or praised. A consistent picture is provided by supervisor respondents at the block and district levels, as shown in Figure 4. Among these supervisors, 76 percent report that bad performance is always discussed at these meetings. The dominance of concerns with poor performance at management meetings would be con- sistent with the research evidence of high absenteeism and low performance among public health workers in India (discussed and cited in the sections above). The problem of absen- teeism, in particular, is something the survey teams encountered when attempting to inter- view the ANMs who are supposed to be located at the village-level health sub-centers. Among the 254 targeted GPs, 226 health sub-centers were found, and out of these, 145 sub-centers (or 64 percent) were closed. The survey teams were able to interview ANMs at the 81 open sub- centers and located 28 others, resulting in an available sample of 109 ANMs, less than half of the expected number. The problem of lack of availability of ANMs at the health sub-centers 19 is corroborated in responses by citizens and local politicians. When asked whether the ANM is usually available at the health sub-center, 73 percent of both citizens and GP politicians re- sponded "sometimes", "rarely", or "never". Our survey thus con rms what was widely reported during our eldwork in Bihar–that village-level health sub-centers are often dysfunctional, and ANMs are not usually available.10 In contrast to the responses of citizens and local politi- cians, when the ASHAs and AWWs were asked this question about the availability of ANMs, only 35 percent responded "sometimes", "rarely" or "never". This pattern is consistent with reluctance on the part of health workers to speak ill of their peers. Other survey questions targeted at asking health workers about their view of peers is likely to su er from this same reluctance to answer the question objectively. One such question posed to ANMs was to think about all the ANMs they knew or had worked with and tell us out of those how many were hardworking, and how many were honest. Among the ANMs who answered these questions, 80 percent indicated that all (100 percent) of the ANMs they know are hardworking, and 88 percent indicated that all (100 percent) are honest. While the regular "scoldings" and discussion of problems of poor performance in management meetings is consistent with the evidence of lack of availability of ANMs on the job, it also suggests a vicious cycle of low expectations sustained by lack of trust among colleagues and peers. This can feed a lack of professional motivation among health workers as we nd in responses to another module of questions on whether ANMs feel attached to their profession, and a sense of e cacy. Half of the ANMs respond that they agree it would have been better if they had taken-up another profession (in contrast, only 16 percent of doctors respond that they would have taken-up another profession). Lack of professional agency or e cacy is particularly demonstrated in responses to the fol- lowing question: "Irrespective of my e orts, the system will not allow health outcomes to im- prove". Figure 5 shows that across health cadres, from the village level up to doctors in public hospitals and health supervisors, more than 70 percent agree with this statement. Lack of professional authority or discretion is evident in Figure 6, showing high rates of respondents 10 Interestingly, in this regard, when comparing responses of citizens, ASHAs, AWWs and politicians at the GP level on the modules to measure norms and motivation, the ANM respondents stand-out as having the lowest measures on Integrity and Public Service Motivation, as reported further below. 20 across health cadres agreeing with the statement that "In my work, I have to take permission for every little thing." Supervisor responses are a useful source for understanding the extent to which the public health system su ers from low-e ort norms. As mentioned earlier, questions posed to health workers about their expectations of their peers–whether they think their peers are hardwork- ing or honest–did not work well in eliciting objective views because respondents are reluctant to speak ill of their peers. Indeed, even supervisors appear reluctant to report any problems: 80 percent of supervisor respondents say that 100 percent of the various cadres of health sta are both hard-working and honest. But, when we restrict attention to the 20 percent of super- visor respondents whose answers are less than 100 percent, these supervisors are reporting quite high rates of dishonesty and shirking among health workers. The average reported shares of dishonest/shirking workers when we only look at the data from those supervisors are: 36 percent of ANMs, 44 percent of Nurses, and 47 percent of Doctors.11 Health supervisors answer other pertinent questions in ways that suggest systemic problems: 44 percent say that good workers get transferred because others feel threatened by them; 80 percent say that irrespective of their e orts, the system will not allow people’s health to improve; 67 percent say they have to get permission for every little thing. Health supervi- sors’ responses to these three questions related to professional e cacy are thus similar to the responses of health workers across the cadres, from the GP-level up. Grouping the set of questions aimed at measuring attachment to the profession, and sense of e cacy into an index, we examine variation across frontline health workers. Table 1 reports correlates of this variation. To highlight a few interesting correlations: one, non-receipt of salary payments is negatively correlated with the index of professional identity and e cacy, and this correlation is statisti- cally signi cant for ANMs but not for ASHAs/AWWs.12 Two, those who report few opportu- 11 The ANMs here include those posted at PHCs, who may be performing signi cantly better than the GP-level ANMs). 12 The average ASHA and AWW reported that in the past year, they had not received the payments due to them for six months of work. The average ANM reported that salary for three months of work had not been received. That is, the ANM reports lower delays in receiving renumeration compared to the lower tier community health 21 nities for interaction with peers during the regular meetings held at the block-level Primary Health Center also report weaker professional identity and e cacy. Three, those who report that informal payments (bribes) are important to get desired transfers and promotions are also likely to report weaker professional identity and e cacy. Four, those who report that politicians create di culties for their work are also likely to report lower e cacy or weaker norms. All of these correlations are reasonably intuitive, and consistent with the importance of management practices and political environment in shaping professional attitudes among health workers. Taken together, the pattern of survey responses across di erent cadres of health workers and their supervisors suggests that the public health system is stuck in an equilibrium of low ex- pectations, low trust, and low performance. Across the board, health workers express a sense of professional ine cacy, that no matter how hard they try, the system will not allow health outcomes to improve. Health workers think they are not recognized and empowered to per- form; those who have any management power or authority think that health workers need to be scolded and handled strongly to get them to perform. This equilibrium stands in stark contrast to the recommendations that emerge from the logic of economic theory about how to organize institutions that are tasked with delivering services with "public good" charac- teristics. The next section discusses how such an equilibrium is consistent with a broader literature examining the role of institutions in development. 5 Institutions of low trust and perverse expectations of behavior in the public sector Available research suggests that low trust and perverse expectations of behavior in the public sector, for rent-seeking rather than service delivery, characterize institutions in poor places like Bihar. Scholars of Bihar argue that the Indian caste system runs deep in the state, which together with a historical legacy of feudal systems of land ownership creates persistent social divisions and lack of trust across groups (Gupta, 1981). workers. 