Performance pay and information: Reducing child undernutrition in India

https://doi.org/10.1016/j.jebo.2015.01.008Get rights and content

Highlights

  • I test for the impact of performance pay to government child care workers.

  • In an experiment covering 145 day care centers, I implement three treatments.

  • These are performance pay to workers, information to mothers and their combination.

  • Combining performance pay and information reduces malnutrition significantly.

  • This complementarity is shown to be driven by better mother–worker communication.

Abstract

This paper provides evidence for the effectiveness of performance pay to government health workers and how performance pay interacts with demand-side information. In a controlled study covering 145 child day-care centers, I implement three separate treatments. First, I engineer an exogenous change in compensation for childcare workers from fixed wages to performance pay. Second, I only provide mothers with information without incentivizing the workers. Third, I combine the first two treatments. This helps us identify if performance pay and public information are complements or substitutes in reducing child malnutrition. I find that combining incentives to workers and information to mothers reduces weight-for-age malnutrition by 4.2 percentage points in 3 months, although individually the effects are negligible. This complementarity is shown to be driven by better mother–worker communication and the mother feeding more calorific food at home. There is also a sustained long-run positive impact of the combined treatment after the experiment concluded.

Introduction

Performance incentives seem to be very effective in private firms (Lazear, 2000, Bandiera et al., 2007). However, there is little evidence of their impact in the public sector, especially in public health. Besides, even if performance incentives could work to boost efforts of government health workers, the gains may be easily undone by informational failures on the demand-side. In this paper, I test if performance pay to government health workers is effective in improving health outcomes and how it interacts with public information.

I look at the specific context of a government childcare worker in India. She may affect the health of children through two channels: (i) providing mid-day meals in her day-care center and (ii) informing mothers on child nutrition. However, Gragnolati et al. (2005) find that leakage of meals to nontargeted beneficiaries is widespread and childcare workers do not give guidance to parents on improving nutrition within the family food budget.1 A recent household survey in 100 Indian districts indicates 96 percent of the villages are served by government child-day care centers, although only 50 per cent provided food on the day of survey and just 19 per cent of the mothers reported that the workers provides nutrition counseling (Hungama Report, 2011). An estimated 1.27 million children die every year in India because they are malnourished.2

In a controlled study covering 145 child day-care centers and 4101 children (aged 3–6 years) in urban slums of Chandigarh, India, I implement three separate treatments.3 First, I engineer an exogenous change in compensation for childcare workers from fixed wages to performance pay. Second, I provide mothers with information without incentivizing the workers. Third, I combine the first two treatments, where along with the change in compensation for workers, I supply nutritional information to mothers directly. This helps us identify if performance pay and information are complements or substitutes in affecting health outcomes.

The key findings are as follows. Changing compensation from fixed wage to performance pay does not change a child's weight on average in 3 months relative to a control group. Only providing nutritional information to mothers also does not change weight relative to control. However, providing incentives to workers and information to mothers reduces weight-for-age malnutrition by 4.2 percentage points in 3 months. This effect is equivalent to increasing the average income of a household by 51 percent to achieve the same reduction in malnutrition (via a simple baseline correlation between child being malnourished and income of family).4 The weight increase of 71 g per month relative to a control is also comparable to the 100 g per month increase achieved from iron and deworming implemented in urban slums of Delhi (Bobonis et al., 2006). This points to a complementarity in increasing weight when incentives and information are supplied together.

Next, I find some evidence on the mechanisms underlying the change in weight. In the combined treatment, there appears to be an increase in calories at home that is significantly greater than when we only provide incentives or only information. There are also changes in the nature of mother–worker interaction. Workers on performance pay start paying more personalized visits to homes and also talk more about the child's diet with the mother. In turn, mothers reduce their visits to the day-care center for meeting the worker.

This paper contributes to the empirical literature on the effects of incentive pay on performance in organizations. The literature highlights potential pitfalls in implementing performance pay. For example, providing incentives for improvement in only malnourished children may lead to the worker applying extra effort at the cost of children who are normal weight. Keeping this in mind, the incentive treatment in the experiment was designed to disincentivize the worker for a decline in weight-for-age grade. Perverse incentives have been found to be active in Vermeersch and Kremer (2005) and Sylvia et al. (2012) where teaching effort was displaced for health input. Similarly, there could be short-term manipulations or plain cheating by workers (Figlio and Winicki, 2005, Jacob and Levitt, 2003). These concerns are addressed by hiring, training and monitoring independent enumerators.

