When financial incentives backfire: Evidence from a community health worker experiment in Uganda☆
Introduction
Hundreds of thousands of community health workers (CHWs) provide access to basic health services in poor countries. However, many CHWs do not perform all of their assigned activities (Strachan et al., 2012; Chen et al., 2004). Unfortunately, little evidence exists on how best to increase CHW effort. Given the expanded role many expect CHWs to play in achieving the sustainable development goals (WHO et al., 2016), ensuring that such programs function effectively is of great concern for the global health community.
In light of this, there is growing support for an entrepreneurial model, where CHWs sell health products door-to-door and retain the profits (Skoll, 2018; Kopf, 2016; Economist, 2012; Bjorkman Nyqvist et al., 2019). Proponents claim such a model increases and sustains CHW effort. However, it is unclear if an entrepreneurial model increases effort compared to more traditional CHW models (e.g., free health product distribution). In this paper, we use a field experiment to examine whether an entrepreneurial model increases CHW effort relative to free health product distribution.
Textbook economic theory suggests the entrepreneurial model should improve CHW effort relative to free distribution because CHWs earn more if they distribute more products. At the same time, theories of intrinsic motivation and social recognition suggest CHWs could put in less effort under an entrepreneurial model than when distributing products for free, if the utility they get from helping their community with free distribution outweighs the utility from higher pay (Deci, 1975). These social motivations are likely to be higher for CHWs than for workers in other settings, as CHWs serve families they know. Moreover, the type of person that chooses to become a CHW might be particularly socially motivated (Deserranno, 2019).
The evidence on how financial incentives affect performance in poor country health sectors is mixed. Basinga et al. (2011) find that financial incentives improved the performance of public sector health workers in Rwanda. Miller et al. (2012) document a decrease in anemia prevalence as a result of financial incentives to public sector workers in China. In contrast, Ashraf et al. (2014) find no effect of financial incentives on female condom distribution in Zambia.
The evidence on how pro-social incentives affect health worker motivation is scarce. Ashraf et al. (2014) find that non-financial social incentives increased health worker performance in Zambia more so than financial incentives. Studying the same CHW program used for the current study, Deserranno (2019) documents that pro-social CHWs perform better than more financially motivated CHWs.
To explore how an entrepreneurial model compares to free distribution in terms of CHW effort, we randomly assigned CHWs in Uganda to either sell oral rehydration salts and zinc (ORS + zinc)—highly cost-effective but underused treatments for child diarrhea—or distribute ORS + zinc for free. We gave CHWs in both study arms free ORS + zinc to distribute and asked them to visit all households in their catchment area to carry out their assigned interventions. CHWs in the home sales arm kept the sales revenue and, therefore, had a stronger financial incentive to put in more effort.
Our main result is that the entrepreneurial model led to substantially less CHW effort than free distribution. Only 35% of households with a CHW assigned to sell ORS + zinc (and retain the revenue) received a home visit, whereas 61% of households with a CHW assigned to free distribution received a home visit. Home visits (our primary measure of CHW effort) are the main job task of a CHW as nearly all job activities occur during these visits. In a rare example of a backward-bending labor supply curve, free distribution led to 74% increase in CHW effort relative to home sales. The low effort among CHWs assigned to the entrepreneurial model left a meaningful amount of money on the table (11%–57% of the average monthly income).
Qualitative interviews revealed the importance of social motivations. CHWs assigned to the entrepreneurial model reported that it was unpleasant to ask their neighbors, many of whom are very poor, to purchase something. They said they would have visited more households had they been distributing products for free. Moreover, CHWs assigned to free distribution reported feeling good about giving products away for free. They explained free distribution meant they were helping their community and it made them look good to their peers.
Prior analysis of this experiment documents that 18% more cases of diarrhea were treated with ORS when the CHW was assigned to free distribution compared to the entrepreneurial model (Wagner et al., 2019). We contribute to these results in several important ways. First, Wagner et al. (2019) focus on the household response to different barriers to ORS + zinc use and provide insight into what interventions are likely to increase ORS + zinc coverage (whether through CHWs or other distribution channels). The current paper focuses on the supply-side response to different CHW models. Thus, these results have implications for CHW program design beyond ORS + zinc. While there is a large body of evidence documenting that household demand is sensitive to the price of health products (see Kremer and Glennerster (2011) or Dupas and Miguel (2017) for reviews), this work is the first we are aware of to document that free distribution can also increase CHW effort relative to a sales model. These results call into question the notion that an entrepreneurial model necessarily increases CHW effort and provide further support for eliminating user fees for some products. Moreover, this work identifies free delivery of health products as a useful tool for motivating CHWs to make household visits, a crucial pillar of most CHW programs.