22 Bihar’s institutions can also be understood within a broader macroeconomic story of what explains why some countries are rich and others persistently poor. One body of work has argued that European colonization established "extractive" institutions in the history of cur- rently poor countries, and those historical institutions have persistent e ects till today, ex- plaining why some places are not able to su ciently grow out of poverty or develop "inclu- sive" political institutions that assure health and education for all (Acemoglu, Johnson and Robinson, 2001 and 2002; ). These ideas are tested by Banerjee and Iyer (2005) and Iyer (2010) by examining variation in institutions within India. They nd that colonial institutions have persistent e ects on agricultural productivity and the delivery of public services. Bihar hap- pens to be the state where the historical institutions used by Banerjee and Iyer (2005) and Iyer (2010) show no within-state variation. All of Bihar was governed directly by the British and through feudal land institutions during the colonial era, which are the institutions these authors nd correlated with low agricultural productivity and fewer public good provision in current times (respectively), long after independence and land reforms. However, the mechanisms of persistent impact remain shrouded in mystery. Research is also lacking on why current formal institutions, such as competitive electoral institutions, courts, police, which most states, including Bihar, possess, are not su cient to overcome the burden of history. Further, these explanations may be unsatisfying because they are so abstract, and provide such little guidance of ways out of the problem. What precisely are "extractive" in- stitutions, and why is to so hard to reverse them after the colonial powers are removed? It is di cult to argue that Bihar’s political institutions are not "inclusive", after decades of compet- itive elections, turnover of political parties in o ce, and rise of lower-caste political leaders, on platforms of empowering the lower castes. Indeed, our survey data are consistent with inclusive politics, uncovering quite astonishing rates of political aspiration, and sophisticated political thinking among citizens in rural Bihar, across caste groups (discussed in the section below). Taking stock of the literature on institutions, two World Bank reports have argued that the reason historical institutions have persistent e ects is because they work through norms– 23 beliefs about how others are behaving– which are slow to change (World Bank, 2016; World Bank, 2017). Political norms are de ned as a subset of social norms, pertaining to beliefs about how others are behaving in the sphere of politics and government. Political norms of behav- ior — what citizens demand from the state, and how they expect others to be acting in the public sector — can explain why well-intentioned reformers, in powerful political positions, nd it di cult to institute change. To illustrate the argument13 , prevalent political norms can be characterized as follows. Rational expectation among “ordinary” (i.e. non-o ce-bearing) citizens is that other citizens will vote for politicians who share their identity or ideology, and who provide targeted private bene ts, even though, in equilibrium, the consequences are harmful for everybody (since voting on the basis of identity and private bene ts weakens po- litical incentives to provide public goods). Of these ordinary citizens, among those who have leadership or entrepreneurial qualities, who become contenders for political power (starting from the village level, as local elections spread within countries across the political spectrum), the rational expectation is that other contenders are entering the fray to seek private rents from public resources. Among o ce-bearing citizens, or state personnel, all the way from high- and mid-level bureaucrats to frontline providers, the rational expectation is that their peers do not care about doing their jobs well because there are few formal or informal (sanc- tions) rewards for (bad) good performance. The service delivery organizations of the state thus lack both incentives and non-pecuniary sources of motivation (such as peer pressure and professional norms). When a reform leader tries suddenly to strengthen incentives in this low- performance setting, those reforms are resisted and thwarted by well-organized in- terest groups, such as unions of teachers and doctors, or other political opponents who seek rents from the status quo. Where even dynamic and motivated leaders nd it hard to bring reforms is where success is di cult to deliver using their executive powers alone, and therefore, also harder for them to take credit for and return to o ce. These di cult areas are human-resource intensive and require sustained change in the day-to-day behavior of a large number of frontline person- nel. It is also precisely in these human-personnel-intensive sectors of delivery that leaders 13 This section draws on Khemani (2019). 24 face a particularly di cult political trade-o — win elections by extending the patronage of government jobs and deploying ideological instruments (such as by exploiting social animus between groups on the basis of religion or ethnicity) versus strengthening bureaucratic insti- tutions so that service providers are professionally motivated and routinely held accountable without need for political intervention. Thachil (2011) and Chidambaram (2012), for example, nd that political parties with religious ideologies respond to citizen demand for health and education by providing services through their party organizations outside of the state bureau- cracy, in order to win citizens’ gratitude and their vote; but, these parties do not appear to have su cient incentives once in o ce to institutionalize service delivery by de-linking it from patronage or ideological politics. In the language of game theory, persistent and systemic problems of low e ort and low per- formance in government agencies can be understood as the non-cooperative equilibrium of a Prisoner’s Dilemma, supported by low expectations for cooperation (that is, non-cooperative norms) among large numbers of players. Dixit (2018) describes this problem in the context of widespread corruption. Although it would be bene cial for society as a whole to reduce corruption, society is instead stuck at high corruption levels because individuals believe that engaging in corruption is the best they can do given how others are behaving. For example, bureaucrats ask for bribes in order to provide public services, and citizens pay these bribes because they believe that most others engage in bribery; if they refuse to pay the bribe, they will get nothing, or worse, they may su er retribution. The sections below present evidence of forces of change that may be harnessed by policy- makers to try to make public health spending more e ective in strengthening public health. 6 Potential for harnessing local political contestation Patterns in the data suggest that citizen demand for public health can be a force for political change in favor of e ective public health spending. One, there is little evidence of "populist" demands from citizens, such as for cash, or subsidies, 25 without regard for the opportunity cost of public spending on health. A clear majority of citizens respond that any additional public spending for their area be allocated to health and nutrition services for their children rather than to cash transfers, job creation programs, or roads (Table 2).14 On a simple question about price subsidies that we tried out–whether gov- ernments should provide electricity for free– as many as 25 percent of respondents answered no, without any quali cation, while 34 percent quali ed that subsidies could be targeted to poor people. These responses from the average citizen respondent stand in contrast to the responses from those who were identi ed in the data as leaders of the village SHG–only 17 percent of SHG leaders answered no, and 52 percent answered with an unquali ed yes, com- pared to only 40 percent of citizens saying yes (Table 3).15 This pattern of citizen responses is even more striking when compared with how the higher income and educated respondents, such as doctors, in our sample answered this question. Among doctors, for example, 57 per- cent answered with an unquali ed yes, that governments should provide free electricity, with only 6 percent