The Effects of Health Insurance within Families: Experimental Evidence from Nicaragua

Health insurance is thought to be an important determinant of healthcare utilization and health outcomes, particularly among children (Olson, Tang, and Newacheck 2005; Kaiser Family Foundation 2002). Studies have shown that children’s health insurance is positively correlated with immunization rates, checkups, and overall healthcare utilization; insured children may also have improved health outcomes and lower mortality (Abdullah et al. 2010; Currie and Gruber 1996; Newacheck, Hughes, and Stoddard 1996, 1998; Palmer et al. 2015; Todd et al. 2006). However, causally attributing these observed differences in outcomes to health insurance coverage is difficult due to adverse selection: parents may choose to enroll in coverage if they have children who are likely to be sick (Polimeni and Levine 2011). On the other hand, parents with more resources or education may be more likely to have health insurance coverage for their children, resulting in advantageous selection (Fang, Keane, and Silverman 2008). Studies examining the effects of health insurance coverage on outcomes are complicated by potential biases due to selection into enrollment.

 This paper overcomes the usual empirical challenges to measure the causal effects of health insurance using an experiment that randomly assigned parents in Managua, Nicaragua, access to free health insurance. We study the effects of insurance among children (age 15 and under). Although recent studies on health insurance also utilize random or quasi-random research designs, these either occur in the United States or estimate impacts only among adults (Banerjee, Duflo, and Hornbeck 2014; Barofsky 2015; Finkelstein et al. 2012; King et al. 2009; Levine, Polimeni, and Ramage 2016; Newhouse 1993; Sheth 2014). The results from these existing studies may not apply to children or to the developing country context because the elasticity of healthcare demand has been found to be larger in the developing country setting and larger for children as compared to adults (Zhou et al. 2011; Gertler and Van der Gaag 1990; Sauerborn, Nougtara, and Latimer 1994).

In addition to analyzing effects among insured children, we also examine how having an insured parent affects healthcare utilization among children who are ineligible for the insurance themselves. This allows us to see how families adjust their health demands in response to parental insurance coverage. A model of healthcare demand within a household predicts that parents efficiently optimize, equating the ratio of marginal benefits to costs across all family members, rather than separately for each individual member (Jacobson 2000). If one member becomes insured, the net effect upon uninsured members’ utilization is ambiguous if parents reallocate time or financial resources across household members in response to the decreased cost of utilization. The net outcome may be either toward or away from uninsured children, depending upon whether the income or substitution effect dominates. Alternatively, following health insurance enrollment of some members, parents could reallocate resources to increase utilization for uncovered members to ensure equity across their children (Adhvaryu and Nyshadham 2014; Ejrnæs and Pörtner 2004; Griliches 1979; Hertwig, Davis, and Sulloway 2002). The empirical literature measuring the effects of parental health insurance enrollment thus far has generated mixed results, and none are based upon randomized studies.¹

To study the effects of insurance on insured and uninsured children, we utilize an age-eligibility cutoff rule that gave coverage only to children under the age of 12. Children age 12 and older were not eligible for insurance coverage. The data include 1,614 families in Managua, Nicaragua, who were interviewed before and after being randomly offered access to free insurance. Our primary analysis measures the impact of parental health insurance on healthcare utilization—measured by number of provider visits and choice of health provider—of eligible and ineligible dependents one year after access to insurance is randomly distributed.

Among families with an insured parent, covered children (under 12 years old) increased healthcare utilization at covered providers by 0.56 visits, with an increased 1.3 visits to all providers. The results suggest no substitution away from other providers such as pharmacies or public facilities. In contrast, ineligible children (12 years old and older) with an insured parent decrease their total health visits by 1.7, driven primarily by reductions in visits to private providers.

We examine several potential mechanisms to explain these results. First, we find that the results are primarily a result of a parent’s enrollment rather than that of siblings. Second, we find no evidence that resources are reallocated in response to improved health of those insured.

While ultimately the exact mechanisms cannot be identified, the results show that public programs such as government-sponsored health insurance can have sizeable effects on the allocation of resources within a family. While insured children benefit by receiving additional healthcare, ineligible members of a family may be negatively affected because families use shared budget and time constraints to make decisions within their households. These potential spillover effects may change the effectiveness of health policies and programs (Basu and Meltzer 2005).

The findings also suggest important complementarities across provider types among children. For insured children, we observe no substitution from pharmacies or private providers in response to health insurance enrollment but rather a sizable increase in utilization across all providers. These results suggest that reducing the price of healthcare at one provider may increase investments in healthcare across other providers. However, complementarities may also exacerbate negative spillovers, for example, among the uninsured.

