This is the second in our series of posts by students on the job market this year.
Relaxing supply-side constraints is not always sufficient to ensure delivery of public services to poor and remote communities. It may be necessary to stimulate demand by exploiting local agents who can link the relevant parties. We thus see the use of intermediaries in a variety of sectors in development; for example through the use of agricultural extension agents (Anderson 2004), loan officers for microfinance (Siwale 2011), and referral incentive programs – like that used by the British colonial army in Ghana (Fafchamps 2013). My job market paper studies the use of intermediaries in the maternal health sector in the Western Province of Kenya. I use an RCT to evaluate the efficacy of financial incentives for Traditional Birth Attendants (TBAs). The program provides payments for TBAs to encourage pregnant women to attend antenatal care (ANC) visits at a local health facility. In this way, TBAs link pregnant women with health facilities, the TBAs’ rivals. This potential competition, which is absent from most intermediary relationships, is a noteworthy feature of this program as it creates a nontrivial incentive problem for the TBA.
Background and Program
The majority of women in rural Kenya fail to meet recommendations for ANC timing and use. Proper care during pregnancy is important for the health of women and their babies. The WHO estimates that between 1/3 to 1/2 of all maternal deaths are from causes related to inadequate care during pregnancy (Lawn 2006). In developing countries, for both ANC and delivery services, women have a choice between seeing a nurse or doctor at a formal health facility or seeing a TBA in their own village. TBAs are community-level agents similar to, but less skilled than, midwives; they serve as informal providers of pregnancy care and operate independently from health facilities. As part of the RCT, village-level randomization was performed across all 172 villages in Vihiga District, such that a TBA in a treatment village would receive financial incentives for encouraging pregnant women to seek ANC at a health facility. The pilot program operated for 9 months (June 2012 − March 2013) and program staff enumerated all of the TBAs in the district, 474 in total; of these, 231 were in treatment villages. I evaluated the impact of the TBA program on the health-seeking behavior of 1,130 pregnant women in the same district.
By paying TBAs for encouraging women to visit ANC clinics, this program breaks with tradition by integrating TBAs with formal health facilities. However, the potential competition between the TBAs and the health facilities renders the effect of the program an empirical question. Even though the incentive payment was large (it represented approximately 25% of what a TBA would receive for a delivery and double what the TBA would receive for conducting her own pre-delivery visit), there was no guarantee that the TBAs would respond to it. If clients encouraged to visit the health facilities for ANC subsequently decide to visit those same facilities for delivery, the TBA might withhold encouragement to increase her chance of retaining the clients for future delivery services, despite the financial incentive. My research thus examines the use of intermediaries in a situation where program incentives potentially oppose the targeted parties’ livelihood interests.
I examine the effect of the program on whether the women attended at least their recommended number of ANC visits. A woman’s number of recommended visits was calculated based on Kenya’s Ministry of Public Health and Sanitation’s schedule regarding the gestation during which ANC visits should occur. The outcome indicator accounts for the fact that, at the start of the program, women were at various stages of their pregnancy. That is, women with less remaining time in their pregnancies would have fewer recommended visits. My findings suggest that living in a TBA treatment village increases the likelihood of a woman attending at least the recommended number of ANC visits by 21 percent.
Furthermore, by examining heterogeneous effects based on characteristics of the TBAs, I find that the program was more successful in places where the TBAs had options beyond working as birth attendants – the program effect was positive in places where the TBAs were more educated and negative in places where the TBAs were less likely to have training and were less experienced. This suggests that client loss may have been a concern for TBAs, despite the fact that there was no evidence of an effect of the program on the rate of deliveries in health facilities.
Interestingly, I find no effect of the program on the total number of ANC visits made. Since the average number of ANC visits was the same between women in the treatment and control groups, I next consider the distribution of visits by recommended visit number, which should vary depending on the stage of pregnancy at the start of the program. The figure below groups women according to their recommended number of visits. It reveals that the positive treatment effects are driven by women who had 2 recommended visits, i.e. women towards the end of their pregnancy but not at the final stage. Furthermore, the figure reveals that negative treatment effects exist for women who had 1 recommended visit – i.e. those at the last stage of their pregnancy. Finally, the figure suggests that there was no treatment effect for women who were recruited early in their pregnancies and had 3 or 4 recommended visits.
One possible explanation for the negative and positive treatment effects for women with 1 or 2 recommended visits, respectively, is that the TBA may have been concerned about encouraging women too close to delivery. If women in advanced pregnancy gestation who attend ANC are more likely to deliver in a health facility (as opposed to with a TBA), the TBA might focus encouragement on women with more remaining time in their pregnancies, for instance women with 2 recommended visits as opposed to those with only 1. However, this explanation does not clarify why I do not find an effect of the treatment on women with 3 or 4 recommended visits.
My findings suggest that despite potentially opposed incentives, TBAs can serve as useful intermediaries in the maternal health sector by linking patients to formal health facilities for ANC. However, program designs that more carefully consider the underlying incentives of the intermediary are likely to have greater success. I find evidence that TBAs are concerned about the potential loss of clients – the TBAs most likely to respond to the incentive program are those less locked into the role of a TBA, and their response appears to encourage women with less risk of subsequently delivering in a health facility.
Nevertheless, programs of this nature can help address maternal health challenges. Despite Kenya being the largest and most advanced economy of East Africa, maternal and infant health outcomes are typical for those of other sub-Saharan African countries; therefore the results estimated here may be generalizable to the wider region. Finding an increase in the likelihood of a woman attending the recommended number of ANC visits suggests that the use of intermediaries has the potential to improve maternal and infant health outcomes.
Nisha Rai is a Ph.D. candidate in the Department of Economics at Georgetown University and is on the job market this year. Her primary research interests concern applied health and education questions in the context of developing countries.