Long-run Consequences of Health Insurance Promotion When Mandates are Not Enforceable: Evidence from a Field Experiment in Ghana

Many poor households in developing countries lack access to health insurance, and their poverty is exacerbated by health-related problems (Dercon, 2002). In the absence of insurance, households bear a high proportion of medical expenses in the form of out-of-pocket (OOP) payments, and face financial constraints that act as significant barriers to health care access (Gertler and Gruber, 2002; Xu et al., 2003; Wagstaff, 2007). 1 Many developing countries have increasingly been instituting social health insurance schemes (SHIs) to help mitigate the effects of adverse health shocks, especially for the poor (WHO, 2005, 2010). 2 However, even though SHIs are theoretically mandates and offer low sign-up costs and generous benefits to increase enrollment, take-up and retention rates remain very low in many countries (Fenny et al., 2016), especially among the poorest households (Acharya et al., 2013). 3 Achieving universal coverage or a high enrollment rate is important in terms of risk pooling and the sustainability of social health insurance; however, it is often difficult for developing countries to successfully impose mandates, primarily due to administrative constraints. For example, if those who are more ill or with larger health care service utilization are selected into social health insurance, financial burden of the program will increase and become less sustainable.

There have been various efforts to promote health insurance enrollment and health care utilization, but many recent studies find such efforts have limited impact. 4 Even after successfully promoting health insurance enrollment in the short run, retention and sustainable improvements in health service utilization and health status remain a challenge.

The long-run effects of an intervention have important implications for policy. For example, an increased retention rate may yield economic and health benefits when individuals engage in health services on a regular and timely basis, which may improve the sustainability of the health insurance program. Nevertheless, this topic remains relatively understudied.

A subsidy is one of the few successful types of interventions used to promote health insurance enrollment (e.g., Thornton et al., 2010). However, an important question emerges regarding the level of subsidy. Different levels of subsidy (price) may attract people with different characteristics, and this selection may affect health care service utilization and health outcomes among the insured. 5 The screening effect of subsidy level has been studied for a few health products and services, such as facility delivery (Grepin et al, 2019), malaria bed nets (Cohen and Dupas, 2010), and chlorine for water purification (Ashraf, Berry, and Shapiro, 2010), but has not been investigated for health insurance in a developing country setting. 6 This study aims to fill this gap in the literature through a field experiment. Similar to Kremer and Miguel (2007) and Dupas (2014), we employ experimental variations in exposure to a health product and follow the behavioral response in the long run. We randomly selected communities for the subsidy intervention and randomized different levels of subsidy (one-third, two-thirds, and full subsidy) for the insurance premiums and fees for one-year’s coverage at the household level. To measure the impact of these interventions, we conduct a baseline survey and two follow-up surveys, one at seven months and the second at three years after the initial intervention.

This experiment has three main objectives. First, we study whether a subsidy for premiums and fees promotes health insurance enrollment in the short run. Second, we study whether the level of subsidy affects health insurance enrollment, health care service utilization, and health status to shed light on the potential selection effect of the level of health insurance subsidy. Third, we study whether a one-time intervention could have sustainable impacts on health insurance enrollment, health care service utilization, and health outcomes in the long run.

Three sets of results emerge. First, we find a significant increase in short-run insurance take-up. Those receiving one-third, two-thirds, and a full subsidy were 39.3, 48.3, and 53.8 percentage points, respectively, more likely to enroll in health insurance in the short-run. Three years after the initial intervention, we still observe increased enrollment. Those who received one-third, two-thirds, and a full subsidy were 17.3, 14.0, and 18.9 percentage points, respectively, more likely to enroll in health insurance in the long run.

Second, we find evidence of selection. Those who enrolled due to our intervention (compliers) are more ill and have larger health expenditures than those who did not enroll regardless of intervention (never-takers). Among compliers, individuals in the partial subsidy group are particularly more ill and have larger health expenditures than those in the full subsidy group. This