saying no. Citizen responses to another set of questions– on whether they would vote for a candidate for village Mukhiya who o ers inducements at the time of elections, such as a gift or cash in exchange for votes–suggest a more sophisticated political way of thinking than one of gullible or cynical voters who are easy prey to vote-buying strategies. Only 10 percent of citizens answer that they would vote for the gift-giving or bribing candidate, with 89 percent saying they would not. It is not easy to dismiss these responses as arising only from social desirability bias because other respondents– those that are identi ed in the data as leaders of the village SHG, and the village ANMs–who may be equally subject to the possibility of such bias–are twice as likely to respond voting for the bribing candidate (20 percent respond they would vote for the bribing candidate). Furthermore, the pattern of responses to some follow-up questions probing whether the respondent thinks the bribing candidate is more likely to win elections, more likely to be corrupt, and more likely to get work done compared to candidates 14 https://www.brookings.edu/blog/future-development/2019/04/08/what-do-poor-people-think-about- direct-cash-transfers/ 15 The SHG responses are similar to the responses of women in the sample, suggesting that the di erence may be entirely due to gender. That is, we do not nd that those women who belong to SHGs are less likely to demand free electricity. 26 who do not bribe, provides a critical view of bribing candidates. Across all respondents– the average citizen, the SHG leaders, and the ANMs–more than 75 percent respond that the bribing candidate is more likely to be corrupt, and more than 80 percent respond that the non-bribing candidate would get more work done. Even those who answered that they would vote for the bribing candidate are more likely to respond that the bribing candidate is more likely to be corrupt and less likely to get work done than a candidate who doesn’t bribe. This pattern suggests that citizens may inherently dislike bribing candidates, but nevertheless be compelled to vote for them for other reasons, such as if there is no choice (all candidates bribe), or if the bribing candidate is more likely to win. In our data, those who respond that they would vote for the bribing candidate also tend to respond that the bribing candidate is more likely to win elections. Two, citizens have political aspirations. When asked whether they would consider running for political o ce, as many as 31 percent of respondents answer "de nitely". Even when we restrict the sample to only the third for which the respondent is a woman, as many as 21 percent respond "de nitely" compared to only 5 percent among those women who are frontline public health workers. The average female respondent is only slightly less likely than SHG leaders to report interest in running for o ce (26 percent of SHG leaders answer "de nitely"). This reported interest in running for o ce in our survey is consistent with the large numbers of candidates actually observed in GP elections in Bihar, with more than 10 on average contesting the Mukhiya position. Other states, such as Andhra Pradesh, in contrast, have 2-3 candidates contesting on average (Afridi et al, 2018). Of course, the motives behind these aspirations to run for political o ce is unclear. On the one hand, these high rates of reported interest in running for o ce could be because holding local political o ce is lucrative, presenting opportunities to extract rents from state-funded public program. Even without overtly corrupt motives, standing for local elections may be one way to get an income earning position, in an economic environment where jobs are scarce. On the other hand, these responses suggest a highly contestable local political market, with low bar- riers to entry, which may enable public service motivated individuals to become local leaders. 27 State government policy-makers may be able to leverage this local political contestability to address problems of implementation and delivery.16 Among those citizens who have already identi ed themselves as political contenders–the in- cumbent village politicians and those who contested for the Mukhiya (village head) position in the previous election–the data show that these village politicians are distinct from other village-level respondents (citizens, ANMs, ASHAs, AWWs, Self Help Group members) in hav- ing higher measures of public service motivation and integrity. Figures 7 and 8 show the cumulative distribution function (CDF) of the public service motivation and integrity indices for di erent categories of village-level respondents. The gures show that village politicians are distributed at higher measures of public service motivation and integrity, compared to the distribution of other village-level respondents. This could be because politicians tend to answer the questions related to these characteristics less sincerely, and hence these measures may not be capturing real di erences in public service motivation or integrity among local politicians. We do nd some evidence, however, that the public service motivation measure among local politicians tends to be positively correlated with health service delivery reported by citizens-that is, citizens are more likely to receive maternal and child health services in villages where the average politician is measured as having higher public service motivation. Table 4 shows that women are more likely to receive iron and folic acid, and food supple- ments during pregnancy in villages where local politicians are measured as having higher public service motivation. The last column of Table 4 shows that women are less likely to report never receiving postnatal care in villages where politicians are measured as having public service motivation. We note that not all our survey measures of access to maternal and child health services are signi cantly correlated with politicians’ public service motivation. The purpose here is not to argue that we have identi ed a variable–namely, politicians’ pub- lic service motivation–as a key predictor of access to health services, but rather to furnish as much evidence as possible on the potential of local political contestation to support the policy 16 Further evidence of low barriers to entry for local political leadership comes from politicians’ responses to questions about their family political history. As many as 35 percent of incumbent Mukhiyas, 80 percent of Mukhiya contenders and 62 percent of Ward members report that no one in their family has ever held political o ce. The majority of these politicians, including those who come from non-political families, report that they would de nitely run again for o ce. 28 directions we propose under the current calls for urgent action. 7 Potential of frontline public health workers Citizens report greater reliance on publicly provided preventive and promotive health services than on curative health services. For example, in response to a question about how often they use government health facilities when someone in the family falls ill, only 12 percent answer "often", while 47 percent answer "rarely" or "never". This is consistent with the ndings from the work of Das et al (2016) that people in rural India tend to rely on private providers for curative health care. In contrast, when women in the sub-sample of households who had experienced a recent pregnancy (within the past 5 years) were asked which type of provider gave them ante-natal care, 85 percent responded by indicating a public provider (the ASHA and AWW are the most frequently cited); only 14 percent indicated a private sector provider. This di erence in the extent to which the public sector is used for preventive and promotive rather than curative care is particularly large for poorer and less politically connected citizens. Table 5 shows that respondents who are in positions of political power–elected politicians of local government (the panchayats at the village, block and district levels), members of the state legislative assembly (MLAs) and members of the national parliament (MPs)–are signif- icantly more likely to use government clinics or hospitals for curative care needs. We are unable to control for measures of income and assets because these questions were not asked in the surveys of higher-level politicians owing to sensitivity issues and time constraints in getting these powerful respondents to sit down for a survey. We hypothesize that the esti- mated in uence of political power in accessing curative care in government hospitals may be even higher if we were to be able to control for income, since higher-level politicians would tend to have higher incomes, and higher incomes are associated with greater use of expensive private care facilities. Among the di erent frontline workers who could provide maternal and child health/nutrition services–the ASHAs, the AWWs, and the village sub-center ANMs–citizens systematically re- port much lower reliance on ANMs than on the quasi-volunteer workers (ASHAs and AWWs). 