The Effects of Access to Health Insurance: Evidencefrom a Regression Discontinuity Design in Peru∗

Noelia Bernal, Miguel A. Carpio and Tobias J. Klein†

February 2016

Abstract

In many countries large parts of the population do not have access to health insurance.Peru has made an effort to change this in the early 2000’s. The institutional setup givesrise to the rare opportunity to study the effects of health insurance coverage exploitinga sharp regression discontinuity design. We find large effects on utilization that are mostpronounced for the provision of curative care. Individuals seeing a doctor leads to increasedawareness about health problems and generates a potentially desirable form of supplier-induced demand: they decide to pay themselves for services that are not part of the benefitpackage.

Key words: Public health insurance, informal sector, health care utilization, regressiondiscontinuity design, supplier-induced demand.

JEL-classification: I13, O12, O17.

∗We are grateful to Eric Bonsang, Dimitris Christelis, Pilar García Gómez, seminar participants at VanderbiltUniversity, Erasmus University Rotterdam, Tilburg University and Diego Portales University, as well as partici-pants of the First Annual Congress of the Peruvian Economic Association, the IX Annual Meeting of the LatinAmerican and Caribbean Economic Association (LACEA) and the 2015 Netspar International Pension Workshopfor helpful comments. Maria Eugenia Guerrero, Nicolas Dominguez and Ana Claudia Palacios provided excellentresearch assistance. We would also like to thank Juan Pichihua from the Peruvian Ministry of Economics andfunctionaries of SIS and MINSA for providing useful information and for their comments and suggestions.

†Bernal: University of Piura; e-mail: noelia.bernal@udep.pe. Carpio: University of Piura; e-mail:miguel.carpio@udep.pe. Klein: Netspar, CentER, Tilburg University; e-mail: T.J.Klein@uvt.nl. The first authoris grateful for financial support from Netspar and the Sociale Verzekeringsbank.

1

1 Introduction

In developing countries, a large number of individuals is not covered by health insurance(Banerjee et al., 2004; Banerjee and Duflo, 2007). The reasons for this are manifold. Onthe one hand, individuals are often used to relying on informal forms of risk-sharing instead ofbeing covered by formal health insurance and therefore do not demand insurance.1 On the otherhand, it has in the past not been seen as the role of the government to provide health insurance.Moreover, the World Health Organization and the World Bank stress that even when there ispublic health insurance then it often does not reach large parts of the population and especiallynot the poorest families because it is only provided to the minority of employees in the formalsector (WHO, 2010; Hsiao and Shaw, 2007). For instance, until the early 2000’s, less than 20percent of the individuals in Peru had health insurance.

This may be a cause of concern, because health insurance does not only protect individu-als against catastrophically high health expenditures (Wagstaff and Doorslaer, 2003). It alsoencourages them to see a doctor instead of simply buying medication, and thereby promotesappropriate treatment of illnesses that is often argued to be absent (Commission on Macroeco-nomics and Health, 2001; International Labour Office et al., 2006).

In reaction, many low and middle income countries have recently introduced Social HealthInsurance (SHI) targeted to the poor, with the goal to improve their health and to provide themwith financial protection against the financial consequences of health shocks. Coverage by SHImay or may not be free and implies that individuals receive medical attention from a serviceprovider. The costs are usually paid out of a designated government budget that is completelyor partially funded by taxes.

However, to date, it is not well understood through which channels health insurance cover-age contributes to the well-being of individuals and how this relates to the incentives providedto health care providers and patients.2 Important questions in this context are to what extent itis possible to encourage individuals to seek medical attention rather than simply buying med-ication in a pharmacy, how they can be motivated to invest into preventive care, and what theeffects of medical attention are on out-of-pocket spending and health outcomes.

Answering those questions is challenging for at least two reasons. First, we lack detaileddata on health care utilization and health outcomes, and second, it is challenging to control forselection into insurance. The second problem means that a regression of utilization or outcomemeasures on insurance coverage will yield biased results and will not estimate the causal effectsof health insurance. In this paper, we make progress in both directions. We use unusually richdata from the National Household Survey of Peru (“Encuesta Nacional de Hogares”, ENAHO)

1See for instance Fafchamps (1999), Jowett (2003), Chankova et al. (2008), Giné et al. (2008) and Dercon et al.(2008).

2See for instance Abel-Smith (1992), International Labour Office et al. (2006), Pauly et al. (2006), and Acharyaet al. (2013). Also for developed countries, our understanding in that respect has increased substantially over thelast decades, but is still far from being complete.

2

to evaluate the impact of access to the Peruvian SHI called “Seguro Integral de Salud” (SIS) forindividuals outside the formal labor market on a variety of measures for health care utilization,self-reported health indicators, and out-of-pocket expenditures. We account for selection byexploiting a sharp regression discontinuity design.

The Peruvian case is interesting because SIS resembles Western public health insurancesystems and private insurance products in that it covers health care expenditures, but does notstrongly incentivize individuals to invest into preventive care. Coverage is free for eligibleindividuals, and those who are not covered by SIS typically lack insurance coverage.3 SIS wascreated in 2001 and subsequently reformed. Prima facie, these reforms have been successful, ascoverage by SIS is substantial and the fraction of the population making use of it has increasedfrom 20 percent in 2006 to more than 40 percent of the total population in 2011, reachinga relatively high rate among the SHI programs in low and middle income countries (Acharyaet al., 2013). Yet, even though aggregate data suggest that some health outcomes improved sincethe program has been implemented—between 2000 and 2010 total maternal and child mortalityrates decreased from 185 to 93 and 33 to 17, per 100,000 and 1,000 thousands of children bornalive, respectively4—to date there is no study evaluating the effects of insurance coverage onpreventive care, health care utilization, out-of-pocket expenditures and health outcomes at themicro level that controls at the same time for selection into insurance.5 In this paper, we userich individual-level data to provide such an evaluation.

In our paper, we make use of the opportunity to control for selection by exploiting theinstitutional setup in Peru that gives rise to a sharp regression discontinuity design (RDD). Itoriginates in a reform that was agreed upon in 2009. Since the end of 2010, a household iseligible for free public health insurance if a welfare index called Household Targeting Index(Índice de Focalización de Hogares, IFH) that is calculated by Peruvian authorities from anumber of variables is below a specific threshold. We have access to this information and useit to re-calculate the composite index of economic welfare. Variation in this index around thethreshold provides the natural experiment that we exploit.

3The latter may, however, seek medical attention on a pay-as-you-go basis in the same national hospitals orhealthcare centers and buy medication without a prescription.

4National Institute of Statistics and Informatics (INEI)-National series, http://series.inei.gob.pe:8080/sirtod-series/. Dow and Schmeer (2003) perform an analysis of the effect of health insurance in Costa Rica on infant andchild mortality. They use aggregate level data at the county level and control for fixed effects. As in Peru, increasesin insurance coverage over time went along with decreases in infant and child mortality. Gruber et al. (2014) findevidence for similar effects in Thailand.

5There are other studies relating enrollment to health care utilization and outcomes. For instance, Parodi (2005)finds that SIS enrollment increases the probability that poor pregnant women give birth in a formal institution.However, he does not control for selection into insurance. Bitrán and Asociados (2009) find that SIS increasesutilization for both preventive and curative services (with biggest impacts on treatments for diarrhea and acute res-piratory infections for children) and that SIS reduces the likelihood that insured individuals incur in out-of-pockethealth expenditures. The authors control for selection into insurance but they do not use the mean test used by SISat the period of analysis. Instead, they use consumption per capita to evaluate eligibility. There are also studies thatare more policy-oriented. For example, Arróspide et al. (2009) discuss the design and effectiveness of the SIS’sinstitutional budget and provide policy recommendations. Francke (2013) analyzes whether the implementation ofthe SIS program has played a role in extending health coverage in Peru.

3

Two aspects of the institutional background in combination with our research design areparticularly appealing. First, all individuals who are eligible for the program can be consideredcovered by it. The reason for this is that enrollment is easy and quick, as it can take place atthe facility at which individuals seek treatment, does not involve any fees, and as individualscan usually receive free treatment within a few days, often on the next day. Second, for ourpopulation of interest, crossing the eligibility threshold implies coverage. That is, the regressiondiscontinuity design is sharp, which means that we will not estimate local average treatmenteffects, but average treatment effects for those individuals whose welfare index has a valueclose to the eligibility threshold. This estimated treatment effect is policy-relevant, as it answersthe question what would happen to individuals who are just not covered if eligibility would beexpanded by increasing the threshold.

Exploiting the rich data from the ENAHO of Peru on health care utilization, out-of-pocketexpenditures and self-reported health outcomes, as well as the discontinuity generated by the in-stitutional rules, we find that insurance coverage has positive effects on the utilization of healthservices. For instance, the probabilities of visiting a doctor increases by 8 percentage points,the probability of receiving medicines increases by 11 percentage points, and the probabilityof receiving medical analysis increases by 3 percentage points. More generally, we find strongpositive effects on forms of care that can be provided at relatively low cost. Besides, insurancecoverage has positive effects on individuals receiving surgery. SIS does not provide strong in-centives to individuals to invest into preventive care. Nevertheless, we find that insured womenof childbearing age are more likely to receive pregnancy care, in line with the stark decrease inmaternal and child mortality that was observed after the program was introduced. At the sametime, as could be expected, we find no effects of insurance coverage on other forms of preven-tive care. Taken together, these findings are very much in line with the view that the price ofhealth care utilization has decreased and that this has led to increased demand.

We interpret our findings in more detail by looking at them through the lens of a simpleconceptual framework, or model, that we present in Section 3 below. Guided by this model,we provide evidence in favor of two arguments that are less common in economics.6 First,access to health care centers leads to increased awareness about health problems, because theyare more likely to see a doctor; and second, this even generates a willingness to pay for servicesthat are not covered, which in the context of Peru is a potentially desirable form of supplier-induced demand. In line with this, we show that health insurance coverage goes along withincreases in the level of health expenditures. Another way to think of the latter would be arevealed preference for medical care that stems from individuals being better informed abouttheir health care needs. Using an estimator of quantile treatment effects, we find that the effectis particularly pronounced in the top end of the distribution.

6An exception is Wagstaff and Lindelow (2008) who focus on the effects of health insurance on financial riskin China and find that health insurance coverage increases the risk of incurring high and catastrophic spending,respectively. They argue that this is because insurance encourages individuals to seek care and this ultimately leadsto higher expenditures that they then cover themselves.

4

Finally, as is common in studies like ours, we do not find clear effects on health reports atthe micro level. Our interpretation of this finding is that on the one hand positive effects willonly materialize in the long run. On the other hand, health reports are subjective and may beinfluenced by the increased inclination of insured individuals to see a doctor who then makesthem aware of their health care needs, as described in our model.

The literature on the impact of SHI for informally employed individuals in low and middleincome countries is scarce, but growing.7

Thornton et al. (2010) find that initial take-up of subsidized, but for-pay insurance “SeguroFacultativo de Salud” among informally employed individuals in Nicaragua was as low as 20percent. Moreover, after the subsidy expired most who previously signed up cancelled theirinsurance. The reason they give for this is that convenience and quality of care were not ade-quately addressed, which means that at the margin, the price of insurance—the cost side fromthe perspective of individuals—plays a role, but that the bulk of individuals does not buy in-surance because the associated benefits are too low. This could also be because resources werewasted either in the administration or at the health care providers. Another reason why the pro-gram did not reach its goal was that over the course of the evaluation of the program, there wasa drastic change in government, and with it the design of the program. The results for the fewwho did sign up and kept their insurance suggest that insurance could have a positive effect inthe sense that average health care expenditures, which are generally seen as too low, increased.This could, however, also be the case because those who bought insurance and kept it constitutea negative selection of risks for whom the effect of insurance is particularly high.