In their excellent review of the performance pay schemes in developing countries, Miller and Babiarz (2013) note that sometimes performance pay on inputs may be more effective as compared to outputs as providers have more control on the usage and quality of services than on the actual outcomes. Basinga et al. (2011) and Gertler and Vermeersch (2012) report that bonuses based on inputs and services-based performance increased the delivery of health services by 23 percent in Rwanda, although, there were no improvements in the number of women completing all prenatal care visits or in children receiving full immunization schedules.5 Thus, input-linked incentives may also have a limited impact if the health inputs chosen for measuring performance are inefficient. Consistent with the observation in Miller and Babiarz's review, I find that performance incentives based on outcomes can help local providers use their knowledge productively even though the outcome is only partly under the control of the provider. On the other hand, health outcomes can often be costly to measure as compared to inputs and providers may lack knowledge and skills to work toward obtaining the performance bonus. In our context, it is difficult to measure inputs objectively as the demand-side (mothers) could collude with the providers when reporting their effort. Secondly, workers appear to be significantly more knowledgeable than the mothers they are supposed to advise according to a baseline quiz.

Empirically, most studies implement neither an exogenous change in compensation schemes nor have a valid control group (Prendergast, 1999, Chiappori and Salanie, 2003). This may be important if there are other management changes that are taking place at the same time or if unobservable factors can influence both outcome and compensation structure. There may also be an endogenous feedback of performance on the type of compensation. This experiment is in the same spirit as Bandiera et al. (2007) who have an exogenous change in compensation for managers in a private firm and a valid control group. Muralidharan and Sundararaman (2011) test for teacher performance pay in Andhra Pradesh in a unique experiment and show that performance pay works to improve student grades and find evidence of no teaching to the test. However, there exists almost no research on exogenously changing incentive schemes for workers in a public health organization within a controlled experiment. Health of young children is especially dependent on the quality of supply-side and demand-side as well as their interaction. Moreover, interaction between workers and mothers is especially important in a developing country context. This study is one of only two such controlled interventions evaluating the impact of performance pay linked directly to health outcomes in a developing country.6 Miller et al. (2012) show that performance incentives provided to primary school principals for reducing anemia in China are effective in reducing anemia prevalence by 25% by the end of the academic year. The incentives led to the principals influencing parents to affect the child's eating behavior at home. Although the bonuses in this paper are targeted toward grass-root level health service providers, they help to change dietary behavior at home in response to being incentivized on health outcomes. Individual providers help solve the principal-agent problem that may arise with organization-level incentives and they can customize approaches and use local information gathered from experience to influence behavior on the demand-side (Miller and Babiarz, 2013). However, even though individual incentives prevent free-riding, they could deter teamwork. As each health provider is only responsible for her set of children in our setting, diminished cooperation among providers is less of a concern as compared to a school setting where multiple teachers are usually responsible for a child's overall performance.

Optimal compensation schemes may be different in public organizations (Dixit, 2002, Besley and Ghatak, 2005). In the public sector, high-powered incentive schemes are rarely seen mainly because of a difficulty in measuring outcomes, multi-tasking by agents and intrinsic motivation.7 In this specific context, we have a measurable health outcome (weight-for-age). Moreover, scope for multi-tasking by these childcare workers is very limited.

Finally, supply-side interventions have been shown to be less effective if there is a failure on the demand-side. Banerjee et al. (2010) find that uptake of immunization is much higher if small non-financial incentives are provided on the demand-side along with an increase in supply.8 However, they are not able to isolate the impact of incentives to demand-side from the complementarity effect by the very nature of their experiment. This is one of the first experiments that decomposes the separate effects of incentives and information as well as their interaction effect. For example, in Iajya et al. (2013) and Soeters et al. (2011), the complementarity between information and incentives is not clear, as there are no incentive-only treatments. I find evidence to support that complementarity between demand and supply-side plays a major role in reducing child malnutrition, where changing a childcare worker's compensation from fixed to performance pay is effective only if specific information is supplied to mothers at the same time. The paper also contributes to the growing literature of information on health. The central findings of this literature are that information does matter if it is customized to the person targeted and if the information is specific and intensive (Madajewicz et al., 2007, Dupas, 2011a, Dupas, 2011b). The recipe book provided to mothers in the experiment uses these findings to list ten recipes which are easy-to-make, economical and use locally available ingredients.