29 For example, when recently pregnant women in the sample were asked who they most relied upon for advice during their pregnancy, 33 percent responded by indicating the ASHA, fol- lowed by 28 percent indicating a family member (such as, their mother-in-law); only 5 percent indicated a nurse or ANM. Similarly, only 5 percent of the women indicated the sub-center nurse as the provider of ante-natal care, compared to 27 percent indicating the ASHA, and 43 percent indicating the AWW.17 In response to a question about where their last child was delivered, only 27 percent of women indicated the health center (including both the village sub-center and the block-level PHC, where the ANMs are the key personnel assisting with child births), while 17 percent indicated a home-birth, 34 percent indicated the district hos- pital, and 21 percent indicated private providers. Furthermore, 41 percent of citizens report that the ANMs are rarely or never there to provide health services at the village sub-center, while only 18 percent respond that the ANMs are often there. Consistent with the evidence of lack of availability of ANMs, we nd that our survey measures of public service motivation and integrity are lower among ANMs than among other cadres health workers. Figures 9 and 10 show that ANMs are distributed at lower levels of integrity and motivation compared to the distribution of other cadres of health workers. This evidence underscores the need for complementary reforms (in addition to increasing public spending on frontline health workers) to address the problem of incentives and norms among frontline health workers. 8 Potential of complementary communication campaigns Theoretical analysis of how changes in norms come about points to a triggering role for polit- ical contestation, and the leaders selected through it. Leaders can play this role as “prominent agents” who signal a shift in beliefs among society at large (Acemoglu and Jackson, 2015). Growing experience with political engagement and the learning that comes from it, such as through frustration and indignation with bad outcomes, can create fertile conditions for 17 The AWWs may be regarded as a signi cant source of ante-natal care because AWWs may be distributing supplements like folic acid. In answering a question about access to iron folic acid, 40 percent indicated that it was provided by the AWW. 30 change in political norms (Bidner and Francois, 2013). Recent theoretical developments on the management of complex organizations generally, both in the private and public sectors, is also pointing to the role of leaders in shaping organizational culture. For example, Akerlof (2015, 2017) de nes the concept of “legitimacy” of leaders in getting lower level personnel to follow the organization’s objectives of their own accord, through peer-to-peer interaction, without incentive payments and monitoring from the top. In each of these theories of how changes in norms come about, information and communi- cation that shift expectations about how others are behaving is the necessary element that brings about change. In some models, the information is communicated through the types of leaders that are selected (Acemoglu and Jackson, 2015). In others, information is gathered and shared over time among citizens through the experience of political participation (Bidner and Francois, 2013). In any problem where norms support a less than desirable outcome, shifting to a new norm requires information sharing and communication among the actors to update their beliefs about how others are behaving. The role of political leaders and processes of po- litical participation as the channels for sharing information that shifts norms in public sector agencies, is consistent with classic work on norms for collective action (Ostrom, 2000). Using these theoretical insights, communication campaigns can be designed to harness the potential of local political contestation in shifting norms towards the public good of public health. Given available platforms of regular meetings among peers–as is the case in the de- partment of health in Bihar–these meetings can also be an arena in which communication strategies can be used to strengthen peer-to-peer pressure for professional norms. Our survey included a detailed module asking about media consumption across all categories of respondents. The primary purpose in designing these modules was to understand the con- text and patterns of media consumption with an eye towards future communication inter- ventions targeted at strengthening norms. There are two elements to consider in this regard: one, which media are likely to provide local (versus national) news; and two, which media are likely to be shared in common across respondents, so that communication about local mat- ters can be expected to reach them all simultaneously. Put another way, to what extent are 31 di erent types of respondents–the health workers, citizens, politicians–accessing completely di erent media outlets for news, versus converging on some common platforms?18 The following patterns emerge: One, among the "middle/upper-class" (urban, educated, higher caste and income) respondents in our survey–the district and block-level politicians, health supervisors, and doctors–98 per- cent respond that they read newspapers regularly, and cite 4 Hindi newspapers as their favored papers. Two, Figure 11 shows that GP-level respondents–who are rural, lower caste, education and income– are far less likely to report reading newspapers regularly (daily or 2-3 times a week). However, there is considerable variation across di erent types of respondents: while only 31 percent of citizens report reading newspapers regularly, 82 percent of ANMs and 58 percent of GP politicians do so. Three, Figure 12 shows GP-level responses to the question "Which medium do you rely on most for news–newspapers, TV, radio, social media/internet?". Even though 82 percent of ANMs report reading newspapers regularly, only about half of ANMs answer that newspapers are their main source of news, with 44 percent citing TV. Similarly, 33 percent of GP politicians cite newspapers as their main source, while 41 percent cite TV. (In fact, the middle/upper- class respondents in point one, also split between TV and newspapers when answering this question of the main media on which they rely for news). Four, Figures 13, 14 and 15 together show that among those who regularly read newspapers, there is much lower dispersion in which newspapers they read compared to a high degree of fragmentation across di erent TV stations of those who cite TV as their main source of news. Those who read newspapers converge on the same 4 Hindi language dailies that are local to Bihar. The TV channels on which there is some convergence across respondents, are all national channels and therefore unlikely to focus on local news. 18 Segmentation of consumers across di erent media markets is being analyzed in the United States as a source of increasing political polarization and non-cooperative norms, because it allows people to select into so-called "echo-chambers" to con rm their priors rather than seek common ground with others. 