Next to this, there are three studies on Mexico, Barros (2008), King et al. (2009), and Sosa-Rubi et al. (2009). All of them investigate the effects of the “Seguro Popular” program, whoseaim it is—as the SIS’s in Peru—to improve access to health insurance for the poor. Unlikein the Peruvian, but like in the Nicaraguan program, coverage in the Mexican program is notfor free. Turning to the effects of introducing the program in Mexico, it is remarkable that thefindings in all three papers consistently suggest that the demand for medical care has shifted toproviders that are part of the system, and in line with this, individuals health care expenditureshave been reduced, including catastrophic health expenditures. In that sense, the program wassuccessful in being a transfer program, but less so in encouraging individuals to seek care whenill. Interestingly, as it is the case in Peru with the SIS program, policy makers have also targeted

7The selection of papers we discuss here is necessarily incomplete, but we believe it is to some extent repre-sentative. Acharya et al. (2013) systematically examine 64 papers on the effects of health insurance and presenta review on the 19 papers that correct for selection into insurance. The review concludes that there is little evi-dence on the impact of insurance on health status, some evidence on utilization, weak evidence on out-of-pockethealth expenditures, and unclear effects for the poorest. However, arguably, given the large variation in incentivesprovided by the respective institutions, it is not surprising that there is heterogeneity in the effect across countries.Giedion et al. (2013) also provide a comprehensive review classifying papers according to findings and researchdesign. They also conclude that specific features of the design have a large impact on the likelihood that spe-cific goals, such as increasing access or improving health, are reached. See also Abel-Smith (1992), InternationalLabour Office et al. (2006), Pauly et al. (2006) and Dercon et al. (2008), and the references therein, for a review ofthe more policy-oriented literature.

5

pregnant mothers, and consequently, as in Peru, there is a positive effect on obstetric utilization.At the same time, the findings do not suggest that utilization has increased for other types ofcare.

Turning to China, Wagstaff et al. (2009) find that the launch of a heavily subsidized volun-tary health insurance program in the rural parts of the country led to increased outpatient andinpatient utilization, but has not reduced out-of-pocket expenses per outpatient visit or inpatientspell. Also overall out-of-pocket expenditures have not decreased.

In contrast, Wagstaff (2010) finds for Vietnam that insurance coverage led to a reduction ofout-of-pocket spending and no impact on utilization.

Turning to Georgia, the design of the program is very similar to the one in Peru. However,and in contrast to our findings, Bauhoff et al. (2011) find no effect of insurance coverage onutilization. They argue that this is due to the fact that individuals were not aware of the fact thatthey were covered or that there were administrative problems that caused them to indeed not becovered, that they did not make use of the services because the program did not cover drugs,and because the perceived quality of the services was low. Therefore, it is not surprising thattheir findings are different from ours for Peru.

Next, turning to Colombia, and comparing the results to the ones in this paper for Peru,it becomes clear that the effects of insurance coverage depend on the design of the system.In Colombia, private insurers mainly receive a capitation fee and therefore have incentives toincrease preventive services on the one hand and to limit total medical expenditures on the other.And indeed, Miller et al. (2013) mainly find effects on preventive care. In Peru, SIS coversboth preventive and curative services and doctors are reimbursed on the basis of the treatmentsthey provide. Hence, participating hospitals and health care facilities do not have an incentiveto discourage curative treatments or medical procedures in favor of preventive services. Thisexplains why in Peru most of the effects are on curative use.

Finally, two recent papers, Gruber et al. (2014) and Limwattananon et al. (2015), investigatethe effect of a large-scale increase in health insurance coverage for the poor in Thailand. Theyfind that the program had positive effects on health care utilization, negative effects on out-of-pocket expenditures, and negative effects on child mortality rates. These findings are similarto ours for Peru, except that we find positive effects on health expenditures at the top end ofthe distribution. Our explanation for this is that individuals, once covered, became aware ofadditional health care needs and payed for some of them out-of-pocket.

To summarize, arguably, as of now more is known about the potential pitfalls than aboutthe effects of a successful SHI program and in particular on how they depend on details of theimplementation. The results presented in this paper suggest that SIS in Peru is an exceptionand belongs to the latter category, as does the Colombian program. Interestingly, the supply-side incentives between those two countries differ in important ways. For that reason, it willbe particularly interesting to compare our findings to the ones for Colombia. Also the Thaiprogram seems to be a notable exception.

6

We proceed as follows. Section 2 discusses the institutional background and provides detailson the SIS program. We present our model of demand for health insurance and health careutilization in Section 3. In Section 4 we provide information on our data and in Section 5 weformally describe the econometric approach. Results are presented in Section 6. A numberof robustness checks are conducted in Section 7. Section 8 concludes. Additional results arepresented in an Online Appendix.

2 Institutional Background

2.1 Seguro Integral de Salud

The public health insurance program “Seguro Integral de Salud” (SIS), whose effect we evaluatein this paper, was introduced in 2001. Its overarching goal is to improve access to health careservices for individuals who lack health insurance, giving priority to vulnerable groups of thepopulation who live in extreme poverty (Arróspide et al., 2009).

The creation of SIS and subsequent reforms led to a substantial increase of health insurancecoverage over time. Bitrán and Asociados (2009) and Francke (2013) provide an interestingdescriptive analysis of this increase and its relevance within the Peruvian health system in gen-eral. Between 2006 and 2011 the fraction of the population making use of services providedby SIS increased from 20.0 to 44.7 percent, which means that by then SIS was the main healthinsurance provider in Peru.8

2.2 Eligibility and Benefit Package

The aim of the government was to target particular, poor groups in the population. For this,ideally, eligibility should be based on accurate information on income at the level of the indi-vidual or family. However, such information is typically not available in developing countriesbecause a large part of the population works outside the formal sector and therefore does notpay income taxes and social security contributions. Eligibility for SIS is therefore based on theso-called Household Targeting System (“Sistema de Focalización de Hogares”, SISFOH). Aunified household registry is maintained and is used to calculate targeting indicators at the levelof the family (see SISFOH, 2010). Data are collected by government officials on a continuousbasis and using a standardized form. It includes questions on, amongst other things, housingcharacteristics, asset possessions, human capital endowments and other factors.

The IFH index is the main eligibility criterion for the sample of individuals we considerin this paper. It is a linear combination of the variables in the household registry that takes

8This increase was higher than the one in other countries, such as Colombia. One of the reasons for this could bethat the price for insurance was truly zero. Although there is no economic reason why there should be a substantialdifference between a zero price and a small positive price, behavioral aspects that lead to individuals perceiving abig difference between the two may play an important role (Shampanier et al., 2007).

7

on lower values for households that are more poor. Eligibility for SIS is based on SISFOH inthe capital Lima from 2011 on, and in the rest of the country from 2012 on. Online AppendixC explains in detail how the IFH is constructed, including the complete list of variables andtheir weights. Individuals are eligible if it is below a regional-specific threshold.9 Importantly,whereas potential beneficiaries intuit the importance of their answers to the questions of thegovernment official, they do not know how exactly the IFH index is calculated and what theircutoff value for eligibility is. SISFOH does not inform households about the value of theirindex and only provides the result of the eligibility evaluation. If eligible, individuals havethe possibility to enroll into SIS at a number of places, including MINSA facilities. They arecovered as soon as eligibility is confirmed, which is usually a matter of days. Then, they receivethe health services that are offered at MINSA facilities and that are part of the benefit package.In this sense, eligibility also means coverage.10

SIS offers a comprehensive package of health care benefits. There are no co-payments,coinsurance, deductibles, or similar fees. It is estimated that SIS covers 65 percent of the totaldisease burden (Francke, 2013). Tables 7 and 8 in Online Appendix A provide a detailed list ofservices covered by SIS together with the maximum levels of coverage. Coverage includes ob-stetric and gynecology interventions, pediatric interventions, neoplasm or tumor interventions,transmittable and non-transmittable disease’s interventions, chronic and degenerative disease’sinterventions and emergency care. It also includes outpatient medical-surgical intervention andhospitalization, as well as coverage of high-cost diseases. There are no waiting times or la-tent periods. But there are maximum levels of coverage in terms of the number of times anindividual can receive medical attention. For instance, for preventive care, SIS covers up to 10treatments related to pregnancy, ultrasounds, lab tests and the receipt of supplements of ironand folic acid. Regarding curative use of outpatient services, doctor visits and minor surgeriesare covered without any limit (including its medications). In the case of inpatient services (withor without surgeries), extra diagnosis and maximum levels are applied.

2.3 Supply Side

The Ministry of Health (“Ministerio de Salud”, MINSA) runs a network of health care centersand hospitals that provides services to individuals covered by SIS. MINSA also serves individu-

9At the same time, it is required that water and electricity expenditures are both below respective regional-specific thresholds. We control for this in our analysis, as we explain in Section 5. For Lima, the thresholds are 55for the IFH, 20 Soles for water expenditures and 25 Soles for electricity expenditures. This corresponds to 7.3 and9.1 U.S. dollars, respectively. According to the Central Bank of Peru, the average informal exchange rate (NuevosSoles per U.S. Dollar) in 2011 was 2.755. As a reference, the interbanking rate and the banking rate was 2.754.Table 13 in Online Appendix C provides the complete set of IFH thresholds by geographic areas. Figure 5 showsthe relationship between water and electricity expenditures and the IFH index. The Online Appendix also containsfurther details on the eligibility rules.

10It could be that some individuals are not aware of being eligible and therefore think of themselves as not beingcovered by health insurance. If this is the case, then we will estimate lower bounds of the effects of insurancecoverage, as explain in Section 7.7 below.

8

als who are not covered by SIS, on a pay-as-you-go basis. MINSA is reimbursed out of the SISbudget. Rates are based on estimates of the costs plus a markup. This means that, importantlyfor our results and as opposed to, for instance, Colombia, the system offers no extra incentivesto health care providers to encourage individuals to invest in preventive care. At the same time,it does not provide incentives that limit curative use.

Also importantly for our findings, in our study period, some MINSA facilities suffered froma number of substantial supply limitations. First, there was a lack of equipment in MINSAhospitals. According to Defensoría del Pueblo (2013), who conducted a survey for a sample ofhospitals at a national level in 2012, 20 percent of them lack at least one piece of equipmentrequired for inpatient surgery and 15 percent report to have problems with at least one otherinput needed for performing surgery. Second, there has been a shortage of dentists and ophthal-mologists. The rate of odontologists per ten thousand inhabitants is one of the lowest among allmedical professionals (Giovanella et al., 2012) and it is even lower when they work as providersfor SIS (Defensoría del Pueblo, 2013). In addition to that, at that time, only a small numberof ophthalmologists provides services to SIS participants, which in turn limits the use of oph-thalmological care. Only recently, and after our study period, the National OphthalmologicalInstitute, the largest provider in Peru, joined the list of SIS providers. Third, even though drugsare officially covered by SIS, according to the information in the ENAHO 2011, 37 percent ofthe covered individuals report to have paid for drugs received at the hospital level and 9.7 per-cent report to have paid for it at the health care centers level (Defensoría del Pueblo, 2013). Thismay be related to a cut that SIS experienced in its budget, which resulted in a failure to trans-fer resources to MINSA, which in turn motivated some hospitals to charge for hospitalization,regardless of insurance status. Patients are referred from health care centers to hospitals whenthe formers do not have specific medical specialties to perform proper diagnosis or treatments.Once at the hospitals, patients are less aware about the services they are freely entitled to asparticipants of SIS or they are not able to find all the medications they need. Taken together,the supply limitations imply that some patients were not able to receive some treatments andmay have been asked to pay for other treatments that were actually formally included in the SISpackage, especially when they received treatment in a hospital. Moreover, they may have hadto pay themselves for medicines that are formally covered.

3 Conceptual Framework

It is instructive to interpret our empirical results through the lens of a simple model of healthcare demand. The model we present below is based on the one in Einav et al. (2013) and adaptedfor our purposes. We keep it as simple as possible to focus on what we believe are the maindriving forces behind our empirical findings.11 The primary purpose of this model is to study

11We will consider a model with risk neutral individuals. Risk neutrality can be justified by thinking of themodel as being one within a given decision period, say a year. Risk aversion may still be present ex ante, before

9

the implications the institutional setup has on the demand side, with a focus on individuals notbeing aware of some of their health care needs when they are not covered by health insurance.

The point of departure is that an individual has health care needs λa and λu. One can think ofthem as measured in amounts of health care spending, as will become clear below. Without himseeing a doctor the individual is aware of the former and unaware of the latter, so “a” standsfor “aware” and “u” stands for “unaware”. Sen (2002) distinguishes in this context between“internal” and “external” views of health and stresses that “the patient’s internal assessmentmay be seriously limited by his or her social experience”, such as seeing a doctor or not.12

Importantly, and in contrast to what is common in developed countries, the individual canbuy all drugs at the pharmacy. That is, there are no prescription drugs. Therefore, the baselinecase is that he buys drugs at the pharmacy to treat his health care needs λa and pays for thishimself. This has been common practice for a long time and may of course ultimately haveadverse effects on health. However, evidence on this is scarce (Laing et al., 2001).