This paper opens the black box of how incentives interact with information in public health and illuminates the behavioral mechanisms at play. Not only is this important for a child's future and her family, it is relevant for policy makers in shaping compensation schemes for childcare workers and health policy in general. Methodologically, this is not a randomized experiment due to the strong possibility of spillovers but a controlled study that uses a three-pronged approach of matching with a contemporaneous control group, difference-in-differences and placebo analysis to address endogeneity and reversion-to-the-mean.

The paper is organized as follows. Section 2 provides the context and develops a conceptual framework. Section 3 illustrates the experiment design. Section 4 describes the data and Section 5 presents the specification and main results. Section 6 reports the mechanisms. Section 7 checks for robustness of these mechanisms. Section 8 provides long-term impacts, Section 9 delineates policy implications and Section 10 concludes.

Section snippets

Context

In India, the government-run Integrated Child Development Services (ICDS) program targets close to 35 million children between the age of 3–6 years, through over 1.24 million government day-care centers or ‘Anganwadis’. Each Anganwadi is run by an Anganwadi worker who takes care of children (aged 3–6 years) in a small room from 9 am to 1 pm and is on a fixed salary of Rs. 2000 ($44.44) per month.9 Although, Muralidharan et al. (2011) and

Matching of Anganwadis

In December 2009, data on malnutrition rates at the Anganwadi-level was collected from the local Health Department.17 We only had information on the proportion of children who are malnourished in each center before the baseline. Chandigarh had 370 Anganwadis divided into 3 blocks. Each geographical block was further divided in

Compliance and attrition

Overall, 4101 children were weighed twice during the experiment. For 94 percent of these children, their mothers were also quizzed two times. Selective taking of the quiz by the smartest mothers may hinder us from finding the causal effect of the treatments. Therefore, it is necessary that the compliance rate among mothers is high. Table 2 illustrates the compliance and attrition rates. The compliance rate of the mothers is high due to three reasons: first, the workers were very cooperative

Specification

The main regression specification for finding the average effect of the treatments on weight of a child is as follows:

wijt=α(post)t+β(incentive)j+γ(recipe)j+ρ(combined)j+η(post*incentive)jt+θ(post*recipe)jt+ω(post*combined)jt+Xijt+ɛijt

wijt is the weight of a child i in Anganwadi j at time t. The variable post is a dummy that is 0 for baseline and 1 for endline. The variables incentive, recipe and combined are 1 if the child is in the treatment specified and 0 otherwise. Xijt are individual and

Mechanisms

This section focuses on the main mechanisms that could be driving the change in weight and relates the findings to extension of conceptual framework.

Propensity score matching

One of the concerns with the main result found through a difference-in-differences approach may be that the sample population in the combined treatment is different along observables as compared to the control group and this could introduce a bias in our estimates. I carry out a robustness check using propensity score matching. Propensity score is defined as the probability that a unit in the full sample receives the treatment, given a set of observed variables at baseline. First, I model the

Long-run impact

In order to check if there was a long-run impact of the combined treatment's effect after the conclusion of the three-month experiment, I measured the weights of children in these Anganwadis in April 2011. There could also have been a lagged effect of providing the recipe book on malnutrition or only performance pay which was not detected after three months. I find in Table 12 that the combined treatment has a positive impact of a similar magnitude even nine months after the incentive scheme

Policy implications

Often, public health programs focus only on distribution of nutritional information or an increase in the supply-side. This study reinforces the view that distribution of nutritional information if complemented with a performance incentive to the supply-side improves the chances of the program's success. Björkman and Svensson (2009) look at the effect of encouraging community-based monitoring on child health and find an increase of 0.14 in the z-score after 1 year. The comparative increase for

Conclusion

This paper provides evidence for the impact of performance pay in the public sector and how it interacts with demand-side information in impacting health. By designing an experiment in the specific context of a government childcare worker in India and implementing three separate treatments, we find that providing performance pay alone may not be effective in improving health outcomes. The worker does respond to incentives by making additional visits to homes, but it does not help improve child

Acknowledgements

I am grateful to Oriana Bandiera and Gerard Padró i Miquel for their guidance. I profited from helpful comments by Harold Alderman, Tim Besley, Ariel Fiszbien, Maitreesh Ghatak, Jun Ishii, Chris Kingston, Rohini Pande, Kate Sims and seminar participants at Amherst College, STICERD (LSE), Graduate Institute (Geneva), University of Leicester, LSE Asia Research Centre, IFPRI, Oxford University, Navarra Center for International Development, Yale University, Mt. Holyoke College, UC Davis, Indian

References (69)

  • A. Banerjee et al.