32 Five, the average citizen in rural areas is more likely to rely on national Hindi TV channels for news rather than on local newspapers. Strikingly, SHG leaders report the lowest reliance on or use of newspapers (around 11 percent), and much more on national TV (43 percent). Citizens and SHG leaders thus appear to have little consumption of or access to local news from established news outlets, relying on a variety of "other" informal sources, such as their social networks. Hindi newspapers emerge as the dominant media for local news that is shared across in uen- tial respondents (ANMs, GP politicians, and middle/upper-classes of respondents above the GP). A communication campaign that seeks to shift views simultaneously, across health ser- vice cadres (ANMS and their supervisors at the block and district level), GP politicians, and the upper-classes of the state, would thus need to seriously consider the role of newspapers as a commonly shared and established media. The existence of newspapers as the shared platform for accessing news is suited to communicating evidence and complex information that requires reading to digest. At the same time, buying time on national Hindi TV channels to devote to Bihar-relevant and speci c messages, would be an important complement, to reach citizens and SHG leaders who do not read newspapers, and to re-enforce the messages delivered through newspapers. 9 Conclusion: Outlining policy directions for public health spending The evidence provided here of a vicious cycle of low trust, perverse expectations, and poor performance in public health systems should inform the design of investments in public health that policymakers are poised to undertake. Designs that do not address the trust de cit will miss an opportunity to use economic theory and research to build trust in public institutions at a time when it is urgently needed for public health. This paper has presented evidence of potential forces of change in local politics and health bureaucracies that can be harnessed by policymakers to establish a cadre of trustworthy and professional frontline public health 33 workers.19 This concluding section outlines policy directions that could be tailored to any context, as we do for Bihar, for reform leaders and international partners to consider in their investments in public health systems. The following three elements characterize these policy directions: • Infuse messages into management meetings in health bureaucracies: the Bihar survey found that across all cadres of health workers, there are regular meetings held with district and block-level management (75 percent of sub-center ANMs say they have weekly meetings; 86 percent of ASHAs say they have at least monthly meetings with the MOIC at the Block PHC), and a signi cant proportion of health workers are part of WhatsApp groups (39 percent of village-ANMs and 52 percent of block-ANMs say they belong to a health workers’ WhatsApp group.). Evidence-based messages about the need for professionalism and motivated service delivery can be crafted for sharing at meetings that are already happening organically and provide an institutionalized space for strengthening peer-to-peer norms.20 In the Bihar and India context in particular, this involves a radical shift away from the culture of "scoldings" and disciplining, towards encouragement and recognition of health workers as professionals whose services are urgently needed for the good of society. • Newspaper and TV campaigns: Bihar provides a context in which readership of Hindi-language daily newspapers provides a platform to share complex messages that require reading and discussion among people. As discussed above, newspaper read- ing is already happening as a way of nding out local, Bihar-speci c news, and thus presents an organic platform that external agents can use to disseminate credible, data- based messages. At the same time, the relative importance of TV viewing among the poorer groups of respondents in villages suggests a role for inserting Bihar-speci c in- 19 The survey on which this paper draws was designed to understand the management environment in which health workers perform their tasks, and did not have scope to address the important area of professional training. We thank Hemant Pawar and the World Bank’s India country team for highlighting the need for better pre- service and in-service training of health workers. Our paper can be used to craft complementary governance interventions so that better trained health personnel are motivated to use their training for health impact. 20 Contrast this against the lack of such institutionalized space for social accountability–86 percent of citizens respond that they have never heard of a village committee on health and nutrition. 34 formation as sponsored programs or ads into national TV channels (the ones people report watching). Such media campaigns are not only hypothesized to reduce rent- seeking by village politicians, and strengthen local incentives for health services, but also to strengthen the hands of state-level reform leaders to invest in a new or reju- venated cadre of frontline health workers. Research suggests that public health issues have not enjoyed political salience in India because of di erentiated demands across socio-economic classes. Upper and middle classes in urban India may lack pertinent information about the value of public health systems in reaching poor households in rural India. Campaigns deployed through the mainstream media in Bihar that reaches citizens across these di erent socio-economic classes can, potentially, enable reform leaders to build political support for investing in frontline public health workers. • New recruitment and steady wages to community-based public health profes- sionals (but crucially, accompanied by the two communication campaigns above): as reviewed above, the economics literature on managing principal-agent relationships in the public sector suggests that there is considerable scope for improving bureaucratic productivity by reducing reliance on incentives and strengthening intrinsic motivation and professional norms. Yet, RCTs that focus on incentives alone, with no compara- ble arm on strengthening professional norms, have captured the policy imagination, to the detriment of trying out di erent approaches with potential. Financing incentive payments, as well as implementing them with delity to the design, require consider- able resources. In normal times, it would be of value to policymakers to test whether lower cost management reforms can improve service delivery compared to implement- ing high-powered incentives. In the current time of crisis, the arguments provided in this paper support a leap of faith in recruiting, training, empowering and equipping frontline health workers with steady wages. Investing in this cadre to manage and pre- vent disease outbreaks in their communities, liaising with the higher-tier health system as advised to do so by the global scienti c and technical research community, is indis- pensable. Doing so by taking the game of local politics and bureaucracy seriously into account will increase the likelihood of success in ghting the current and future health 35 crises. The policy package described above is similar in spirit to what Brazilian Governors in the state of Ceara adopted over 1987-1994, as they confronted a situation of poverty recently made worse by a drought; some of the highest rates of infant mortality (102 deaths per 1,000) in the world; and no functioning public health system. The Governors combined meritocratic recruitment of a new cadre of public health workers with a blizzard of communication using local radio, the dominant media to in uence local politics and governance. Tendler and Freed- heim (1994) provide a case study of how these steps built trust in the public health system and dramatically turned around health outcomes in the state within a span of a few years. Infant mortality fell by 36 percent and vaccination coverage increased from 25 to 90 percent. Perhaps even more importantly, the Ceara model was scaled-up across Brazil as the country’s Family Health Program (now called Family Health Strategy), relying on state-recruited community health agents to deliver basic services targeted at poor households. Rochas and Soares (2010) estimate that this program has been successful in improving health outcomes in poor areas. More examples from other parts of the world are provided by Wagsta and Claeson (2004) of how professional norms in teams of health workers leads to better health outcomes in the communities they serve. This paper has focused on leveraging theoretical and new empirical research for ideas on how to bring about these professional norms in institutional contexts where they are missing to start with. Policymakers who are poised to increase public health spending to respond to the global health and economic crises can immediately use the ideas presented in this paper. Fixing di cult problems like public health requires a complex mix of policy instruments, going be- yond health to institutions of governance and trust. This paper aims to bring these complex ideas to the fore at a time when international nancing is being mobilized to strengthen public health systems. Fast tracked health projects to combat the new coronavirus outbreak should use the evidence provided in this paper when designing those projects, and help developing countries move beyond the crisis to strengthening public institutions for health. The research division of the World Bank stands ready to help in clarifying and tailoring the ideas for im- 36 plementation in speci c contexts, along with evaluating its impact over time. 10 References Acemoglu, Daron, and Matthew O. Jackson. 