We specify utility to be quadratic in the difference between health care consumption m andhealth care needs. It is quasi-linear in money, which is given by income y minus the fraction c ·mthat is paid out-of-pocket, minus the cost p of going to a health care center instead of simplybuying drugs at a pharmacy. In our model, c will be one if the individual has no insurancefor his health care needs, meaning that the individual pays himself, and equal to zero if theindividual has insurance and the health care need is covered. In practice, some of the needs ofthe individual will be covered and other will not, but we keep this implicit here, as we want tofocus on our main argument. Formally, utility is of the form

u(m) = (m−λa−λu)−1

2ω(m−λa−λu)

2 + y− c ·m− p

and the individual will perceive λu to be zero before seeking medical attention at a health carecenter.13 ω is a parameter capturing moral hazard, as we show below.

In the baseline case the individual is not covered by insurance and utility is given by

u0(m) = (m−λa)−1

2ω(m−λa)

2 + y−m

the individual enters that decision period, and is over the out-of-pocket spending in the period—as in Einav et al.(2013). Here, we are primarily interested in choices that take place within the year.

12See also the discussion of various biases in self-assessed health measures that are discussed in Murray andChen (1992).

13It would be possible to extend the model such that individuals know the distribution of λu before going to thedoctor, and choose whether or not to go based on their beliefs. Moreover, in such a model, they would chooseoptimal consumption as to maximize expected utility, where the expectation is over realizations of λu. Here, wechose to use a simpler model, because we believe that it already captures the main mechanism that is at play in thecontext of health care demand in developing countries like Peru.

10

so that the optimal level of health care consumed follows from the first order condition and is14

m0 = λa.

Notice that this optimal consumption does not depend on the parameter ω capturing moralhazard. The associated indirect utility is

u∗0 = y−λa.

Suppose now that the individual considers seeking professional care at fixed cost p and withco-payment rate c. Before doing so he is only aware of his health care needs λa. Hence, heperceives his utility function to be

uunaware1 (m) = (m−λa)−

12ω

(m−λa)2 + y− c ·m− p

and he expects to consumemunaware

1 = λa +ω · (1− c).

Notice that now, health care expenditures depend on the parameter ω . Suppose the treatmentis covered by insurance such that c = 0. Then it follows that ω is the amount individualsconsume in addition if they are covered by insurance. Following Einav et al. (2013) we willthink of this as moral hazard.15

The associated indirect utility is

uunaware∗1 = ω · (1− c)− 1

2ω· (ω · (1− c))2 + y− c · (λa +ω · (1− c))− p.

Individuals seek professional care at a health care center if uunaware∗1 > u∗0. This is the case

ifω

2+ y− p > y−λa

orp <

ω

2+λa.

In words, they seek professional care if the opportunity cost of time that it takes to enroll and see

14Here and in the following it is easily verified that the second order conditions hold.15This increase in the demand for medical care once we control for the risk is commonly termed ex post moral

hazard. In contrast, a reduction of preventive effort or care that is due to them being covered by health insurance istermed ex ante moral hazard. If higher risk types, in the absence of moral hazard, buy insurance, then one speaksof adverse selection, following Akerlof (1970). See for instance Zweifel and Manning (2000). The empiricalliterature on moral hazard and adverse selection is still scarce, but growing. Chiappori (2000) provides a broadreview of the early literature. Chiappori and Salanié (2000), Abbring et al. (2003), and Abbring et al. (2003)investigate moral hazard in the market for car insurance. Finkelstein and Poterba (2004), Bajari et al. (2006),Fang et al. (2006), Aron-Dine et al. (2012) and Einav et al. (2011) study adverse selection and moral hazard inthe context of health insurance in developed countries. Einav et al. (2013) study the interrelation between adverseselection and moral hazard and term it “selection on moral hazard”.

11

a doctor is smaller than the health care expenditures they save, λa, plus the value they associatewith the additional free health care consumption, ω/2.

Conversely, one reason not to go a health care center is the perception that even thoughhealth insurance gives individuals access to doctors this is not valuable because advice obtainedfrom them is often of low quality, and therefore making use of insurance coverage is not worthits (opportunity) cost, including the time it takes to register.16 One way to think about this inour model is in terms of high values of p relative to λa.17

If individuals are heterogeneous in their λa and ω , then this will mean that there is selectionin the sense that those individuals who seek professional care are aware of higher health careneeds λa, are more responsive to the decrease in the price of care (have high ω), or face a lowp. It does not depend on λu.

Once individuals visit a doctor, he makes them aware of their health care needs λu. Then,health care expenditures are given by maware

1 = λa +λu +ω · (1− c) in the general case and forc = 0 we get

maware1 = λa +λu +ω.

This is a form of what has been termed supplier-induced demand (McGuire, 2000). Strauss andThomas (1998) argue that this is an important potential determinant of health care expendituresin developing countries. These additional expenditures may or may not be covered by theirhealth insurance, but in any case the increase in expenditures is given by λu. If they are notcovered, then going to the doctor may go along with an increase in out-of-pocket expendituresthat may or may not be beneficial to the individual. However, one can make the argument thatspending money reveals the preference of the individual for these increased expenditures andthat the supplier-induced demand is therefore beneficial to the individual.

Turning to the empirical predictions of the model, we expect utilization to increase onceindividuals are covered by health insurance, and out-of-pocket health care expenditures to eitherincrease (when individuals are made aware of many useful expenditures that are not covered)or decrease (when the majority of the treatments are covered and the overall out-of-pocketexpenditures decrease because less money is spent at the pharmacy and in health care centerstogether).

When asked about their health, individuals will base their answer on λai and munaware1i when

they are not covered by insurance and λai, λui and maware1i when they are. Thus, it is an empirical

16Das et al. (2008) provide evidence pointing towards such low quality advice, at least in other low-incomecountries. Non-enrollment into free (net of the opportunity cost of time) health insurance is a well-documentedphenomenon in the U.S. See, for instance, Blank and Card (1991), Blank and Ruggles (1996), and Currie andGruber (1996). Also in other contexts, it is argued that individuals make dominated choices (see for example Choiet al., 2011, and the references therein). In our empirical analysis, we allow individuals not to sign up for insurancewhen becoming eligible.

17Individuals can always buy private insurance that may or may not be more generous. Regular dependentemployees are covered by another social insurance scheme, independent of whether or not they are poor. However,in our empirical analysis we focus on individuals who become covered by public insurance. Therefore, we abstractfrom this in our model.

12

question whether the effect of insurance coverage on their subjective health report is positiveor negative. It may be positive because health care needs are satisfied and the individual evenreceives more treatment than he would buy himself. But it may just as well be negative becauseexposure to a health care professional has made the individual aware of his health care needs,even if they have been treated.

4 Data

This paper uses cross-sectional data from the ENAHO survey for the year 2011, which is repre-sentative at the level of each of the 24 departments in Peru. It is the only data set that providesinformation on health care utilization, out-of-pocket expenditures, self-reported indicators, in-surance status, and the information needed to re-compute the IFH index.18 Data are collectedusing face-to-face interviews with one or more respondents per household, who are also askedto provide information on the other household members.

SIS is targeted to individuals who work in the informal sector. For those individuals, theIFH index is the most important criterion to determine eligibility. Therefore, for our analysis,we select individuals that belong to a household in which no member is formally employed.19

This group comprises approximately 60 percent of the entire sample.In 2011, almost one third of the population lived in the Lima Province and half of Peru’s

Gross Domestic Product (GDP) was generated there. For two reasons, we focus on individualsfrom that province. First, the regulatory framework mandates that the IFH targeting rule shouldalready be applied in this area in 2011 and only afterwards to the rest of the country. Second,the Lima Province is very densely populated and therefore there are enough health care centersso that we can exclude that either a large distance or absence of the staff explain that individualsdo not demand health care.20

18The ENAHO is not a panel but a yearly survey in which the respondents change every year (repeated crosssection). There are also data for the years 2012, 2013 and 2014. For political reasons, however, the eligibilitycriteria were applied less strictly in these later years. Therefore, unlike in 2011, we cannot explot a regressiondiscontinuity design in those later years.

19We define formality as having monetary income from any wage activity. This does not include any other mon-etary income or income from self-employment. This definition is closest to the one used by the authorities. Theydistinguish between those individuals whose wage is observed, who are mainly employees with a formal contract,and others. We have also explored other definitions, including being a wage worker in the main occupation, anyindication of having a formal contract in the main occupation, and working in an enterprise that keeps accountingbooks and is affiliated to a pension system. Results remain qualitatively the same.

20According to Banerjee et al. (2004) these are two prime reasons why households in Rajasthan in India spenda considerable fraction of their budget on health care, essentially buying drugs. In other parts of Peru, utilizationof health services has been limited by supply constraints. The Office of the Ombudsman reports that most of the4,500 health care centers around the country are not sufficiently equipped to provide inpatient care (Defensoría delPueblo, 2013). An official technical committee concludes that the biggest challenge faced by the Peruvian healthsystem between 2009 and 2011 is the shortage of supply of health services in many parts of the country, because itlacks adequate capacity infrastructure, equipment and human resources (Comité Técnico Implementador del AUS,2010). Finally, also statistics from the World Bank shows that, while the average of hospital beds per 1,000 peopleis 1.83 for Latin America, it is only 1.55 for Peru. This also occurs with other measures of supply health services,including the number of health workers such as physicians, nurses and midwives (World Bank, 2013).

13

Table 1: Descriptive Statistics 1/2

(1) (2) (3)Variable Dummy Total Covered 7/. Not covered 7/.

N = 4,161 N = 1,397 N = 2764Mean Std. Mean Std. Mean Std.

DemographicsWoman D 0.511 - 0.502 - 0.516 -Age 33.014 22.250 28.498 21.046 35.296 22.496Years of education 8.126 4.854 6.542 4.481 8.926 4.839Number household members 4.607 2.101 4.490 1.882 4.665 2.202Woman head of household D 0.251 - 0.260 - 0.247 -Annual household income (thousand Soles) 1/. 30.62 27.07 18.80 13.360 36.79 30.24

More general forms of care usually provided by health care centerDoctor visits D 0.319 - 0.341 - 0.309 -Medicines D 0.456 - 0.466 - 0.452 -Analysis D 0.063 - 0.059 - 0.065 -X-rays D 0.037 - 0.034 - 0.039 -Other tests D 0.013 - 0.011 - 0.014 -Glasses D 0.041 - 0.021 - 0.051 -Other treatments D 0.234 - 0.201 - 0.251 -

Care provided by hospitalHospital D 0.060 - 0.057 - 0.061 -Surgery D 0.041 - 0.034 - 0.044 -Child birth 2/. D 0.033 - 0.041 - 0.028 -Dental care D 0.118 - 0.091 - 0.131 -Ophthalmological care D 0.054 - 0.024 - 0.068 -

Likely preventive careVaccines D 0.109 - 0.137 - 0.094 -Birth control D 0.060 - 0.068 - 0.055 -Pregnancy care 2/. D 0.074 - 0.098 - 0.062 -Planning 3/. D 0.119 - 0.171 - 0.091 -Iron 4/. D 0.137 - 0.148 - 0.130 -Kids check 5/. D 0.263 - 0.233 - 0.287 -Preventive campaign 6/. D 0.037 - 0.037 - 0.037 -

Curative careExperience health problem and seek medical attention D 0.199 - 0.190 - 0.203 -Experience health problem and doctor visits D 0.196 - 0.186 - 0.201 -Experience health problem and medicines D 0.187 - 0.178 - 0.191 -Experience health problem and analysis D 0.056 - 0.054 - 0.058 -Experience health problem and X-rays D 0.033 - 0.030 - 0.035 -Experience health problem and other tests D 0.009 - 0.006 - 0.010 -

Notes: Data from the ENAHO 2011. See Table 9 in the Online Appendix for variable definitions. 1/. Question applied at household level:total N = 1,129, covered N = 393, non-covered N = 736. 2/. Question applied for women in fertile age: total N = 1,182, covered N = 410,non-covered N = 772. 3/. Family planning: total N = 1,181, covered N = 410, non-covered N = 771. 4/. Reception of iron supplements forpregnant women and children less than three years old: total N = 343, covered N = 135, non-covered N = 208. 5/. Question applied for kidsunder age of 10: total N = 649, covered N = 283, non-covered N = 366. 6/. Information on prevention of sickness . 7/. Eligible using IFH,water and electricity criteria.