Second, the estimated effects of the interventions on ORS use reported by Wagner et al. (2019) capture both how prices affect household demand and how prices affect CHW effort. In this study, we shed light on the relative importance of the demand and CHW effort channels. We estimate that extra CHW effort caused roughly half of the increase in ORS use under free distribution.
We also go beyond Wagner et al. (2019) by investigating whether an entrepreneurial model could have efficiency benefits relative to free distribution. First, charging could improve efficiency though targeting to higher-risk cases if caretakers have higher willingness to pay for such cases. Second, charging could screen out people uninterested in using ORS + zinc and thus reduce wastage. We find mixed evidence on whether charging for ORS + zinc better targets higher-risk cases of diarrhea. Consistent with prior literature1, we find that the entrepreneurial model did not do a better job of targeting younger children, who have a higher mortality risk from diarrhea. However, there is some (imprecisely estimated) evidence that user fees bettered target children with ‘severe’ cases of diarrhea (defined as either blood in the stool or concurrent fever).
The entrepreneurial model did a modestly better job of screening out uninterested households. Of the caretakers who acquired ORS from the CHW, 89% used ORS to treat a case of diarrhea if distributed for free, compared to 94% if purchased (p = 0.06).2 However, this possible improvement in screening comes at the cost of substantially lower usage; 76% of cases used ORS under free distribution compared to 64% in the entrepreneurial model. As such, our results are consistent with prior research, which finds that moving from free to positive prices strongly reduces usage of health products (Kremer and Glennerster, 2011; Dupas and Miguel, 2017).
A final way in which we extend Wagner et al. (2019) is by estimating the efficiency gains from non-monetary hassle costs imposed by CHWs. In a recent study, Dupas et al. (2016) find that imposing small hassle costs for free point-of-use water treatment is a more efficient screening mechanism than prices, because it screens out non-users while not dampening demand among users. We build on (Dupas et al., 2016) by randomizing half of the CHWs assigned to free distribution to deliver ORS + zinc to the door and the other half to deliver vouchers that could be redeemed for free ORS + zinc at the CHW's home. Among households with a case of diarrhea, vouchers did not do a statistically distinguishable better job than free distribution of screening for caretakers who used ORS (no difference in take-up or coverage).3 However, hassle costs led to lower acquisition of ORS among households that did not experience a diarrhea episode, which could reduce wastage and implementation costs.
Taken together, our study suggests that, compared to free distribution, the entrepreneurial model that we test in this study reduces ORS + zinc coverage and reduces CHW effort, while only modestly improving screening. While other entrepreneurial designs might produce better results, this study sheds light on the importance of testing and comparing different CHW program design features. The rest of the paper proceeds as follows. Section 2 provides a background on CHW programs, sections 3 Study design, 4 Summary statistics provide an overview of the study design and sample characteristics, section 5 presents results on CHW effort, sections 6.1 Did prices or hassle costs improve targeting to high-risk cases?, 6.2 Did prices or hassle costs reduce wastage? present the targeting and screening effects of user fees, and section 7 concludes.
Section snippets
CHWs: effectiveness, incentives, and motivation
Community health workers are an important part of the health care system in many developing countries. Nearly all countries in sub-Saharan Africa and South Asia have a CHW program. CHWs are generally members of the community who live near the households they serve. CHWs off very basic health care, provide some health-related products, and refer severe cases to the formal health-care system.