2015. “History, Expectations, and Leadership in the Evolution of Social Norms.” Review of Economic Studies 82 (1): 1–34. Acemoglu, D., Kremer, M. and Mian, A., 2008. Incentives in markets, rms, and governments. The Journal of Law, Economics, and Organization, 24(2), pp.273-306. Acemoglu, D., Johnson, S. and Robinson, J.A., 2002. Reversal of fortune: Geography and institutions in the making of the modern world income distribution. The Quarterly journal of economics, 117(4), pp.1231-1294. Acemoglu, D., Johnson, S. and Robinson, J.A., 2001. The colonial origins of comparative de- velopment: An empirical investigation. American economic review, 91(5), pp.1369-1401. Afridi, Farzana. 2017. Governance and Public Service Delivery in India. S-35407-INC-1, Inter- national Growth Centre (April). Ahmed, T., Arur, A., de Walque, D. and Shapira, G., 2019. Incentivizing Quantity and Quality of Care: Evidence from an Impact Evaluation of Performance-Based Financing in the Health Sector in Tajikistan. Policy Research Working Paper 8951. The World Bank. Akerlof, R. 2015. “A Theory of Authority.” Working Paper, University of Warwick Akerlof, R., 2017. The importance of legitimacy. The World Bank Economic Review, 30(Sup- plement 1), pp.S157-S165. Alesina, A. and Tabellini, G., 2007. Bureaucrats or politicians? Part I: a single policy task. American Economic Review, 97(1), pp.169-179. Alesina, A. and Tabellini, G., 2008. Bureaucrats or politicians? Part II: Multiple policy tasks. Journal of Public Economics, 92(3-4), pp.426-447. Alesina, A. and Giuliano, P. 2015. “Culture and Institutions.” Journal of Economic Literature 37 53 (4): 898–944. Ashraf, N., Bandiera, O. and Lee, S.S., 2015. Do-gooders and go-getters: career incentives, selection, and performance in public service delivery. Harvard Business School. Ashraf, N., Bandiera, O. and Jack, B.K., 2014. No margin, no mission? A eld experiment on incentives for public service delivery. Journal of Public Economics, 120, pp.1-17. Banerjee, Abhijit, Esther Du o, and Rachel Glennerster. 2008. “Putting a Band-aid on a Corpse: Incentives for Nurses in the Indian Public Health Care System.” Journal of the Eu- ropean Economic Association 6 (2-3): 487-500. Banerjee, Abhijit, and Lakshmi Iyer. 2005. “History, Institutions, and Economic Performance: The Legacy of Colonial Land Tenure Systems in India.” American Economic Review 95 (4): 1190–213. Bang, A.T., Bang, R.A., Baitule, S.B., Reddy, M.H. and Deshmukh, M.D., 1999. E ect of home- based neonatal care and management of sepsis on neonatal mortality: eld trial in rural India. The Lancet, 354(9194), pp.1955-1961. Bang, A.T., Bang, R.A., Tale, O., Sontakke, P., Solanki, J., Wargantiwar, R. and Kelzarkar, P., 1990. Reduction in pneumonia mortality and total childhood mortality by means of community- based intervention trial in Gadchiroli, India. The Lancet, 336(8709), pp.201-206. Breza, E., Kaur, S. and Shamdasani, Y., 2018. The morale e ects of pay inequality. The Quar- terly Journal of Economics, 133(2), pp.611-663. Chaudhury, Nazmul, Je rey Hammer, Michael Kremer, Karthik Muralidharan, and F. Halsey Rogers. 2006. “Missing in Action: Teacher and Health Worker Absence in Development Coun- tries.” Journal of Economic Perspectives, 20 (1): 91-116. Claeson, Mariam; Elmendorf, A. Edward; Miller, Daniel Charles; Musgrove, Philip A.; Preker, Alexander S. [editor]. 2003. Public health and World Bank operations. Health, Nutrition and Population (HNP) discussion paper. Washington, DC: World Bank. 38 Dal Bó, Ernesto, Frederico Finan, and Martin A. Rossi. 2013. “Strengthening State Capabilities: The Role of Financial Incentives in the Call to Public Service.” 128 (3): 1169-1218. Das, Jishnu, Alaka Holla, Aakash Mohpal, and Karthik Muralidharan. 2016. “Quality and Accountability in Health Care Delivery: Audit-study Evidence from Primary Care in India.” American Economic Review 106 (12): 3765-99. Dhaliwal, Iqbal, and Rema Hanna. 2014. “Deal with the Devil: The Successes and Limitations of Bureaucratic Reform in India.” No. w20482. National Bureau of Economic Research. Filmer, Deon and Adam Wagsta , 2020. "How service providers are paid matters as much as how they are paid". In Artuc, Erhan; Cull, Robert J.; Dasgupta, Susmita; Fattal Jaef, Roberto N.; Filmer, Deon P.; Gine, Xavier; Jacoby, Hanan G.; Jolli e, Dean Mitchell; Kee, Hiau Looi; Klapper, Leora; Kraay, Aart C.; Loayza, Norman V.; Mckenzie, David J.; Ozler, Berk; Rao, Vi- jayendra; Rijkers, Bob; Schmukler, Sergio L.; Toman, Michael A.; Wagsta , Adam; Woolcock, Michael. 2020. Toward Successful Development Policies : Insights from Research in Develop- ment Economics (English). Policy Research working paper; no. WPS 9133. Washington, D.C. : World Bank Group. Flabbi, Luca, and Roberta V. Gatti. 2018. A Primer on Human Capital. Policy Research Work- ing Paper No. 8309. The World Bank. Francois, Patrick. 2000. “‘Public Service Motivation’ as an Argument for Government Provi- sion.” Journal of Public Economics 78 (3): 275–99. Fujiwara, T., 2015. Voting technology, political responsiveness, and infant health: Evidence from Brazil. Econometrica, 83(2), pp.423-464. Gupta, Shaibal. 1981. “Non-Development of Bihar: A Case of Retarded Sub-Nationalism.” Economic and Political Weekly 16 (37): 1496-1502. Kandpal, Eeshani, Benjamin P. Loevinsohn, Christel M. J. Vermeersch, Elina Pradhan, Mad- hulika Khanna, Mark Kaleb Conlon and Wu Zeng. 2019. “Impact Evaluation of Nigeria State Health Investment Project.” Report. The World Bank 39 Khemani, Stuti. 2019. “What is State Capacity?” Policy Research Working Paper no. WPS 8734. Washington, D.C. : World Bank Group Khemani, Stuti. 2015. “Buying Votes versus Supplying Public Services: Political Incentives to under-Invest in pro-Poor Policies.” Journal of Development Economics 117 (C): 84–93. La Forgia, Gerard, Shomikho Raha, Shabbeer Shaik, Sunil Kumar Maheshwari, and Rabia Ali. 2015. “Parallel Systems and Human Resource Management in India’s Public Health Services: A View from the Front Lines.” Public Administration and Development 35 (5): 372-389. Lo, T.Q., Marston, B.J., Dahl, B.A. and De Cock, K.M., 2017. Ebola: anatomy of an epidemic. Annual review of medicine, 68, pp.359-370. Mani, A. and Mukand, S., 2007. Democracy, visibility and public good provision. Journal of Development Economics, 83(2), pp.506-529. Miller, N.P., Milsom, P., Johnson, G., Bedford, J., Kapeu, A.S., Diallo, A.O., Hassen, K., Ra que, N., Islam, K., Camara, R. and Kandeh, J., 2018. Community health workers during the Ebola outbreak in Guinea, Liberia, and Sierra Leone. Journal of global health, 8(2). Moon, S., Sridhar, D., Pate, M.A., Jha, A.K., Clinton, C., Delaunay, S., Edwin, V., Fallah, M., Fidler, D.P., Garrett, L. and Goosby, E., 2015. Will Ebola change the game? Ten essential reforms before the next pandemic. The report of the Harvard-LSHTM Independent Panel on the Global Response to Ebola. The Lancet, 386(10009), pp.2204-2221. Moore, Celia, James R. Detert, Linda Klebe Treviño, Vicki L. Baker, and David M. Mayer. 2012. “Why Employees Do Bad Things: Moral Disengagement and Unethical Organizational Behavior.” Personnel Psychology 65:1–48. Muralidharan, Karthik, and Venkatesh Sundararaman. 2011. “Teacher Performance Pay: Ex- perimental Evidence from India.” Journal of Political Economy 119 (1): 39-77. Nooruddin, Irfan, and Pradeep Chhibber. 2008. “Unstable Politics: Fiscal Space and Electoral Volatility in the Indian States.” Comparative Political Studies 41(8): 1069-1091. 40 Nooruddin, Irfan, and Joel W. Simmons. 2015. “Do Voters Count? Institutions, Voter Turnout, and Public Goods Provision in India.” Electoral Studies 37(1): 1-14. Pandey, Priyanka. 2010. “Service Delivery and Corruption in Public Services: How Does History Matter?” American Economic Journal: Applied Economics 2 (3): 190–204. Singh, Prakarsh, and William A. Masters. 