.

.

14

Table 2: Descriptive Statistics 2/2

(1) (2) (3)Variable Dummy Total Covered Not covered

Mean Std. Mean Std. Mean Std.

Health reportSymptom D 0.396 - 0.454 - 0.367 -

Illness D 0.144 - 0.147 - 0.142 -Chronic illness D 0.415 - 0.355 - 0.446 -Relapse D 0.097 - 0.085 - 0.102 -Accident D 0.023 - 0.026 - 0.021 -Num. days with symptom 0.134 1.122 0.169 1.387 0.116 1.003Num. days with illness 0.144 1.099 0.165 1.173 0.133 1.122Num. days with relapse 0.250 2.295 0.189 1.957 0.280 2.501Num. days with accident 0.069 1.123 0.071 0.946 0.068 1.133

Out-of-Pocket health care expendituresAny out-of-pocket health care expenditures D 0.571 - 0.536 - 0.589 -Health expenditures 401.133 1154.340 218.258 640.813 493.563 1331.651Absolute deviation health expenditures 532.713 1015.747 292.912 569.897 653.914 1159.973Absolute value residual expenditures 525.749 1008.115 292.415 566.358 643.683 1151.798Square residual expenditures 1292463 1.16e+07 406038 4792799 1740487 1.38e+07Expenditures exceed median D 0.487 - 0.497 - 0.482 -Expenditures exceed 75th percentile D 0.249 - 0.249 - 0.249 -

Health expenditures as a share of incomeShare expenditures 0.057 0.187 0.051 0.178 0.060 0.191Absolute deviation share 0.076 0.170 0.069 0.164 0.079 0.173Absolute value residual share 0.076 0.170 0.069 0.164 0.079 0.173Square residual expenditures 0.035 0.478 0.032 0.508 0.036 0.462Share exceeds median50 D 0.500 - 0.500 - 0.500 -Share exceeds 75th percentile D 0.250 - 0.250 - 0.250 -

Catastrophic health expendituresExceeds 5% of per capita household income D 0.231 - 0.218 - 0.237 -Exceeds 10% of per capita household income D 0.136 - 0.127 - 0.141 -Exceeds 15% of per capita household income D 0.096 - 0.086 - 0.102 -Exceeds 20% of per capita household income D 0.069 - 0.059 - 0.075 -Exceeds 25% of per capita household income D 0.051 - 0.043 - 0.055 -

Notes: Data from the ENAHO 2011. See Table 10 in the Online Appendix and Section 6.3 for variable definitions..

15

Our sample contains information on 4,161 individuals after the two exclusions criteria areapplied. Table 1 and 2 provide summary statistics for the main variables that we use in theanalysis. We distinguish between four sets of variables: demographics, health care utilization,self-reported health and out-of-pocket expenditures. The columns in the two tables contain thesummary statistics for the whole sample and for the sample broken down by insurance coverage.

On average, individuals in the sample are 33.0 years old, half of them are woman, individ-uals have around 8 years of education, and average annual household income is 30,620 Soles,or 11,114.3 U.S. dollars. Covered individuals are younger, less educated, and earn less. This isnot surprising since the SIS program is targeted to the poor.

Turning to utilization of health services, we find that, on average, 31.9 percent of the indi-viduals have visited a doctor in the last month, 45.6 percent have received medicines and 6.3percent have had medical analysis in the same period. 4.1 percent of the individuals have re-ceived an intervention or have undergone surgery in the last 12 months. Focusing on women,we observe that those who received pregnancy care in the last 12 months represent 7.4 percentof the sample of the women who are in fertile age.

Shifting attention to self-reported health, 39.6 percent of the individuals say that they haveexperienced a symptom in the last month. At the same time, only 14.4 percent report that theysuffered from illnesses. Regarding out-of-pocket expenditures, Table 2 shows that 57.1 percentof the individuals had some health expenditures in the last 12 months. The average annualexpenditures are around 401.1 Soles, or 145.6 U.S. dollars.

5 Econometric Approach

In this paper, we estimate the impact of SIS coverage on a host of variables characterizing healthcare utilization, out-of-pocket expenditures and self-reported health. Based on the institutionalsetup described in Section 2.2 we do this by means of a RDD using the IFH index as thecontinuous forcing variable.21

An individual is eligible for public insurance if she lives under poor conditions, which ismeasured at the household level. In the Lima Province, the condition for this is that the IFHindex is below or equal to a value of 55, provided that both, water and electricity expendituresdo not exceed 20 and 25 Soles, respectively. Hence, provided that the condition on water andelectricity expenditures holds, we have a sharp RDD.

We will estimate the effects using the standard ordinary least squares estimator with estima-tion equations of the form

yi = β 0+β 1zi+β 2elig_i f hi+β 3zielig_i f hi+β4not_elig_wei+β5not_elig_wei ·elig_i f hi+ε i

21This approach goes back to at least Thistlethwaite and Campbell (1960). See Hahn et al. (2001) for a moremodern exposition and Imbens and Lemieux (2008) for a discussion of practical issues.

16

and also controlling for demographics. Here, zi is the IFH index centered at its threshold,elig_i f hi is an indicator for eligibility based on the IFH index (that is, an indicator for zi < 0),and not_elig_wei is an indicator for ineligibility based on water and electricity consumption.The parameter of interest is β2, which is the effect of health insurance coverage for individualswho become covered because their IFH index crosses from above to just below the eligibilitythreshold.22 This parameter is policy-relevant because it is directly related to the question whatthe effects of expanding insurance coverage through increasing the threshold value would befor the individuals who would then receive coverage.23

In the model in Section 3, the effects of insurance coverage are a combination of ex post

moral hazard (ω) and increased awareness about medical needs (λu). Formally, we will not beable to separately quantify the two. However, estimating them for a host of different outcomevariables and interpreting the results in light of the institutional rules will allow us to concludethat most of the effects are likely related to the latter.

The first assumption we need to make for our analysis is that if no insurance would beassigned to everybody around the threshold, then the distribution of the outcome conditionalon the index would be smooth in the index zi around zero.24 Then, β2 is indeed the effect ofcoverage. This assumption cannot be tested directly and is therefore the main assumption wewill make. As we have argued before, the institutional rules suggest that it holds, as no otherprograms or rules are based on this eligibility threshold. Moreover, this assumption is supportedby further evidence that we present in Section 7 below.25

The second assumption is that insurance status is monotone in eligibility. This holds byconstruction, as changing from a value of the index slightly higher than the threshold valueto a value lower than the threshold value will make an individual eligible for insurance cover-age.26 The final, third assumption is an exclusion restriction. It is that in a small neighborhoodaround the eligibility threshold, the value of the index, zi, is independent of the outcomes, andin particular εi.27 It would be violated if households would manipulate their answers to thegovernment official in order to influence the value of the IFH index. As discussed in Section 2this is unlikely to be the case. We nevertheless test for manipulation in Section 7.1.

22A more general model would also include an interaction term between not_elig_wei and zi. We have alsoestimated this model and have found similar results. Moreover, the coefficient on the interaction term was alwaysinsignificant or very small.

23For individuals ineligible due to their high consumption of water and electricity, the effect of the IFH indexcrossing from above to just below the eligibility threshold is the sum of β2 and β5. Our results show that this is notsignificantly different from zero.

24This is slightly stronger than needed. Usually, it is enough to assume that the conditional expectation of yigiven zi is smooth around the threshold. We make this stronger assumption here in order to be able to estimatequantile treatment effects as well, as described below.

25For instance, we show in Figure 5 and Table 17 below that water and electricity consumption do not exhibit adiscontinuity at the eligibility threshold, which shows that there is independent variation in the IFH index.

26The assumption would be violated if an individual could become ineligible and lose coverage when crossingthe threshold from above. See Battistin et al. (2009) and Klein (2010) for related discussions.

27Again, for the same reason as above, mean independence suffices, but we make this stronger assumption inorder to also estimate quantile treatment effects.

17

Under the same assumptions, it is also possible to exploit the RDD and estimate quantiletreatment effects, as described in Frandsen et al. (2012). The underlying idea is straightforward.Instead of an average, the quantile treatment effect is the change in, say, the median of thedistribution of an outcome that results from being covered by public health insurance for thosewho select to enroll when becoming eligible. Results are presented in Section 6.3.

Before presenting the results, it is worth noting that our econometric approach does not in-volve a “first stage”, as it is usually the case in similar studies exploiting a regression disconti-nuity design. It is easiest to see this by inspecting the estimation equation above. If individualsare anyway not eligible and hence not covered by health insurance because of their water orelectricity consumption, then we will control for this.28 Consequently, β2 is the effect of be-coming eligible due to crossing the eligibility threshold for all other individuals. As we haveexplained above, given the insitutional rules eligibility essentially implies coverage. Hence, thisis not only the effect of elegibility, but also the effect of coverage. This is as if the first stage isone, which is always the case in a sharp regression discontinuity design (Hahn et al., 2001). Werevisit this point in Section 7.7 below.

6 Results

We start by showing the relationship between four measures of health care utilization and theIFH index.29 Higher values of the index indicate a higher level of welfare. Individuals are cov-ered by SIS when the index is below the eligibility threshold. In the figures, we plot estimatesof the expectations of the outcomes against the IFH index minus the eligibility threshold, whichis why we expect a downward jump of utilization at zero. We see that insurance coverage hasa positive effect on the probability to visit a doctor, to receive medicines, to receive medicalattention, and to undergo surgery. Moreover, we see that in general, utilization is positivelyrelated to welfare.

Next, as described in Section 5 above, we estimate the effect of health insurance coverageon a number of measures of health care utilization, out-of-the-pocket expenditures, and self-reported healthindicators. Throughout, we control for the value of the index (separately to theleft and to the right of the eligibility threshold), age, gender, whether the head of the householdis female, the number of household members, and years of education. Sensitivity analysis arecarried out in Section 7 below.

28We did not drop these observations in order to be able to estimate β1 and consequently the effects of interestmore precisely. Dropping them instead led to very similar point estimates. Moreover, the estimated effect ofcrossing the eligibility threshold should be zero for individuals who live in households with high enough water orelectricity consumption. This is testable and generally, we found that β2 +β5 was not significantly different fromzero.

29See Figure 7 in Online Appendix E for additional figures.

18

Figure 1: Health Care Utilization

(a) Doctor visits0.

000.

100.

200.

300.

400.

500.

60M

ean

of d

octo

r vi

sits

−60 −40 −20 0 20 40IFH index minus threshold

(b) Medicines

0.00

0.10

0.20

0.30

0.40

0.50

0.60

Mea

n of

med

icin

es

−60 −40 −20 0 20 40IFH index minus threshold

(c) Surgery

0.00

0.05

0.10

0.15

Mea

n of

sur

gery

−60 −40 −20 0 20 40IFH index minus threshold

(d) Medical attention

0.00

0.10

0.20

0.30

0.40

Mea

n of

med

ical

atte

ntio

n

−60 −40 −20 0 20 40IFH index minus threshold

Notes: Based on ENAHO data for the year 2011 for the Lima Province. See Section 4 and Online Appendix C for details on thedata and on how the IFH index is computed. The dots denote averages, with a bin width of 3. Their size represents the number ofobservations. The regression lines with corresponding 95 percent confidence intervals stem from separate linear regressions to theleft and to the right of the threshold using the individual-level data.

6.1 Health Care Utilization

Table 3 shows estimates of the effect of SIS on the utilization of 19 health services, includingthe first three in Figure 1 (the fourth is for the first outcome in Table 4). Section 2 explainswhy we should expect the strongest effects on simple forms of curative care and therefore, wemake a distinction between preventive and curative care and order the latter by the likely careprovider.

Looking first at more general forms of care typically provided by easily accessible healthcare centers, health insurance coverage increases the probability of visiting a doctor in the fourweeks prior to the interview by 8 percentage points, from a baseline of 31 percent (see lastcolumn of Table 1). The probability of receiving medicines in the same four weeks increasesby 11 percentage points, from a baseline of 45 percent. Moreover, health insurance coverageincreases the probability of performing at least some medical analysis by 3 percentage points,from a baseline of 7 percent.