    Improving immunisation coverage in rural India: clustered randomized controlled evaluation of immunisation campaigns with and without incentives

    BMJ

    (2010)
  • J.R. Behrman et al.

    Returns to birthweight

    Rev. Econ. Stat.

    (2004)
  • J.R. Behrman et al.

    Long-Term Impacts of the Oportunidades Conditional Cash Transfer Program on Rural Youth in Mexico

    (2004)
  • J.R. Behrman et al.

    Will developing country nutrition improve with income? A case study for rural South India

    J. Polit. Econ.

    (1987)
  • T. Besley et al.

    Competition and incentives with motivated agents

    Am. Econ. Rev.

    (2005)
  • I. Bhushan et al.

    Achieving the Twin Objectives of Efficiency and Equity: Contracting Health Services in Cambodia

    (2002)
  • M. Björkman et al.

    Power to the people: evidence from a randomized field experiment on community-based monitoring in Uganda

    Q. J. Econ.

    (2009)
  • E. Bloom et al.

    Contracting for Health: Evidence from Cambodia

    (2006)
  • G.J. Bobonis et al.

    Anemia and school participation

    J. Hum. Resour.

    (2006)
  • L.E. Caulfield et al.

    Undernutrition as an underlying cause of malaria morbidity and mortality in children less than five years old

    Am. J. Trop. Med. Hyg.

    (2004)
  • Census of India

    (2001)
  • N. Chaudhury et al.

    Missing in action: teacher and health worker absence in developing countries

    J. Econ. Perspect.

    (2006)
  • P.A. Chiappori et al.

    Testing contract theory: a survey of some recent work in advances in economics and econometrics

  • V. Datta

    Factors Affecting Job Performance Of Anganwadi Workers: A Study of Three Districts of Maharashtra. A Study sponsored by UNICEF

    (2001)
  • A. Deaton et al.

    Nutrition in India: Facts and Interpretations. Working Paper 1071

    (2008)
  • R. Dehejia et al.

    Causal effects in nonexperimental studies: reevaluating the evaluation of training programs

    J. Am. Stat. Assoc.

    (1999)
  • DHS India (2005–2006). Final Report

    (2007)
  • A. Dixit

    Incentives and organizations in the public sector: an interpretative review

    J. Hum. Resour.

    (2002)
  • E. Duflo et al.

    Additional Resources versus Organizational Changes in Education: Experimental Evidence from Kenya. mimeo

    (2009)
  • P. Dupas

    Health behavior in developing countries

    Annu. Rev. Econ.

    (2011)
  • P. Dupas

    Do teenagers respond to HIV risk information? Evidence from a Field Experiment in Kenya

    Am. Econ. J.: Appl. Econ.

    (2011)
  • S. Fishman et al.

    Malnutrition and the Global Burden of Disease: Underweight in Comparative Quantification of Health Risks: The Global and Regional Burden of Disease due to 25 Selected Major Risk Factors

    (2004)
  • P. Gertler et al.

    Using Performance Incentives to Improve Health Outcomes. World Bank Policy Research Working Paper No. 6100, Impact Evaluation Series No. 60

    (2012)
  • S. Ghosh et al.

    Nutritional problems in urban slum children

    Indian Pediatr. J.

    (2004)
  • Cited by (23)

    • Impact of caregiver incentives on child health: Evidence from an experiment with Anganwadi workers in India

      2017, Journal of Health Economics
      Citation Excerpt :

      Performance pay was also more expensive and difficult to provide, however, so policymakers might prefer fixed payments designed to enhance workers’ intrinsic motivations. The trial reported in this study builds on Singh (2015), continuing a series of trials designed to inform performance pay in the ICDS system in India. Related research concerns the use of tournament-type contests among Anganwadi workers in ICDS centers, the specific kinds of effort that workers use to achieve children’s weight gain, and complementarity or substitution between what they provide and children’s diets or care practices at home.

    • Incentives, information and malnutrition: Evidence from an experiment in India

      2017, European Economic Review
      Citation Excerpt :

      HA treatment leads to a large decrease in severe malnutrition given the time period and could sustain itself after three months because of immunity and resistance acquired by children. Long-term positive effects after discontinuation of a similar incentive scheme were reported in Singh (2015) in Chandigarh. The increase in weight in that incentive scheme was 213 g over three months relative to the control group and this increase was sustained until one year after baseline after the discontinuation of the treatment.

    View all citing articles on Scopus
    View full text