2017. “Impact of Caregiver Incentives on Child Health: Evidence from an Experiment with Anganwadi Workers in India.” Journal of Health Economics 55: 219–31. Wagsta , Adam and Mariam Claeson, 2004. "The Millenium Development Goals for Health: Rising to the challenges." The World Bank WHO and World Bank, 2017. Tracking Universal Health Coverage: Health Shocks and Poverty, Global Monitoring Report, World Health Organization and International Bank for Reconstruc- tion and Development Wolf, Martin. 2020. "The virus is an economic emergency too." Financial Times, March 18, 2020 World Bank. 2015. World Development Report 2015: Mind, Society, and Behavior. Washing- ton, DC: World Bank. World Bank, 2016. Making Politics Work for Development: Harnessing transparency and cit- izen engagement, Policy Research Report, Development Research Group, World Bank: Wash- ington, D.C. World Bank, 2017. World Development Report 2017: Governance and the Law. Washington, DC: World Bank 41 11 Figures and Tables Figure 1: Principal-Agent Relationships of Government Source: World Bank (2016). Figure 2: Map of Bihar. Note: Sample districts are colored in red and sampled blocks, within sampled districts, colored in green. 42 Figure 3: Share of health sta saying management meetings discuss bad performance and involve scoldings Note: This gure reports the share of health sta , in each survey, stating always or sometimes bad perfor- mance is discussed (left panel) or supervisors scold workers (right panel). Figure 4: Supervisor reports consistent with dominance of discussing bad performance in meetings Note: This gure reports the distribution of supervisors’ responses about how often workers are rewarded (grey bars); how often they are formally reported (orange bars); and how often bad performance is discussed (blue bars). 43 Figure 5: Share who agree with: Irrespective of my e orts, the system will not allow people’s health outcomes to improve. Note: This gure reports the share of health workers, in each group, that fully agree or somewhat agree that irrespective of their e orts the system will not allow people’s health outcomes to improve. Figure 6: Share who agree with: In my work, I have to take permission for every little thing. Note: This gure reports the share of health workers, in each group, that fully agree or somewhat agree that in their line of work they have to take permission for every little thing. 44 1 .8 Cumulative Probability .6 .4 .2 0 -4 -2 0 2 4 Index Public Service Motivation Citizens Politicians ANM ASHA/AWW SHGs Mukhiya Figure 7: Distribution of Public Service Motivation across Village Respondents Note: This gure reports the cumulative distribution functions (CDF) of the index of Public Service Motiva- tion for each respondents’ category. 45 1 .8 Cumulative Probability .6 .4 .2 0 -4 -2 0 2 4 Index Integrity Citizens Politicians ANM ASHA/AWW SHGs Mukhiya Figure 8: Distribution of Integrity across Village Respondents Note: This gure reports the cumulative distribution functions (CDF) of the index of Integrity for each respondents’ category. 46 1 .8 Cumulative Probability .6 .4 .2 0 -5 0 5 10 Index Integrity CHW ANM PHC ANM Subc Staff Nurse Doctor Supervisor Figure 9: Distribution of Integrity across Health Worker Cadres Note: This gure reports the cumulative distribution functions (CDF) of the index of Integrity for each respondents’ category. 47 1 .8 Cumulative Probability .6 .4 .2 0 -4 -2 0 2 4 Index Public Service Motivation CHW ANM PHC ANM Subc Staff Nurse Doctor Supervisor Figure 10: Distribution of Public Service Motivation across Health Worker Cadres Note: This gure reports the cumulative distribution functions (CDF) of the index of Public Service Motiva- tion for each respondents’ category. 48 How often read newspaper? 100 5.6 4.7 24.6 0.9 6.5 31.7 80 47.5 10.5 21.5 2.7 70.4 % of total 60 4.1 16.3 12.4 4.6 12.7 10.6 40 3.6 4.5 60.7 11.5 9.9 45.8 12.3 20 2.8 21.8 25.3 2.7 4.2 7.6 0 Citizens Politicians ANM ASHA/AWW SHG Daily 2-3 times a week Once a week 2-3 times a month Rarely Never Figure 11 Note: This gure reports the distribution of responses, for each surveyed group, about how often they read the newspapers. 49 Medium most used for news 100 2.8 11.2 11.9 20.2 11.2 8.9 1.9 10.0 29.5 80 5.9 4.7 12.9 10.8 % of total 11.2 43.9 60 6.5 40.7 49.6 40 32.6 42.8 20 40.2 33.2 23.1 23.8 10.5 0 Citizens Politicians ANM ASHA/AWW SHG Newspaper TV News Radio News Social Media/internet Other Figure 12 Note: This gure reports the distribution of responses, for each surveyed group, about which medium they use the most to access news. 50 Which is your favorite newspaper? 100 6.0 2.0 21.9 11.0 30.7 80 42.7 1.7 10.0 7.5 2.5 64.9 6.6 5.3 % of total 60 23.0 7.2 2.8 4.5 5.9 29.3 20.9 40 16.6 4.1 2.0 2.5 48.0 6.6 20 32.9 33.4 27.4 19.9 0 Citizens Politicians ANM ASHA/AWW SHG Dainik Jagran bihar Hindustan Prabhat khabar Dainik Bhaskar Bihar Other Doesn't read Figure 13 Note: This gure reports the distribution of responses, for each surveyed group, about which is their favorite newspaper. 51 Which is your favorite newspaper? | Read Newspapers 100 5.0 2.2 2.1 3.6 9.7 11.7 7.7 11.6 7.9 8.4 10.3 5.6 10.3 10.6 80 7.2 18.8 24.5 30.2 % of total 29.0 37.5 40 60 56.8 47.8 51.1 48.2 42.2 20 0 Citizens Politicians ANM ASHA/AWW SHG Dainik Jagran bihar Hindustan Prabhat khabar Dainik Bhaskar Bihar Other Figure 14 Note: This gure reports the distribution of responses, for each surveyed group, about how often they read the newspapers, conditional on responding they read some newspaper. 52 Which is your favorite TV channel for news? 100 19.6 30.3 34.0 80 41.6 49.9 16.8 % of total 11.5 2.8 60 2.7 13.5 4.2 9.3 13.9 3.1 7.1 6.5 6.3 1.4 3.3 6.1 14.6 40 3.7 11.2 5.6 4.2 5.2 1.9 6.2 1.5 5.4 2.9 8.0 10.3 6.3 3.4 3.4 3.2 12.3 5.6 20 9.1 6.5 3.1 10.1 7.0 5.3 15.2 15.0 11.4 10.4 7.7 0 Citizens Politicians ANM ASHA/AWW SHG AajTak ABP News DilliAajTak DD News Zee News ETV Bihar - bhojpuri India News India TV Other Doesn't watch news on TV Figure 15 Note: This gure reports the distribution of responses, for each surveyed group, which is their favorite TV channel for news. 53 100 Which is your favorite TV channel for news? | Watch news 16.5 20.9 20.5 23.8 29.1 3.9 80 2.5 6.1 3.5 4.7 5.6 10.2 11.6 9.6 3.1 % of total 6.3 5.9 8.4 60 7.2 8.7 8.1 6.7 7.5 9.3 6.4 13.8 14.0 7.7 9.6 11.2 40 2.3 15.6 12.8 5.2 6.2 17.6 15.3 10.6 8.1 20 12.0 21.7 18.6 17.3 20.7 13.3 0 Citizens Politicians ANM ASHA/AWW SHG AajTak ABP News DilliAajTak DD News Zee News ETV Bihar - bhojpuri India News India TV Other Figure 16 Note: This gure reports the distribution of responses, for each surveyed group, which is their favorite TV channel for news, conditional on watching news on TV. 54 Table 1: Determinants of Professional Identity and E cacy (1) (2) (3) ASHA/AWW ANM Pooled Demographics Married 0.150 [0.139] -1.144 [0.807] 0.154 [0.134] ST, ST or OBC -0.102 [0.110] -0.054 [0.230] -0.030 [0.098] Age -0.000 [0.005] 0.012 [0.016] 0.000 [0.005] Have ration card 0.019 [0.066] 0.045 [0.205] 0.014 [0.060] Uses whatsapp -0.078 [0.095] 0.077 [0.196] -0.049 [0.085] Asset Index (ICW) -0.049 [0.036] -0.047 [0.130] -0.055* [0.033] Behavioral traits Index Personality Traits -0.026 [0.052] -0.112 [0.146] -0.052 [0.051] Index Integrity 0.055 [0.040] 0.136 [0.137] 0.070* [0.038] Index Public Sector Motivation 0.033 [0.042] 0.107 [0.130] 0.033 [0.040] Index Enterpreneurship 0.085* [0.043] 0.284* [0.146] 0.091** [0.042] Supervisors & Peers Job tenure in years -0.004 [0.008] -0.004 [0.011] 0.001 [0.007] How many months of salary not received -0.014 [0.010] -0.091** [0.039] -0.019** [0.009] ANM supervisor has good or very good management skills -0.024 [0.139] -0.465 [0.398] -0.006 [0.133] Sometimes or always supervisor scold peers who don’t perform 0.056 [0.086] 0.057 [0.273] 0.099 [0.082] Supervisor report employees who don’t work and they lose job -0.128* [0.075] -0.053 [0.194] -0.126* [0.069] Sometime or always there is recognition for good work -0.014 [0.070] -0.043 [0.218] -0.021 [0.067] Sometimes or always interact with peers in BPHC meetings 0.320** [0.143] 0.201 [0.291] 0.258** [0.119] Quite a few or all ANM known are hardworking 0.252 [0.240] 1.102** [0.537] 0.242 [0.218] Quite a few or all ANM known are honest -0.140 [0.278] 1.041** [0.465] -0.047 [0.252] Share ANM colleagues hardworkers 0.267 [0.392] 0.764 [1.280] 0.414 [0.378] Share ANM colleagues honest -0.039 [0.368] -2.026 [1.677] -0.170 [0.355] Politics Politicians often/sometimes create work di culties -0.185** [0.082] -0.036 [0.237] -0.186** [0.076] Politicians often/sometimes support work -0.076 [0.084] 0.083 [0.278] -0.045 [0.080] Good workers transferred due others feel threatened -0.125* [0.066] 0.279 [0.235] -0.091 [0.062] Work somewhat or often hampered by delay of funds release -0.016 [0.073] -0.229 [0.226] -0.025 [0.068] Had to use money or connections to get job 0.054 [0.105] 0.084 [0.336] 0.078 [0.098] Important to get transfer or promotion: political connections -0.080 [0.112] -0.258 [0.342] -0.153 [0.102] Important to get transfer or promotion: informal payments -0.249** [0.108] 0.291 [0.314] -0.224** [0.101] Important to get transfer or promotion: both -0.132 [0.120] 0.089 [0.427] -0.190* [0.111] Constant -0.264 [0.452] 0.358 [1.617] -0.459 [0.425] Observations 896 81 979 Distric FE Yes No Yes R-Squared 0.452 0.420 0.434 Note: Ordinary Least Squares (OLS) regression estimates of the predictors of professional identity and e cacy (dependent variable is the index) among frontline health workers. Robust standard errors in brackets (* p<0.1, ** p<0.05, *** p <0.01) 55 Table 2: Preferences for Public Spending on Health vs Jobs, Cash or Roads (1) (2) (3) Health Over Jobs Health Over Cash Health Over Roads Indicators for di erent respondents (omitted category: citizens) SHG -0.047** [0.022] -0.036** [0.017] 0.007 [0.018] Contender for Mukhiya 0.049*** [0.018] 0.022* [0.013] 0.022 [0.016] Ward Member 0.034** [0.018] -0.010 [0.014] 0.019 [0.016] Mukhiya 0.048* [0.028] 0.034* [0.020] -0.005 [0.028] Panchayat Samiti -0.090** [0.045] -0.016 [0.031] 0.117*** [0.024] Zilla Parishad 0.083 [0.068] 0.105*** [0.010] 0.138*** [0.037] MLA/MP -0.364*** [0.073] -0.110* [0.060] -0.037 [0.060] CHW 0.101*** [0.016] 0.049*** [0.012] 0.050*** [0.014] ANM Subcentre 0.063 [0.042] 0.023 [0.030] -0.012 [0.039] ANM PHC 0.071*** [0.022] 0.016 [0.016] 0.117*** [0.015] Sta Nurse 0.051* [0.027] 0.047*** [0.017] 0.105*** [0.017] Doctors 0.045* [0.024] 0.066*** [0.014] 0.105*** [0.019] Supervisors -0.085*** [0.029] -0.006 [0.020] 0.096*** [0.019] Media Access Watch TV Daily 0.014 [0.010] 0.034*** [0.007] 0.065*** [0.009] Receive Newspaper Daily 0.045*** [0.013] 0.009 [0.010] 0.001 [0.011] Read Newspaper Daily 0.011 [0.014] 0.013 [0.010] -0.016 [0.012] Rely on Newspaper for News 0.020* [0.011] 0.007 [0.008] 0.034*** [0.009] Social Indicators Male 0.058*** [0.013] 0.005 [0.010] -0.022* [0.011] Married 0.026 [0.020] -0.017 [0.014] -0.008 [0.017] Age -0.000 [0.000] -0.000 [0.000] 0.000 [0.000] Muslim -0.009 [0.016] -0.057*** [0.014] -0.070*** [0.015] Constant 0.648*** [0.025] 0.878*** [0.018] 0.794*** [0.023] Observations 9247 9247 9247 R-Squared 0.023 0.018 0.032 Note: OLS estimates to compare the policy preferences (health vs. other uses of public spending) of di erent respondents to the preferences of citizens, who are the omitted category. Robust standard errors in brackets (* p<0.1, ** p<0.05, *** p <0.01) 56 Table 3: Respondents NOT Demanding Free Electricity and Loan Waivers (1) (2) No - Free Electricity No - Farming Loan Waivers Indicators for di erent respondents (omitted category: citizens) SHG -0.036* [0.020] -0.036** [0.017] Contender for Mukhiya 0.108*** [0.025] -0.001 [0.018] Ward Member 0.058** [0.024] -0.025 [0.016] Mukhiya 0.115*** [0.036] -0.016 [0.023] Panchayat Samiti -0.235*** [0.030] 0.261*** [0.046] Zilla Parishad -0.182*** [0.066] 0.172** [0.085] MLA/MP -0.320*** [0.038] 0.087 [0.062] CHW -0.002 [0.018] -0.037** [0.015] ANM Subcentre 0.103* [0.057] 0.038 [0.049] ANM PHC -0.072*** [0.023] 0.253*** [0.024] Sta Nurse -0.080*** [0.027] 0.254*** [0.030] Doctors -0.297*** [0.020] 0.056** [0.025] Supervisors -0.260*** [0.023] 0.058** [0.026] Media Access Watch TV Daily 0.088*** [0.012] 0.006 [0.010] Receive Newspaper Daily 0.004 [0.016] 0.010 [0.013] Read Newspaper Daily 0.031* [0.017] -0.024* [0.013] Rely on Newspaper for News 0.007 [0.013] -0.008 [0.010] Social Indicators Male 0.073*** [0.014] 0.024* [0.012] Married -0.014 [0.021] -0.014 [0.020] Age 0.000 [0.000] 0.001** [0.000] Muslim -0.047*** [0.018] -0.030*** [0.011] Constant 0.194*** [0.029] 0.089*** [0.023] Observations 9247 9247 R-Squared 0.043 0.042 Note: OLS estimates to compare the demand for Free Electricity and Loan Waivers among di erent survey respondents to demand among citizens, who are the omitted category. Robust standard errors in brackets (* p<0.1, ** p<0.05, *** p <0.01) 57 Table 4: Services Received from Community Health Workers (1) (2) (3) IFA Received Mother Received Supp Food Never Received PNC Selection traits : GP (Village) Politicians Average PSM Index - GP Level 0.251*** [0.080] 0.097* [0.058] -0.089* [0.047] Average Personality Index - GP Level -0.155* [0.089] 0.010 [0.065] 0.114** [0.045] Average Integrity Index - GP Level -0.051 [0.068] -0.010 [0.054] 0.054 [0.039] Average Enterpreneurship Index - GP Level 0.121* [0.072] 0.039 [0.057] -0.122*** [0.045] Selection traits: Community Health Workers (CHW) Average PSM Index - GP Level -0.137** [0.068] 0.077 [0.058] 0.008 [0.038] Average Personality Index - GP Level 0.026 [0.078] -0.011 [0.061] 0.026 [0.041] Average Integrity Index - GP Level -0.110* [0.063] -0.095** [0.046] 0.030 [0.032] Average Enterpreneurship Index - GP Level 0.110* [0.062] -0.001 [0.064] -0.012 [0.034] Average E cacy Index - GP Level 0.074 [0.073] 0.046 [0.053] 0.025 [0.040] Selection traits: Self Help Group (SHG) leaders Average PSM Index - GP Level -0.029 [0.051] -0.020 [0.043] 0.046* [0.027] Average Personality Index - GP Level 0.055 [0.069] 0.038 [0.054] -0.021 [0.034] Average Integrity Index - GP Level 0.044 [0.068] 0.016 [0.039] 0.004 [0.027] Average Enterpreneurship Index - GP Level -0.032 [0.055] 0.069* [0.039] 0.037 [0.028] Political Competition Working on health issues can help re-election -0.494** [0.225] 0.213 [0.161] 0.250** [0.101] Would vote for candidate who bribes -0.013 [0.083] 0.119* [0.067] 0.015 [0.046] Socio-economic Indicators Have ration card 0.072 [0.076] -0.014 [0.059] -0.006 [0.038] Asset Index (ICW) -0.026 [0.026] 0.010 [0.022] -0.014 [0.015] SC/ST 0.267** [0.114] 0.125 [0.094] -0.086 [0.058] OBC 0.118 [0.095] 0.093 [0.075] -0.052 [0.050] Muslim -0.257** [0.105] -0.061 [0.081] 0.077 [0.062] Below Primary 0.019 [0.117] 0.128 [0.083] 0.040 [0.063] Below Secondary 0.066 [0.089] 0.021 [0.071] 0.003 [0.043] Above Secondary -0.020 [0.093] -0.067 [0.070] -0.029 [0.046] Above High -0.156 [0.119] -0.087 [0.098] -0.015 [0.053] No of Adults 0.010 [0.016] 0.014 [0.011] -0.004 [0.008] No of Kids -0.044*** [0.016] -0.019 [0.012] -0.003 [0.010] Landless -0.064 [0.087] 0.053 [0.077] -0.072 [0.052] Small Holder -0.253*** [0.092] 0.043 [0.084] -0.036 [0.054] Low Expenditure -0.145 [0.096] -0.072 [0.088] -0.081 [0.055] Medium Expenditure -0.002 [0.076] -0.006 [0.062] -0.068* [0.039] Constant 0.758*** [0.184] 0.404*** [0.145] 0.465*** [0.116] Observations 674 705 707 R-Squared 0.172 0.120 0.113 Note: OLS regressions estimating predictors of women reporting that they received Iron and Folic Acid, Supplementary Food during their pregnancy, and Never receiving Post Natal Care from Community health workers. All speci cations include district xed e ects. Robust standard errors in brackets (* p<0.1, ** p<0.05, *** p <0.01) 58 Table 5: Use of Government Health Facilities for: (1) (2) Common Illnesses Hospitalization Indicators for respondents in increasing order of political power (omitted category: citizens) Contender for Mukhiya 0.035* [0.019] 0.025 [0.021] Ward Member 0.095*** [0.018] 0.085*** [0.020] Mukhiya 0.153*** [0.031] 0.061* [0.035] Panchayat Samiti 0.389*** [0.042] 0.260*** [0.046] Zilla Parishad 0.339*** [0.085] 0.262*** [0.094] MLA/MP 0.335*** [0.066] 0.172** [0.074] Socio-economic Indicators Muslim 0.006 [0.019] 0.027 [0.021] SC/ST 0.047** [0.019] 0.087*** [0.021] OBC 0.058*** [0.016] 0.096*** [0.018] Below Primary 0.014 [0.022] -0.013 [0.024] Below Secondary -0.003 [0.016] 0.022 [0.018] Above Secondary -0.000 [0.017] 0.004 [0.019] Above High -0.030 [0.022] 0.007 [0.025] Male 0.019 [0.012] 0.003 [0.014] Age 0.001** [0.000] 0.002*** [0.001] Constant 0.159*** [0.029] 0.239*** [0.032] Observations 6345 6345 R-Squared 0.029 0.018 Note: OLS regressions to estimate di erences among respondents with increasing political power, compared to the omitted category of citizens, in accessing government health facilities for curative care. Robust stan- dard errors in brackets (* p<0.1, ** p<0.05, *** p <0.01) 59