At the same time, we do not find effects on the probability that X-rays are taken or othertests are performed. Eyeglasses are not part of the benefit package. We nevertheless estimated

19

the effects of insurance on the probability to obtain glasses, because being insured has an effecton seeing a doctor who could then advise the individual to get new glasses when performinganother treatment. However, we do not find support for this hypothesis.30

As for care provided by hospitals, Table 3 shows that health insurance coverage has bor-derline significant effects on the probability of visiting a hospital. In Section 2.3, we explainthat hospitals lacked equipment and were plagued by budget problems. This resulted in themcharging for services that are formally covered by public health insurance and could explainour finding. Nevertheless, we do find large effects of health insurance coverage on individualsreceiving surgery. Starting from a baseline of 4.4 percent in Table 1, we estimate the effect ofhealth insurance to be an increase of 3.7 percentage points. This can also explain the increasein out-of-pocket spending that we document in Section 6.3 below.

MINSA health care centers provide only basic services. Individuals are sent to hospitals inorder to visit a specialist, which also includes dentists and ophthalmologists. Table 3 shows thathealth insurance coverage has no significant effects on utilization of dental and ophthalmolog-ical care during the previous three months. However, this can be explained by another supplylimitation described in Section 2.3, namely the shortage of dentists and ophthalmologists .

Turning to care that is more likely of a preventive nature, we generally find smaller effects.One exception is our finding that health insurance coverage has significantly positive effects onthe probability that women at fertile age receive pregnancy care. It increases by 9 percentagepoints, from a baseline of 6 percent.31 This is in line with one of the motivations for introducingSIS in the very beginning, namely to reduce child and maternal mortality. At the same time, theprobability that a child is born in a hospital does not increase significantly.

The picture that emerges is that the effects of health insurance coverage are bigger for formsof care that are of a more general nature and can be provided by MINSA health care centersat relatively low cost. The effects on care provided by hospitals and on preventive care aremixed, which can be explained by supply limitations. As we have explained in Section 3, dueto these limitations the effect of health insurance on out-of-pocket expenditures could even bepositive. At the same time, insurance coverage does not seem to have an effect on more intenseprocedures or more elaborate forms of testing. Peru is a country in which poor individuals areaccustomed to not receiving any professional diagnosis and where drugs can also be boughtin a pharmacy without a prescription. Therefore, our findings nonetheless point towards theexpansion of the program being a success.

6.2 Curative versus Preventive Use

Our data include additional information that allow us to better tell apart curative from preventiveuse. We use this information to construct an indicator for experiencing a health problem, which

30Alternatively, one could see this as a placebo test in addition to the tests we perform in Section 7 below.31We do not observe whether a woman is pregnant. Therefore, in principle, this includes the effect of insurance

on becoming pregnant.

20

Table 3: Effect of Health Insurance on Health Care Utilization

Estimates Ste.More general forms of care usually provided by health care center

Doctor visits 0.0818*** (0.0305)Medicines 0.1111*** (0.0319)Analysis 0.0323** (0.0158)X-rays 0.0222* (0.0126)Other tests 0.0111 (0.0076)Glasses 1/. 0.0005 (0.0123)Other treatments 0.0482* (0.0267)

Care provided by hospitalHospital 0.0267* (0.0160)Surgery 0.0365*** (0.0130)Child birth 2/. 0.0265 (0.0264)Dental care 0.0058 (0.0205)Ophthalmological care -0.0004 (0.0146)

Likely preventive careVaccines 0.0398* (0.0209)Birth control -0.0242 (0.0156)Pregnancy care 2/. 0.0921** (0.0367)Planning 3/. 0.0391 (0.0388)Iron 4/. 0.0567 (0.0781)Kids check 5/. -0.0608 (0.0590)Preventive campaign 6/. 0.0024 (0.0121)

All providers and all forms of care togetherAny of the above 0.0753** (0.0295)

Notes: See Table 9 for variable definitions. N = 4,161, except for child birth, pregnancy care, familyplanning, iron and kids check. 1/. Not covered by SIS. 2/. Questions applied for women in fertileage, N = 1,182. 3/. Family panning, N = 1,181. 4/. Reception of iron supplements for pregnantwomen and children less than three years old, N = 343. 5/. Question applied for kids under age 10,N = 649. 6/. Information on prevention of sickness. Standard errors in parentheses. * p < 0.10 **p < 0.05 *** p < 0.01.

.

21

Table 4: Effect of Health Insurance on Curative Care Use

Estimates Ste.Experience health problem and seek medical attention 0.0875*** (0.0261)

Experience health problem and doctor visits 0.0845*** (0.0260)Experience health problem and medicines 0.0814*** (0.0256)Experience health problem and analysis 0.0310** (0.0151)Experience health problem and X-rays 0.0168 (0.0119)Experience health problem and other tests 0.0094 (0.0058)

Any of the above; or hospital, surgery, or child birth(last 3 reported in Table 3) 0.1153*** (0.0282)Notes: See main text for details on how we define curative care, . Standard errors in parentheses. * p<0.10 ** p<0.05*** p<0.01.

.

we then interact with the first five variables in Table 3.32 The measured effect is thus on thejoint event of experiencing a health problem and, say, going to the doctor. In that sense it iscurative use.

Table 4 shows the results. The first outcome indicates whether individuals have a healthproblem and then seek any type of medical attention. The effect of health insurance on thisoutcome is positive and is of a similar magnitude as the last outcome reported in Table 3. Theestimated effects of health insurance coverage on the other constructed joint outcomes, such asexperiencing a health problem and seeing a doctor, are almost as strong as the ones in Table3. When we think of the outcomes in Table 4 as being less prone to moral hazard, becauseinsurance coverage and a real medical need motivates them to see a doctor, for instance, thenthis suggests that the main effect of insurance coverage is to indeed allow individuals who havea health problem and need medical attention by a health care professional actually get it.

To characterize the effect of insurance coverage on a very comprehensive measure of cura-tive care use, we group the joint outcomes and three important additional outcomes from Table3 that likely have a curative nature together. We find that insurance increases the probability ofusing at least one curative service by 11.5 percentage points. This effect is significant at the 1percent level.

Overall, Table 3 and 4 show strong effects on curative use and only weak effects on usageof preventive care. It is interesting to compare these results to the ones by Miller et al. (2013)for Colombia, where the system provides larger incentives to invest in preventive care, as al-ready discussed in Section 1. And indeed, Miller et al. (2013) find stronger positive effects onpreventive care.

32We say that individuals experienced a health problem if at least one of the first 5 questions in Table 10 wasanswered with “yes”. See Table 2 for summary statistics. To be precise, in the questionnaire, individuals are askedwhether they saw a doctor, for instance. On top of that, they are asked whether they experienced health problemsand then, if they answer with "yes", again whether they saw a doctor. In our data, the answer to the first, generalquestion is always equal to the one to the third question.

22

6.3 Expenditures

Next, we estimate the effect of health insurance on the distribution of out-of-pocket spending.33

Obviously, health insurance coverage could reduce out-of-pocket spending because individualsdo not have to pay for certain treatments anymore. At the same time, as we argue in Section 3,it could also increase spending if receiving medical attention motivates individuals to actuallyspend more on their health themselves, because they become aware of additional health careneeds. This positive effect could originate in, or be reinforced by the supply side limitationsdescribed in Section 2.3 and 3.

We use a number of different outcome measures related to moments of the distribution ofhealth expenditures that have also been used in other studies. They either attempt to measureexpected health expenditures, their cross-sectional variability—interpreted as health risk—, orthe likelihood to incur catastrophic health expenditures.34

Table 5 presents the results.35 The first dependent variable is an indicator for incurring atleast some health expenditures. We find no significant effect on this outcome, suggesting thathealth expenditures are, if at all, mainly affected at the intensive margin.

Turning to the intensive margin, the second outcome is the level of annual health care expen-ditures. We find that insurance coverage leads to an increase of annual spending by about 168Soles on average, which corresponds to 61 U.S. dollars,—in line with the idea that individualsare motivated to spend more on their health when using medical services more often (whichthey do according to Table 3). This is about 1 percent of the average household income amongthe insured (18,800 Soles according to Table 1).

With our next outcome measure, we examine a possible effect on the variability of medicalspending in the cross section. It is the mean absolute deviation of health expenditures, calculatedseparately by insurance status, and similar to the one used by Miller et al. (2013). Healthinsurance has no significant effect on it.

The fourth and fifth measure are constructed from residuals obtained from a regression ofhealth expenditures on the value of the index, insurance status and the interaction of these twovariables. Our aim is to measure the variation of health care expenditures, and therefore we usethe absolute value and square of the residual instead of the more commonly used absolute valueof the expenditures and their square, respectively. Effects are only significant at the 10 percentlevel.

Looking at the results for the next two outcome measures, we find effects on the probabilitythat health expenditures exceed the median or the 75th percentile of the distribution of health

33We interchangeably use the terms spending and expenditures in combination with out-of-pocket, health care,or health.

34Most studies measure the cross-sectional variability of health expenditures and then interpret it as health riskat the individual level. This is only valid if there is no persistence in health expenditures. Otherwise, health risk isoverestimated. See for instance French and Jones (2004) for evidence in favor of such persistence in the U.S.

35See also the variable definitions in Table 10. We also experimented with more sophisticated variability mea-sures and found similar results.

23

Table 5: Effect of Health Insurance on the Distribution of Out-of-Pocket Spending

Estimates Ste.

Any health expenditures 0.0482 (0.0320)

Health care expendituresHealth expenditures 168.0253** (72.0087)Absolute deviation health expenditures -40.4636 (64.8124)Absolute value residual expenditures 110.8115* (63.7510)Squared residual expenditures 1.411e+06* (8.041e+05)Expenditures exceed median 0.1731*** (0.0320)Expenditures exceed 75th percentile 0.1421*** (0.0272)

Health expenditures as a share of incomeShare expenditures 0.0148 (0.0127)Absolute deviation share 0.0034 (0.0117)Absolute value residual share 0.0083 (0.0117)Squared residual expenditures 0.0320 (0.0378)Share exceeds median 0.1209*** (0.0322)Share exceeds 75th percentile 0.0836*** (0.0276)

Catastrophic health expendituresexceeds 5% of per capita household income 0.0699*** (0.0268)exceeds 10% of per capita household income 0.0247 (0.0219)exceeds 15% of per capita household income 0.0216 (0.0186)exceeds 20% of per capita household income 0.0226 (0.0162)exceeds 25% of per capita household income 0.0094 (0.0141)

Notes: See Table 10 for variable definitions. N=4,161. Standard errors in parentheses. * p<0.10 ** p<0.05 *** p<0.01.

24

Figure 2: Effect of health insurance on the distribution of out-of-pocket spending

01,

000

2,00

03,

000

4,00

0E

xpen

ditu

res

0.0 0.2 0.4 0.6 0.8 1.0Quantile

Not coveredCovered

Notes: The figure shows the percentiles of the distribution of expenditures with and without health insurance,along with 95 percent confidence intervals. See Frandsen et al. (2012) for details on the implementation.

expenditures in the entire population.In order to control for possible income differences we also analyze the effect of insurance

on the share of annual health expenditures spent out-of-pocket at the individual level, relativeto the annual per capita household income. For the last two outcomes in the second panel,we calculate the 50th and 75th percentile of the distribution of the share and find that healthinsurance increases the probability that this share exceeds the 50th and 75th percentile by 10 and6 percentage points respectively. Now we only find a positive effect on the last two outcomes.

Finally, we look into whether SIS changes the probability of an individual incurring catas-trophic health expenditures. Health expenditures are defined to be catastrophic if the share ofthe expenditures relative to per capita household income exceeds pre-defined threshold values.We follow Wagstaff and Lindelow (2008) and use the thresholds 5, 10, 15, 20 and 25 percentfor this. We find that the probability that individual health expenditures exceed 5 percent of theper capita household income increases by 7 percentage points, from a baseline of 24 percent inTable 2. We do not find significant effects for higher cutoffs.