We caution that the results may be specific to the Nicaraguan context and healthcare structure. The effects on this specific sample—children of middle-class urban entrepreneurs—may not be generalizable to other populations. The analyses in the paper are also limited in the set of health indicators that are measured within the setting and time frame. The free health insurance was only for six months, which may not have been enough time for there to be large effects, and the scope of our study is not able to detect changes in rare events such as child mortality. This study is also unable to speak to potential longer-run effects of health insurance, or whether these results would apply to different subsidies or cost-sharing schemes. However, this paper underscores the importance of considering negative spillovers within families; positive benefits for some may result in net losses for others.

This paper is structured as follows: In section 1, we provide background on the health insurance program and study design. In section 2, we outline the empirical strategy. In section 3, we present our main results on utilization. In section 4, we examine potential mechanisms driving our results. Section 5 concludes.

1. Background

In Nicaragua, formal-sector employees are automatically enrolled in the Nicaraguan Social Security Institute’s (INSS) health insurance program. The INSS insurance provides subscribers with a comprehensive package of preventive, diagnostic, and curative health services and medications at 17 INSS-contracted facilities in Managua (referred to as EMPs: Empresas Médicas Previsionales).² There are no co-pays at the time of service; rather, individuals who enroll pay a monthly flat fee of approximately |${\$}$|15 to the Social Security Institute for coverage. In addition to the subscriber, the subscriber’s wife is eligible for maternity services, including prenatal, childbirth, and postnatal care. Dependent children under the age of 12 are also fully covered for pediatric care and vaccinations, while those over the age of 12 are excluded from coverage.³

While this insurance plan covers those in the formal sector, this represents only a small proportion of the adult population—just under half a million adults, or approximately 13.5 percent of the adult population (INIDE and MINSA 2007). Uninsured individuals have access to free public-sector clinics and hospitals run by the Ministry of Health Services (MINSA). However, these services are often under-resourced and the source of complaints of long waiting times, frequent supply shortages, and general poor quality (Magnoni, Natilson, and Bolaños 2005). Rather than seeking treatment at MINSA facilities, self-medication from pharmacies for basic care is common.⁴ For those who can afford it, higher-quality, more expensive private facilities are available. On the other hand, many may be unable to pay the high out-of-pocket costs and forego care altogether. Thus, the INSS health insurance package may not only change the cost but also the quality of healthcare for insured individuals.

In January 2007, the government of Nicaragua implemented a demonstration project aimed at extending the Nicaraguan Social Security Institute’s (INSS) health insurance program to the population of informal-sector workers. In addition to potentially improving healthcare access and overall health, enabling middle-class informal-sector workers to seek care at EMPs has the potential to alleviate the overburdened MINSA healthcare system, leaving more resources for the poor. Coverage and cost was designed to be as similar as possible to the standard insurance for formal-sector workers, and enrollment into the program was voluntary. Both this paper and Thornton et al. (2010) report results from this evaluation. In 2007, a baseline survey was conducted among randomly selected uninsured informal-sector workers in the three largest open-air markets in central Managua.⁵ The survey asked detailed questions about utilization and spending, conditional on being ill 12 months prior to the survey, for the adult respondent and for any child in the household under the age of 16.

At the end of the baseline survey, respondents were given either an informational brochure about the insurance product or the brochure plus a six-month subsidy for insurance worth approximately USD|${\$}$|100, nearly half of the sample’s median household income.⁶,⁷ Respondents could enroll in the insurance plan at the INSS or at local microfinance institutions, and there was no deadline for enrollment.⁸ Upon enrolling, the insurance took effect the first day of the following month. Government ID numbers were collected to match respondents to health insurance enrollment data provided by the INSS. One year after the baseline survey, respondents were approached for a follow-up survey. Overall, 93 percent of the adult respondents were re-interviewed (N = 2,608). Of the 2,608 adults who were in both rounds of the initial evaluation, 62 percent had at least one child under the age of 16 (N = 1,614).

The overall take-up rate of insurance for respondents with insurance-eligible children was 35 percent among those who were offered the six-month subsidy and 2.22 percent among those who were not.⁹ The primary sample consists of 2,996 children in 1,614 households in both waves of the survey; this sample is 74 percent of all children at baseline. As with the full adult sample in Thornton et al. (2010), there was no differential sample attrition between those who were offered the subsidy and those who were not (available upon request).

Baseline statistics of households (panel A), parents (adult respondents with at least one child under 16, panel B), and children (panel C) are presented in table 1. Households have on average 4.8 members, including 1.3 children age 11 and under. Average monthly household income is 5,399 Córdobas (US|${\$}$|298). Parents have 9.3 years of education and work in their market stall on average ten hours per day; 76 percent of adults report ever being sick in the past year, with a similar percentage ever seeing a health provider (76 percent). The average number of visits to all providers is 4.28. Average health expenditures are C|${\$}$|828 (US|${\$}$|46), with a median value of C|${\$}$|161 (US|${\$}$|8). Children have similar rates of being sick and utilization of healthcare services as their parents, although total costs are somewhat lower: C|${\$}$|569 (US|${\$}$|31), with a median value of C|${\$}$|204 (US|${\$}$|11).