In most nations the Ministry of Health funds the CHW program. For example, in Uganda, government sponsored
Study overview
This study was a cluster randomized controlled trial with random assignment occurring at the village level (118 villages). We worked with BRAC to select 6 branches in central Uganda to enroll in the study. Branches are local offices used to coordinate all of BRAC's operations in the surrounding villages. We then enrolled all villages affiliated with the selected branches where a CHW was active (about 20 per branch) resulting in 118 villages. All branches were within a 3-h drive from Kampala,
Summary statistics
Table 2 presents the characteristics of the sample and how these characteristics compare between arms. The sample size was fairly well distributed across treatment arms and randomization appears to have been successful at ensuring balance between arms. Caretakers in the home sales arm were slightly younger and slightly more educated. CHWs assigned to home sales had fewer households in their catchment areas, although this should not affect our estimates because our survey team only visited the
CHW effort
Our experimental design allows us to estimate the impact of an entrepreneurial model of ORS + zinc distribution on CHW effort relative to a free distribution model (either through home delivery or vouchers). We use the following equation to estimate the unadjusted differences in CHW effort between treatment arms:where yiv is an outcome related to CHW effort (mainly whether the household was visited) for household i in village v at endline. The terms Delivery
Other potential benefits of entrepreneurial CHW models: targeting and screening
We have so far demonstrated that the entrepreneurial CHW model of ORS + zinc dispensing that we tested performed worse in terms of both CHW effort and household ORS + zinc use relative to free delivery. However, there are other potentially important benefits of entrepreneurial models that are not captured by our analysis of effort and use. In particular, having CHWs charge for ORS + zinc could produce efficiency gains by targeting treatment to the highest-risk cases. Presumably, caretakers of
Summary
Understanding how to motivate CHWs is of growing importance. This study provides a head-to-head comparison of competing models for motivating CHWs to distribute health products and make household visits. Asking CHWs to distribute ORS + zinc for free induced more effort and better performance than asking CHWs to sell ORS + zinc door-to-door. In this setting, the social incentives associated with providing a household with free treatment for child diarrhea appear to be more powerful than the
Conclusion
Entrepreneurial CHW models are growing in popularity. This is at odds with a large body of economics literature that documents that households are very sensitive to the price of health products (Kremer and Glennerster, 2011; Dupas and Miguel, 2017). Our study suggests that an entrepreneurial model for ORS + zinc dispensing not only dampens demand but also leads to less CHW effort than a model of free distribution. Implementors of CHW programs should think carefully before incorporating
CRediT authorship contribution statement
Zachary Wagner: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Validation, Visualization, Writing - original draft, Writing - review & editing. John Bosco Asiimwe: Conceptualization, Funding acquisition, Investigation, Project administration, Supervision, Writing - review & editing. David I. Levine: Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Supervision,
References (37)
- et al.
No margin, no mission? a field experiment on incentives for public service delivery
J. Public Econ.
(2014) - et al.
Effect on maternal and child health services in Rwanda of payment to primary health-care providers for performance: an impact evaluation
The Lancet
(2011) - et al.
Therapeutic effects of oral zinc in acute and persistent diarrhea in children in developing countries: pooled analysis of randomized controlled trials
Am. J. Clin. Nutr.
(2000) - et al.
Impacts and determinants of health levels in low-income countries
- et al.
Human resources for health: overcoming the crisis
The Lancet
(2004) - et al.
Improving health in developing countries
Handb. Health Econ.
(2011) - et al.
regional, and national causes of child mortality in 2000–13, with projections to inform post-2015 priorities: an updated systematic analysis
The Lancet
(2015) - et al.
Explaining causal findings without bias: detecting and assessing direct effects
Am. Pol. Sci. Rev.
(2016) - et al.
Effects of the integrated community case management of childhood illness strategy on child mortality in Ethiopia: a cluster randomized trial
Am. J. Trop. Med. Hyg.
(2016) - et al.
Independent evaluation of the integrated community case management of childhood illness strategy in Malawi using a national evaluation platform design
Am. J. Trop. Med. Hyg.
(2016)
Can higher prices stimulate product use? evidence from a field experiment in Zambia
Am. Econ. Rev.
The moderator–mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations
J. Personal. Soc. Psychol.
Community Health Worker Incentives and Disincentives: How They Affect Motivation, Retention, and Sustainability. Published by the Basic Support for Institutionalizing Child Survival Project (Basics Ii) for the united states Agency for International Development
Reducing child mortality in the last mile: experimental evidence on community health promoters in Uganda
Am. Econ. J. Appl. Econ.
Community health workers and mobile technology: a systematic review of the literature
PLoS One
Free distribution or cost-sharing? evidence from a randomized malaria prevention experiment
Q. J. Econ.
Financial incentives as signals: experimental evidence from the recruitment of village promoters in Uganda
Am. Econ. J. Appl. Econ.
Targeting health subsidies through a nonprice mechanism: a randomized controlled trial in Kenya
Science
Cited by (0)
- ☆
We are grateful to Will Dow, Paul Gertler, Pascaline Dupas, Jack Colford, Neeraj Sood, Lia Fernald, Edward Miguel, Aleksandra Jakubowski, Eran Bendavid, Drew Cameron and Munshi Sulaiman for helpful comments. We would like to thank BRAC, particularly Sharmin Sharif and Robert Mpiira, for administrative support. Finally, we would like to thank Betty Namutebi, Sam Njunwamukama, and Francis Bbosa for excellent research assistance. A pre-analysis plan for this study was pre-registered at the American Economic Association's registry for randomized controlled trials (registry number AEARCTR-0001288). This study was funded by the EASST Collaborative and Weiss Family Program Fund (PI: Wagner).