In addition, Figure 2 shows the estimates of the quantiles of the distribution of health ex-penditures with and without health insurance coverage. As explained in Section 5, also theseestimates are obtained exploiting the discontinuity at the eligibility threshold.36 Interestingly,and consistent with the results we have presented above, we find that insurance has only a pos-

36Recall that we have a sharp regression discontinuity design, which means that coverage changes from 0 to 100percent at the eligibility threshold.

25

itive effect on the higher end of the distribution. To explore this further, we have conducted theanalysis underlying Figure 2 separately for each type of health expenditures. Figure 6 in theOnline Appendix shows the results for the two types of health expenditures for which we havefound an effect, spending on doctor visits and on medicines. The former is borderline signifi-cant and plausibly related to some treatments not being covered. The latter is likely related tothe supply limitations we have described in Section 2.3, namely that health care centers chargedfor medicines that were formally covered by public insurance or that these medicines were notavailable at the health care centers.37

To complement this systematic evidence with more direct one, we have conducted a smallsurvey in two health care facilities, in October and December 2014. We asked 19 individualsabout their experiences. Their answers revealed that the facilities did not have enough drugsin 2011, so that they bought them themselves at private pharmacies, in line with the effect onexpenditures for drugs we estimate. They also confirmed that the lack of equipment meantthat certain treatments could not be undertaken, which resulted in patients paying themselvesfor treatment that was received elsewhere. For instance, one person needed to receive prostatesurgery and said that he will probably get it elsewhere instead of waiting until the next year.

Overall, it is remarkable that we never find a significant negative effect on either expectedhealth expenditures or measures of variability or risk of high expenditures. Miller et al. (2013),in contrast, find for Colombia that insurance lowers both mean inpatient medical spending andits variability. Likewise, Limwattananon et al. (2015) find that health insurance coverage leadsto a decrease in out-of-pocket spending in Thailand.

Rises in health care expenditures as we find them are usually seen in a critical way, es-pecially if some treatments are formally covered but individuals have to nevertheless pay forthem. However, one may question whether this is justified here. On the one hand, this increasein expenditures can be seen as an additional burden to the individuals, possibly also increasingthe variability of expenditures in the cross-section. On the other hand, the alternative couldalso be that individuals are not treated at all because they are not aware of their health careneeds. In that sense the increase in spending for medicines and other treatments could also beseen as a desirable consequence of insurance coverage, which leads to increased accessibility,and thereby gives individuals the idea of using medical services in response to being insured.Moreover, looking at it in yet another way, some treatments are at least partly covered by SISso that the overall price of a treatment is generally lower, which means that the law of demand(that lower prices mean more demand) would also predict an increase in usage. So also in thatsense our findings could be less worrying than they may first seem.

37The survey contains one question that is asked to individuals who report having a health incident and seekmedical attention at a health care center. They are asked whether they found the prescribed drugs at the healthfacility. 43.6 percent answer with no, suggesting that they may have bought them themselves in a pharmacy.

26

Table 6: Effect of health insurance on health reports

Estimates Ste.Symptom 0.0259 (0.0317)Illness 0.0623*** (0.0228)Chronic illness 0.0465 (0.0294)Relapse -0.0029 (0.0191)Accident 0.0140 (0.0104)Incident (any of the above) 0.0887*** (0.0311)

Num. days with symptom 0.0187 (0.0881)Num. days with illness 0.1383** (0.0685)Num. days with relapse 0.1639 (0.1446)Num. days with accident 0.0738 (0.0648)Num days with any incident 0.3946** (0.1937)Notes: N = 4,161. Standard errors in parentheses. * p<0.10 ** p<0.05 ***p<0.01.

6.4 Effects on Health

We have argued in Section 3 that it is unclear how the effect of health insurance on self-reportedhealth measures can be interpreted. The reason for this is that individuals who are covered andwho see a doctor more often are more aware of their health problems. Therefore, even if theeffect on objective health measures is positive, it could be that the effect on subjective healthmeasures is negative or not significantly different from zero.

Table 6 reports our estimates of the effects of insurance coverage on self-reported healthindicators. We find positive effects on reports of illnesses, in line with the literature (see forinstance Acharya et al., 2013). But we—unlike the literature, and for the reasons given above—feel inclined to interpret them as indicating increased awareness of health problems.

7 Sensitivity Analysis

In this section, after having presented the main results, we assess whether they are sensitiveto the particular specifications we have used, and whether the identifying assumptions we havemade can be supported by additional empirical evidence. We start by examining whether house-holds may have manipulated the IFH index in order to become eligible for public insurance.Thereafter, we assess whether there were discontinuities at other values of the welfare index.This would raise concerns, as our approach builds on the premise that there is only one discon-tinuity of the expected outcomes in the welfare index, at least locally. Third, we conduct morelocal analyses by selecting subsamples of observations that are closer to the threshold. Through-out the analysis, we have controlled for covariates. Therefore, fourth, we assess whether therewere jumps in the expectations of covariates at the eligibility threshold. In passing, we con-

27

firm that also the expectation of water and electricity consumption, respectively, do not exhibitsuch a jump. After that, we conduct a non-parametric analysis. Then, we assess whether theexistence of other programs could challenge the validity of our results. And finally, we discusswhen our results could be considered lower bounds of the effects of interest. All correspondingtables and figures can be found in Online Appendix D.

7.1 Manipulation of the Running Variable

A common threat to studies based on a RDD is the incentive to manipulate the running variable.For this, individuals need information on how the IFH is calculated. Then, they need to usethis knowledge to manipulate their answers to the questions posed by the government officialin order to qualify for SIS. This is unlikely to be the case for two reasons. First, even thoughthe information on how the index is computed is, technically speaking, public, it is not easy toobtain and process it. Second, the set of variables included in the IFH construction are verifiedby the government officials and therefore difficult to manipulate. Therefore, the manipulationof the running variable would at most be partial.

We nevertheless analyzes this potential thread using the McCrary (2008) test. The idea isthat if manipulation takes place, then the density of the running variable will be discontinuous atthe cutoff. In our context, the density function would show many households barely qualifyingfor SIS, that is, to the left of the cutoff, and fewer failing to qualify, that is, to the right of thecutoff. The formal procedure is twofold: firstly a finely gridded histogram is obtained and thenthis histogram is smoothed with a local linear regression for each side of the cutoff.38

Figure 4 presents the results. There is no evidence for a jump of the density at the eligibilitythreshold, supporting the assumptions made in the main analysis.

7.2 Jumps at Non-Discontinuity Points

Our analysis implicitly assumes that the only discontinuities occur at the eligibility threshold,55, as we have specified conditional expectations to be linear in the forcing variable, separatelyto the left and to the right of this threshold. A first way of assessing this is to conduct a graphicalinspection. Figure 7 shows that the discontinuities do indeed mainly arise at the eligibilitythreshold.

Besides, following Imbens and Lemieux (2008) we conduct separate additional RDD analy-sis for the subsamples of covered and non-covered individuals and use the respective medians ofthe index as the threshold values. For example, the subsample at the left of the threshold wouldbe comprised by those with zi(IFHindex) < 55 and we test for a discontinuity at the median,to maximize the power of the test. By only using data on the left of the official threshold, weavoid conducting the regressions at a point where it is known to have a discontinuity.

38The IFH index is measured at the household level. Therefore, we conduct the analysis here also at the house-hold level.

28

Results are presented in Tables 14. In general, we observe no significant effects on healthoutcome variables when we run the regressions at the medians of the subsamples, with theexception of vaccines and “other treatments”. For vaccines, there is a negative effect on thesubsample to the left and a positive one on the subsample at the right of the original threshold.In our main analysis, we have found no effect of SIS on receiving vaccines. Figure 7 in theOnline Appendix suggests that what is picked up by this robustness check could be that SIS hasindeed a positive effect on receiving vaccines, but that the threshold that has in practice beenapplied for those is higher than the official IFH threshold. This could also explain our findingfor “other treatments”.

7.3 More Local Analysis

One could argue that the linearity assumption is strong and that therefore, the analysis shouldbe conducted on the population with IFH index values closer to the threshold. To see whetherresults are sensitive to that, we reduce the sample to the 75 and 50 percent of the populationwith IFH index values closest to the threshold, separately for each side.

Tables 15 and 16 shows the results for these reduced samples. Some of the coefficient es-timates increase in magnitude while the precision of the estimates decreases. For instance, forthe 75 percent sample, we find that the effects on the likelihood to visit a doctor, receivingmedicines, conducting analysis and performing surgery increase from 8.2, 11.1, 3.2 and 3.7 re-ported in Table 3 to 8.7, 13.8, 4.3 and 4.5 percentage points, respectively. At the same time, dueto the decreased number of observations, the precision of the estimates decreases, as expected.The estimates of the effects of insurance on pregnancy care is therefore no longer significantlydifferent from zero.

Importantly, however, the picture remains qualitatively and also quantitatively the same: theeffects on utilization are strongest, in particular for curative use, health expenditures increase,and if anything, individuals report to be in worse health.

7.4 Testing for Discontinuities in Covariates

It is standard practice to test whether the expectation of covariates such as age or gender is acontinuous function in the welfare index around the eligibility threshold. When it it is foundnot to be, then one may be concerned that the assumptions underlying our analysis do not hold.

We first conduct both a graphical and a formal analysis in which we replace the dependenthealth variables by the observed covariates age, gender, whether the woman is the head of thehousehold, the number of household members, years of education, household income, house-hold expenditures, and also water and electricity consumption. In Section 6, we use the firstfive covariates as controls in order to be able to obtain more precise estimates. We would beconcerned if water and electricity consumption would exhibit a discontinuity because eligibilityis only based on the index if both of them are not big enough.

29

Figure 5 and Table 17 summarize the results. The latter reports estimates of the effectof insurance on these variables, conducted either at the household or at the individual level,depending on the variable. We do not find evidence for discontinuities.

We also conducted the main analysis without controlling for covariates. Table 15 shows thatthe main conclusions we have drawn remain the same.

This supports the empirical approach to control for covariates and to focus on discontinuitiesrelated to the IFH index that we have taken in the main analysis.

7.5 Non-Parametric Analysis

To address the concern that linearity is too strong of an assumption even in smaller subsampleswe conduct a non-parametric analysis. For this, we use local linear and local quadratic regres-sions to predict the expected outcome and the probability to be covered by insurance using onlydata to the left or to the right of the discontinuity, respectively. We then calculate the differencebetween the prediction for the outcome from the right and from the left. This leads an estimateof the jump at the eligibility threshold that can be compared to the ones reported above.39

Importantly, results are not directly comparable to the ones we have presented above, be-cause here we ignore that some individuals do actually not become eligible when the IFH indexcrosses the threshold, as their water and electricity consumption is too high. We do control forthis in our main analysis. As this applies to about 50 percent of the individuals (see last outcomein Table 19), we can compare the results presented here to the other results after multiplying itby 2, following the logic of Imbens and Angrist (1994).

Table 18 and 19 show the results for a selection of dependent variables and for differentbandwidths. We chose the dependent variables out of the ones in Table 3, Table 4, Table 5,and Table 6, for which we found significant effects at the 5 percent level. Generally speaking,estimates are less precise, but after multiplying them by 2 point estimates are similar to the onesreported before. This suggests that the point estimates presented in Section 6 are not driven bythe linearity assumption.

7.6 Juntos and Food Aid Program

Our identification strategy is based on the assumption that discontinuities at the eligibilitythreshold can be attributed to SIS. There are some programs whose presence could in principlechallenge this assumption.

39It is in principle possible to also control for covariates. This, however, would involve estimating partiallylinear models where we impose that they enter linearly. We do not do so here because this goes along with a largeincrease in the computational burden when we bootstrap the standard errors. Therefore, the results reported hereare closest to the ones in Table 15 in the Online Appendix. Arguably, since these have in turn be close to the mainresults where we control for covariates and since we have found that none of the covariates exhibits a jumps at thediscontinuity, this seems a reasonable way to proceed given that this is meant to be a robustness check.

30

One of them is Juntos, a conditional cash transfer program. It combines a geographic target-ing of the poorest districts with individual targeting, based on the IFH index and the presenceof children up to the age of 14. However, Juntos is a rural program and our study focuses on theLima Province, and our data confirm that no individual in the sample belongs to Juntos.