2. Empirical Strategy

Characteristics of children whose parents were and were not offered the subsidy are balanced across the treatment groups (table 1, panel C). Magnitudes of the differences are also small. Characteristics are also balanced by treatment group separately among eligible and ineligible children and across smaller age groups (0–5 and 6–11) (supplementary online appendix table S1.2).

3. Results

Before presenting the main effects of the insurance on children, we report the effects of being insured among parents according to equation (1) as a benchmark (supplementary online appendix tables S1.3 and S1.4). Enrolled parents are 41 percentage points more likely to attend an EMP (covered provider) and 13.7 percentage points less likely to attend public facilities. Similarly, parents increase the number of visits to EMPs by 1.3 visits. The overall number of visits increases by 0.8 visits, but this estimate is not statistically significant.¹³

Turning to the effects of insurance among children, table 2 shows that having an enrolled parent increases the likelihood that an eligible child visits a covered provider (EMP). The increase is large at 23 percentage points (table 2, panel A, column 3), although not as large as the increase in adults’ own utilization. There is also an increase on the intensive margin with an increase in 0.56 visits to an EMP (panel B, column 3). In addition, eligible children with an insured parent increase total utilization at all providers, not just those that are covered by the insurance, by 1.26 additional visits (panel B, column 1). Note that the magnitude of the effect size on EMP utilization is about half of that among parents, although there are larger effects on children’s total utilization.

Age-ineligible children (ages 12–15) with parents enrolled in health insurance experience fairly substantial decreases in the likelihood of visiting providers, and the total number of provider visits, compared to children of the same ages, whose parent is uninsured. Among these children, having an insured parent reduces the likelihood of attending a private facility by nearly 30 percentage points (p-value = 0.003; column 5) and reduces the total number of healthcare provider visits by 1.8 (p-value = 0.048; table 2, panel B, column 1). These results are robust to examining the difference in utilization as an outcome, as in a panel data IV model (supplementary online appendix table S1.5).¹⁴

The reduced-form estimates measuring the impacts of offering health insurance to parents on children are presented in table 3.¹⁵ Eligible children are seven percentage points more likely to have ever visited an EMP in the past year, while ineligible children in insured families are 6.5 percentage points less likely to have ever visited EMPs (panel A, column 3). Among ineligible children, being in an insured family decreases the likelihood of ever visiting a private health facility by 0.078 percentage points (panel A, column 5). While offering parents health insurance increases overall utilization among eligible children by 0.392 visits (panel B, column 1), including 0.174 at covered providers (panel B, column 3), there is a large and significant decrease in utilization among ineligible children in insured families. As a result of health insurance, ineligible children have 0.847 fewer visits overall (panel B, column 1), including 0.162 fewer visits at covered providers and 0.292 fewer visits at private facilities (panel B, column 5).¹⁶

4. Possible Channels

While the analysis shows that the effects of parental insurance differ by the insurance status of the child, the previous analysis does not explain why these patterns arise. Are parents reallocating time or financial resources between household members in response to the price decrease from enrollment? Or are children healthier and therefore imparting a positive externality on their siblings, reducing the demand for healthcare?

First, we examine whether the effects of insurance are in response to having enrolled parents, siblings, or both. While all ineligible children in insured families have an insured parent, not all have an insured sibling. Using the sample of ineligible children (N = 824), we test whether the effect of parental insurance differs by whether the child has an eligible sibling.¹⁷ The results in table 4 show that the effects of health insurance do not differ by whether the ineligible child has a covered sibling, suggesting that the results are primarily driven by having insured parents. There are no significant added effects of having an eligible sibling.¹⁸

An alternative mechanism consistent with the results is that, if health insurance improves health, there may be positive health externalities among uninsured members of the household. As a result, ineligible siblings might not require as many visits to healthcare providers. Unfortunately, the data do not contain objective measures of children’s health and are restricted to parents’ self-reports, which may be systematically biased due to increased access to more highly qualified providers or increased/decreased healthcare utilization. In addition, any results presented on the effects of insurance on health status—a function of both preventative and curative care—are not comparable to the healthcare utilization results, which are conditional on being sick.