Besides, there is a number of food aid programs oriented to the poor. To be precise, they areoriented to different groups of the population, such as mothers, children and school students.Our data show that 29 percent of the individuals of our sample receive at least support fromone of them.40 Importantly, since these programs do not use SISFOH’s targeting rules and inparticular not the IFH index, it is unlikely that a discontinuity at the eligibility threshold canbe attributed to them. Our finding in Section 7.4 that household expenditures do not exhibit adiscontinuity at the insurance threshold provides additional support for this interpretation.

7.7 Re-Interpreting our Results as Lower Bounds

In our analysis, we estimate the effect of becoming eligible for SIS due to crossing the elibilitythreshold. As we describe in Section 5, we control for ineligibility that is due to other reasonsand explain why we therefore face a sharp regression discontinuity design and estimate theaverage effect of becoming eligible. We argue in the Introduction and in Section 2.2 that thiseffect is essentially the effect of insurance coverage, because enrolling involves filling in a formand if eligible, individuals can usually come back the next day to receive treatment.

It could nevertheless be that individuals do not know whether or not they are actually cov-ered by health insurance. This by itself would not be a problem if they would always try toenroll and then learn that they are actually not eligible. If, however, they wrongly believe thatthey are not eligible and therefore do not even try to enroll, then they may behave as if theyare not covered by health insurance. Our analysis, however, assumes that they are covered.Consequently, the effects we estimate can be re-interpreted as lower bounds of the effects.41

8 Conclusions

Until recently, large parts of the population in developing countries did not have access to publichealth insurance. While it is commonly believed that the effects of health insurance coverage arepositive, opportunities to control for selection by exploiting natural experiments or by conduct-ing field experiments are rare, and therefore we still lack empirical evidence on its impact onhealth care utilization, out-of-pocket expenditures, and self-reported health indicators. Besides,

40The percentage of individuals that receive food aid is 29 percent among those not covered by SIS and 51percent among those who are covered. There is no information on the reception of food aid for 874 individuals, or20 percent of our sample.

41It would be possible to remedy this if individuals were asked whether they believe that they are eligible forSIS coverage (which they were not). Then, one could simply condition on this in the same way as we condition onineligibility due to high water or electricity consumption, as we explain in Section 5.

31

it is not yet fully understood through which channels health insurance coverage ultimately leadsto better health outcomes and to what extent it is possible to encourage individuals to invest intopreventive care.

In this paper, we use rich survey data from Peru to study the effects of the large-scale socialhealth insurance program called “Seguro Integral de Salud” (SIS). The SIS program is targetedto poor individuals working in the informal labor market. We make use of the institutional de-tails that give rise to a sharp regression discontinuity design. We estimate the effect of insurancecoverage on a wealth of measures for health care utilization, health expenditures and health.

We find strong effects of insurance coverage on arguably desirable, from a social welfarepoint of view, treatments such as receiving surgery and on forms of care that can be provided atrelatively low cost, such as visiting a doctor in the first place, receiving medication and medicalanalysis. Effects on preventive care—with the exception of pregnancy care—are much lesspronounced. This is not surprising, as the system does not provide any incentives to actually doso. Furthermore, we find positive effects on health care expenditures, mostly at the high end ofthe distribution, and no clear effects on self-reported health measures.

Based on this evidence, we develop two arguments that are less common in economics.First, access to health care centers leads to increased awareness about health problems; andsecond, this even generates a willingness to pay for services that are not covered, which in thecontext of Peru is a potentially desirable form of supplier-induced demand.

We formalize this interpretation of our findings using a simple model in which individualsare initially not aware of some of their health care needs. Once covered, they see a doctor andlearn about the needs they were unaware of and pay themselves for some of the treatments thatare not covered by the health insurance or are in short supply.

Overall, the evidence suggests that when compared to health care systems in other devel-oping countries, the Peruvian one is a notable exception. It seems to reach its goal to provideaccess to medical care to a sizable fraction of the poor. A key determinant of this success seemsto be that the monetary cost of enrolling is zero, instead of being small but positive, which it iselsewhere. As of now, there is no evidence on the effects this will have on objectively measuredhealth, but it is imaginable that increased access will ultimately lead to better health outcomes.

ReferencesAbbring, J., P. Chiappori, and J. Pinquet (2003). Moral hazard and dynamic insurance data. Journal of the

European Economic Association 1(4), 767–820.Abbring, J., J. Heckman, P. Chiappori, and J. Pinquet (2003). Adverse selection and moral hazard in insurance:

Can dynamic data help to distinguish? Journal of the European Economic Association 1(2-3), 512–521.Abel-Smith, B. (1992). Health insurance in developing countries: Lessons from experience. Health Policy and

Planning 7(3), 215–226.Acharya, A., S. Vellakkal, F. Taylor, E. Masset, A. Satija, M. Burke, and S. Ebrahim (2013). The impact of health

insurance schemes for the informal sector in low and middle income countries. Policy Research Working Paper,The World Bank.

Akerlof, G. A. . (1970). The market for lemons: Quality uncertainty and the market mechanism. Quarterly Journalof Economics 84, 488–500.

32

Aron-Dine, A., L. Einav, A. Finkelstein, and M. Cullen (2012). Moral hazard in health insurance: How importantis forward looking behavior? NBER Working Paper No. 17802. National Bureau of Economic Research,Cambridge, MA, U.S.A.

Arróspide, M., K. Rozas, and J. Valderrama (2009). Evaluated budget: Integrated health insurance. Ministry ofEconomics and Finance, Peru.

Bajari, P., H. Hong, and A. Khwaja (2006). Moral hazard, adverse selection and health expenditures: A semi-parametric analysis. NBER Working Paper 12445. National Bureau of Economic Research, Cambridge, MA,U.S.A.

Banerjee, A., A. Deaton, and E. Duflo (2004). Wealth, health, and health services in rural rajasthan. The AmericanEconomic Review 94(2), 326–330.

Banerjee, A. V. and E. Duflo (2007). The economic lives of the poor. The Journal of Economic Perspectives: ajournal of the American Economic Association 21(1), 141.

Barros, R. (2008). Wealthier but not much healthier: Effects of a health insurance program for the poor in Mexico.Working Paper, Stanford University, Stanford, CA.

Battistin, E., A. Brugiavini, E. Rettore, and G. Weber (2009). The retirement consumption puzzle: Evidence froma regression discontinuity approach. The American Economic Review 99(5), 2209–2226.

Bauhoff, S., D. Hotchkiss, and S. O. (2011). The impact of medical insurance for the poor in georgia: A regressiondiscontinuity approach. Health Economics 20, 1362–1378.

Bitrán and Asociados (2009). Impact of health insurance on access to health services, health services use andhealth status in the developing world. Global Health Financing Initiative - The Brrokings Institution.

Blank, R. M. and D. E. Card (1991). Recent trends in insured and uninsured unemployment: is there an explana-tion? The Quarterly Journal of Economics 106(4), 1157–1189.

Blank, R. M. and P. Ruggles (1996). When do women use aid to families with dependent children and food stamps?the dynamics of eligibility versus participation. Journal of Human resources, 57–89.

Card, D. and L. Shore-Sheppard. (2004). Using discontinuous eligibility rules to identify the effects of the federalmedicaid expansions on low-income children. Review of Economics and Statistics 86(3): 752-66.

Chankova, S., S. Sulzbach, and F. Diop (2008). Impact of mutual health organizations: evidence from west africa.Health Policy and Planning 23(4), 264–276.

Chiappori, P. (2000). Econometric models of insurance under asymmetric information. In G. Dionne (Ed.),Handbook of Insurance, Huebner International Series on Risk, Insurance, and Economic Security, pp. 365–393.Boston, MA, U.S.A.: Kluver.

Chiappori, P. and B. Salanié (2000). Testing for asymmetric information in insurance markets. Journal of PoliticalEconomy 108(1), 56–78.

Choi, J. J., D. Laibson, and B. C. Madrian (2011). $100 bills on the sidewalk: Suboptimal investment in 401 (k)plans. Review of Economics and Statistics 93(3), 748–763.

Comité Técnico Implementador del AUS (2010). Avances del proceso de implementación del aseguramientouniversal en salud (2009-2011).

Commission on Macroeconomics and Health (2001). Macroeconomics and health: investing in health for economicdevelopment. World Health Organization, Geneva, CH.

Currie, J. and J. Gruber (1996). Health insurance eligibility, utilization of medical care, and child health. TheQuarterly Journal of Economics 111(2), 431–466.

Das, J., J. Hammer, and K. Leonard (2008). The quality of medical advice in low-income countries. The Journalof Economic Perspectives 22(2), 93–114.

Defensoría del Pueblo (2013). Serie Informes Defensoriales - Informe Defensorial No. 161.Dercon, S., M. Kirchberger, J. W. Gunning, and J. P. Platteau (2008). Literature review on microinsurance. Mi-

croinsurance Paper No. 1. International Labor Office, Geneva.Dow, W. H. and K. K. Schmeer (2003). Health insurance and child mortality in Costa Rica. Social Science &

Medicine 57(6), 975–986.Einav, L., A. Finkelstein, S. Ryan, P. Schrimpf, and M. Cullen (2011). Selection on moral hazard in health

insurance. NBER Working Paper No. 16969. National Bureau of Economic Research, Cambridge, MA, U.S.A.Einav, L., A. Finkelstein, S. P. Ryan, P. Schrimpf, and M. R. Cullen (2013). Selection on moral hazard in health

insurance. American Economic Review 103(1), 178–219.Fafchamps, M. (1999). Risk sharing and quasi-credit. Journal of International Trade & Economic Develop-

ment 8(3), 257–278.Fang, H., M. Keane, and D. Silverman (2006). Sources of advantageous selection: Evidence from the medigap

insurance market. NBER Working Paper No. 12289. National Bureau of Economic Research, Cambridge, MA,U.S.A.

Finkelstein, A. and J. Poterba (2004). Adverse selection in insurance markets: Policyholder evidence from the ukannuity market. Journal of Political Economy 112(1), 183–208.

33

Francke, P. (2013). Peru’s comprehensive health insurance and new challenges for universal coverage. UniversalHealth Coverage Studies, The World Bank.

Frandsen, B., M. Frolich, and B. Melly (2012). Quantile treatment effects in the regression discontinuity design.Journal of Econometrics (168), 382–395.

French, E. and J. B. Jones (2004). On the distribution and dynamics of health care costs. Journal of AppliedEconometrics 19(6), 705–721.

Giedion, U., E. Andrés Alfonso, and Y. Díaz (2013). The impact of universal coverage schemes in the developingworld: a review of the existing evidence. Universal Health Coverage Studies Series No. 25.

Giné, X., R. Townsend, and J. Vickery (2008). Patterns of rainfall insurance participation in rural india. The WorldBank Economic Review 22(3), 539–566.

Giovanella, L., O. Feo, M. Favia, and S. Tobar (2012). Instituto Suramericano de Gobierno en Salud ISAGS.Gruber, J., N. Hendren, and R. M. Townsend (2014). The great equalizer: Health care access and infant mortality

in thailand. American Economic Journal. Applied economics 6(1), 91.Hahn, J., P. Todd, and W. van der Klaauw (2001, January). Identification and estimation of treatment effects with

a regression-discontinuity design. Econometrica 69(1), 201–209.Hsiao, W. C. and P. Shaw (2007). Social health insurance for developing countries. The World Bank.Imbens, G. and T. Lemieux (2008). Regression discontinuity designs: A guide to pratice. Journal of Economet-

rics 142, 615 – 635.Imbens, G. W. and J. D. Angrist (1994, March). Identification and Estimation of Local Average Treatment Effects.

Econometrica 62(2), 467–475.International Labour Office, Deutsche Gesellschaft für technische Zusammenarbeit, and World Health Organi-

zation (2006). Berlin recommendations. International Conference on Social Health Insurance in DevelopingCountries Berlin, 05-07 December 2005. Final version 3 July 2006.

Jowett, M. (2003). Do informal risk sharing networks crowd out public voluntary health insurance? evidence fromvietnam. Applied Economics 35(10), 1153–1161.

King, G., E. Gakidou, K. Imai, K. Lakin, T. Moore, C. Nall, N. Ravishankar, M. Vargas, M. Tellez-Rojo, J. Her-nandez Avila, M. Hernandez Avila, and H. Hernandez Llamas (2009). Public policy for the poor? A randomizedassessment of the Mexican universal health insurance programme. The Lancet 373, 1447–1454.