Nevertheless, we find no significant effects of parent insurance on the likelihood of ever being sick among parents or eligible children (supplementary online appendix table S1.13). On the extensive margin, however, eligible children with insured parents are reported to be sick 0.67 times more than those without an insured parent. In contrast, ineligible children with insured parents are reported to be sick 1.06 times less than ineligible children without insured parents. There are several possible explanations for these results. It is possible that health insurance actually made eligible children with insured parents sick more often despite their increased utilization at providers through iatrogenic illnesses from waiting rooms, for example (Steel et al. 2004). However, more plausibly, increased healthcare utilization may also have increased parents’ knowledge, or salience, of sick episodes, or changed the criteria for which a child is sick enough to attend the doctor.

There are several possible explanations for the increase in non-covered provider visits among eligible children with insured parents. The increased overall utilization is not likely due to increased available money for more healthcare—we found no significant effects of the insurance on out-of-pocket expenditures (supplementary online appendix table S1.6). The increased utilization could be due to moral hazard, because the price decrease is specific to EMPs. Increased utilization could also have resulted in additional diagnoses for covered children. Economies of scale in time—taking children to the doctor or visiting a pharmacy—could also explain the results. Unfortunately, data on time costs, or when visits to the different providers took place, were not collected. While the data do not allow us to definitively identify the mechanism, the results are most consistent with the Jacobson (2000) model, where the substitution effect from cheaper care for insured members dominates the income effect for uninsured members of the household.

5. Conclusion

This paper estimates the direct and indirect causal effects of health insurance on children’s utilization, using a randomized allocation of insurance subsidies and an age eligibility cutoff. Children who were covered by their parent’s insurance have substantially more visits at covered providers and increase their total number of visits at all providers combined. In contrast, children who were ineligible for coverage due to an age restriction substantially decreased their overall utilization.

The results do not support a model in which parents are concerned with equity of resources. Instead, parents appear to be driven by efficiency. Parents respond to health insurance by decreasing utilization among those for whom healthcare is relatively more expensive. The findings suggest that the marginal value of reduced healthcare costs and/or increased returns to health investments (better-quality service at covered providers) for the insured are greater than the loss associated with reduced provider visits for older children. Unfortunately, we lack the data to test whether the change in utilization among eligible and ineligible children is due to economies of scale with respect to travel or time costs, such as parents taking insured children with them to the doctor.

The findings in this paper have implications for providing families with insurance rather than separate individual insurance plans for children or adults. The results suggest that the primary benefit of parental health insurance to children is improved access to care, but only for eligible children.

Insured children not only substitute from uncovered providers to covered providers but also increase their overall utilization, suggesting important complementarities in healthcare utilization across provider types. Healthcare subsidies could potentially be used to promote further investments in children’s health. At the same time, negative health shocks or spillovers may be exacerbated and increase health inequalities, particularly if programs are not designed to cover all family members.

In contrast to results for adults, there is no substitution away from public facilities and toward covered providers. Thus, children’s health insurance is unlikely to decrease demand at overburdened public facilities, a stated goal of the program. These findings suggest an unmet need for healthcare among children covered by the program.

The results of this paper highlight that families allocate resources—including health demands—according to a specified budget and/or time constraints. These changes in health demands may be sizeable as well: in our context, the impact of health insurance on ineligible members is larger than the impact on eligible members, although in opposite directions. While this study is not powered to detect rare or serious ailments among ineligible children, one may assume that reducing healthcare utilization in a resource-poor setting could potentially have true negative consequences on health.

As health insurance programs, including that of the United States, have moved toward covering children, it is important to understand the net benefits and the costs to families from parental health insurance. When considering the effect, and the cost-effectiveness of health insurance as a social policy, it may be empirically important to account for the positive as well as the negative effects that influence healthcare demands on all members of the household.

Author Biographical

Anne Fitzpatrick is an Assistant Professor at the University of Massachusetts–Boston; her email address is [email protected]. Rebecca Thornton (corresponding author) is an Associate Professor at the University of Illinois at Urbana-Champaign; her email address is [email protected]. Funding for this study was provided by USAID’s Private Sector Partnerships-One (PSP-One) project, the Global Development Network (GDN), and the ILO Microinsurance Facility. The evaluation was coordinated for GDN by EA Consultants. The authors thank the extensive contributions of the field team at ALVA Consultores, including Dr. Ana del Carmen Rojas and Rosario Duarte, as well as the project coordination and support of Barbara Magnoni, EA Consultants, and the contributions in the project design of Tania Dmytraczenko, World Bank. This paper has benefited from comments from Taryn Dinkelman, Emily Oster, and Jeff Smith, seminar participants at the University of Michigan; and two anonymous referees. Funding for this study was provided by the US Agency for International Development’s Private Sector Partnerships-One (PSP-One) project, the Global Development Network (GDN), and the ILO Microinsurance Facility. A supplementary online appendix for this article can be found at The World Bank Economic Review website.

Footnotes

References