Klein, T. J. (2010). Heterogeneous treatment effects: Instrumental variables without monotonicity? Journal ofEconometrics 155(2), 99–116.

Laing, R., H. Hogerzeil, and D. Ross-Degnan (2001). Ten recommendations to improve use of medicines indeveloping countries. Health Policy and Planning 16(1), 13–20.

Limwattananon, S., S. Neelsen, O. O’Donnell, P. Prakongsai, V. Tangcharoensathien, E. van Doorslaer, andV. Vongmongkol (2015). Universal coverage with supply-side reform: The impact on medical expenditurerisk and utilization in Thailand. Journal of Public Economics.

McCrary, J. (2008). Manipulation of the running variable in the regression discontinuity design: A density test.Journal of Econometrics 142(2), 698–714.

McGuire, T. G. (2000). Physician agency. Handbook of health economics 1, 461–536.Miller, G., D. Pinto, and M. Vera-Hernandez (2013, October). Risk protection, service use, and health outcomes

under Colombia’s health insurance program for the poor. American Economic Journal: Applied Economics 5(4),61–91.

Murray, C. J. and L. C. Chen (1992). Understanding morbidity change. The Population and Development Review,481–503.

Parodi, S. (2005). Evaluating the effects of the integrated health insurance over maternal health equity within thecontext of non-economics barriers on service access. CIES, GRADE.

Pauly, M. V., P. Zweifel, R. M. Scheffler, A. S. Preker, and M. Bassett (2006). Private health insurance in developingcountries. Health Affairs 25(2), 369–379.

Sen, A. (2002). Health: Perception versus observation. British Medical Journal 324(7342), 860–861.Shampanier, K., N. Mazar, and D. Ariely (2007). Zero as a special price: The true value of free products. Marketing

Science 26(6), 742–757.SISFOH (2010). Methodology to calculate the household targeting index, ifh. Ministry of Economics and Finance,

Peru.Sosa-Rubi, S. G., O. Galarraga, and J. E. Harris (2009). Heterogeneous impact of the “seguro popular” program

on the utilization of obstetrical services in mexico, 2001–2006: a multinomial probit model with a discreteendogenous variable. Journal of Health Economics 28(1), 20–34.

Strauss, J. and D. Thomas (1998). Health, nutrition and economic development. Journal of Economic Literature 36,766–817.

Thistlethwaite, D. and D. Campbell (1960). Regression-discontinuity analysis: An alternative to the ex post factoexperiment. Journal of Educational Psychology 51, 309–317.

34

Thornton, R., L. Hatt, E. Field, M. Islam, F. Solis Dias, and M. Gonzales (2010). Social security health insurancefor the informal sector in nicaragua: A randomized evaluation. Health Economics 19, 181–206.

Wagstaff, A. (2010). Estimating health insurance impacts under unobserved heterogeneity: the case of vietnam’shealth care fund for the poor. Health Economics 19(2), 189–208.

Wagstaff, A. and E. v. Doorslaer (2003). Catastrophe and impoverishment in paying for health care: with applica-tions to vietnam 1993–1998. Health Economics 12(11), 921–933.

Wagstaff, A. and M. Lindelow (2008). Can insurance increase financial risk?: The curious case of health insurancein china. Journal of Health Economics 27(4), 990–1005.

Wagstaff, A., M. Lindelow, G. Jun, X. Ling, and Q. Juncheng (2009). Extending health insurance to the rural pop-ulation: An impact evaluation of china’s new cooperative medical scheme. Journal of health economics 28(1),1–19.

WHO (2010). World health statistics 2010. World Health Organization.World Bank (2013). World development indicators.Zweifel, P. and W. G. Manning (2000). Moral hazard and consumer incentives in health care. Handbook of Health

Economics 1, 409–459.

35

Online Appendix

This Online Appendix contains additional information on the health insurance plan (AppendixA), detailed variable definitions (Appendix B), a detailed description of the way the welfare in-dex is calculated (Appendix C), results related to the sensitivity analysis in Section 7 (AppendixD), and finally a number of supplementary tables and figures (Appendix E).

APl

anC

hara

cter

istic

s

Tabl

e7:

Bas

icPl

anof

Hea

lthIn

sura

nce

-PE

AS

Var

iabl

es1/

.Ty

peTr

eatm

ent

Cov

erag

eM

axim

umC

over

age

PRE

VE

NT

IVE

SERV

ICE

S

MA

AM

KC

1C

RE

Dch

eck

onC

hild

ren

Mat

eria

ls,s

uppl

ies.

1m

onth

......

2ch

ecks

1ye

ar...

.....

11ch

ecks

2ye

ars

......

.6ch

ecks

4ye

ars

......

.8ch

ecks

.

MA

AI

KC

2C

heck

onne

wbo

rnw

eigh

ted

less

than

2,50

0gr

.C

onsu

ltatio

ns,d

osag

eof

hem

oglo

bin,

iron

supp

lem

enta

tion.

15da

ys...

.....5

chec

ks3

mon

ths.

......

4ch

ecks

1ye

ar...

......

..9ch

ecks

.

MI

3M

icro

nutr

ient

supp

lem

enta

tion

Prov

isio

nof

Vita

min

A,I

ron

supp

lem

ent.

Con

sulta

tion,

Prin

ted

educ

atio

nalm

ater

ial.

1ye

ar...

......

.3ch

ecks

2ye

ars

......

...4

chec

ks.

OT

APC

4Pr

egna

ncy

diag

nosi

sL

abor

ator

ypr

oced

ure.

No

limit

MA

MPC

5Pr

enat

alC

are

Mat

eria

ls,s

uppl

ies.

10ch

ecks

OT

APC

6C

ompl

ete

labo

rato

ryte

sts

of

preg

nant

wom

enFu

llse

tofl

abor

ator

yte

sts.

No

limit

OT

APC

7Tr

eatm

entf

orpr

egna

ntw

omen

w/H

IV

EL

ISA

test

forH

IVdu

ring

preg

nanc

y,ch

ildbi

rth

and

post

part

um.P

roph

ylac

tic

trea

tmen

tto

HIV

posi

tive

preg

nant

wom

en.

No

limit

XM

PC8

Obs

tetr

icul

tras

ound

exam

inat

ions

Mat

eria

ls,s

uppl

ies.

No

limit

MA

AM

D9

Ora

lhea

lthSu

pplie

s.2

chec

ks/y

ear

MA

MD

10Pr

even

tion

ofto

oth

deca

yM

ater

ials

,sup

plie

s.

Den

talp

roph

ylax

is...

...2

chec

ksTo

pica

lflou

rcoa

ting

...4

chec

ksD

esta

rtra

je...

......

......

....1

chec

kTo

pica

lflou

rgel

......

....2

chec

ksSe

alan

tsap

plic

atio

n..

...1

chec

ksR

esto

ratio

npr

actic

e..

...4

piec

esIo

nom

erin

activ

atio

n...

.4pi

eces

.

DV

BC

MP

11Fa

mily

plan

ning

Gui

danc

ean

dco

unse

ling

sess

ion.

Del

iver

yof

cont

race

ptiv

es.

Not

deci

ded

......

...4

chec

ksD

ecid

ed...

......

......

....c

heck

s.

DV

O12

Vis

uala

cuity

impa

irm

ent

dete

ctio

nFu

llbe

nefit

.Exc

lude

spr

ovis

ion

ofle

nses

.Chi

ldre

nan

dA

dole

scen

ts.

1ch

eck/

year

MA

M13

Nor

mal

post

part

umca

reC

heck

s.In

clud

essu

pplie

sus

edin

care

.2

chec

ks

MA

M14

Prop

hyla

ctic

trea

tmen

tkid

s

HIV

Prov

isio

nof

artifi

cial

milk

for6

mon

ths.

Mon

itori

ng.

No

limit

Not

es:

1/.

Var

iabl

esob

serv

edin

the

EN

AH

O.D

octo

rV

isits

(DV

);M

edic

alA

ttent

ion

(MA

);M

edic

ines

(M);

Ana

lysi

s(A

);K

ids

Che

ck(K

C);

Iron

(I);

Plan

ning

(P);

Oth

erte

sts

(OT

);Su

rger

y(S

);Pr

egna

ncy

Car

e(P

C);

Bir

thC

ontr

ol(B

C);

Chi

ldB

irth

(CB

);D

enta

l(D

);O

phth

alm

olog

ical

care

(O);

Hos

pita

l(H

).Se

eR

.M.N

°24

0-20

09/M

INSA

form

ore

deta

ils.

.

.

A1

Tabl

e8:

Bas

icPl

anof

Hea

lthIn

sura

nce

-PE

AS

(con

tinue

d)

Var

iabl

es1/

.Ty

peTr

eatm

ent

Cov

erag

eM

axim

umC

over

age

RE

CU

PER

AT

IVE

AN

DR

EH

AB

ILIT

AT

ION

SERV

ICE

S

MA

AM

15Im

med

iate

atte

ntio

nto

the

norm

alne

wbo

rnSu

pplie

s,dr

ugs,

lab

test

s.N

olim

it

MA

OT

MH

16N

ewbo

rn’s

inpa

tient

care

with

outs

urgi

calp

roce

dure

Dru

gs,a

uxili

ary

test

s,in

puts

duri

ngst

ayin

the

heal

thfa

cilit

y.N

olim

it

MA

OT

MH

17N

ewbo

rn’s

inpa

tient

care

with

surg

ical

proc

edur

eD

rugs

,aux

iliar

yte

sts,

inpu

tsdu

ring

stay

inth

ehe

alth

faci

lity.

No

limit

MA

MC

B18

Vag

inal

birt

hca

reD

rugs

,mat

eria

ls,s

uppl

ies.

No

limit

MA

MC

B19

Cae

sare

anse

ctio

nD

rugs

,aux

iliar

yte

sts,

supp

lies,

mat

eria

ls.

No

limit

DV

M20

Spec

ializ

edm

edic

al

cons

ulta

tion

Dru

gs,s

uppl

ies,

diag

nost

icsu

ppor

t.N

olim

it

DV

MD

21A

mbu

lato

ryca

re:d

octo

rD

rugs

,sup

plie

s,di

agno

stic

supp

orti

nhe

alth

faci

litie

s.N

olim

it

MA

MD

22D

enta

lres

tora

tion

Mat

eria

ls,p

ortio

nfo

rrep

lace

men

tequ

ipm

enta

ndin

stru

men

tatio

n.3

dent

alre

stor

atio

ns/y

ear

MA

MD

23D

enta

lext

ract

ion

Mat

eria

ls,p

ortio

nfo

rrep

lace

men

tequ

ipm

enta

ndin

stru

men

tatio

n.3

extr

actio

ns/y

ear

MA

M24

Am

bula

tory

care

Mat

eria

ls,s

uppl

ies.

Inhe

alth

faci

litie

s:L

evel

I&II

2/.

No

limit

MA

M25

Em

erge

ncy

care

Dru

gs,a

uxili

ary

test

s,m

ater

ials

,sup

plie

s.N

olim

it

AM

26D

iagn

ostic

supp

ort

Dia

gnos

ticsu

ppor

t.N

odi

agno

stic

capa

bilit

ies

incl

ude

supp

ortn

otta

riff

ed.

No

limit

MA

MS

27M

edic

alou

tpat

ient

surg

ery

Dru

gs,a

uxili

ary

test

s,m

ater

ials

,sup

plie

sdu

ring

surg

ery

and

patie

nt’s

stay

inth

e

faci

lity.

No

limit

MA

M28

Inpa

tient

inhe

alth

faci

lity

with

outs

urge

ryD

rugs

,mat

eria

ls,s

uppl

ies

duri

ngsu

rger

yan

dpa

tient

’sst

ayin

the

faci

lity.

No

limit

OT

MS

29In

patie

ntw

ithsu

rger

yA

uxili

ary

test

s,dr

ugs

duri

ngsu

rgic

alpr

oced

ure,

expe

nditu

res

incu

rred

duri

ngpa

tienc

e

stay

until

disc

harg

e.N

olim

it

MH

30A

dmis

sion

inIC

UM

ater

ials

,sup

plie

s,dr

ugs.

Onl

yin

faci

litie

sw

here

serv

ice

can

beve

rifie

d.N

olim

it

MA

D31

Spec

ializ

edde

ntal

care

Proc

edur

espu

lpec

tom

y,pu

lpot

omy

acce

ssop

enin

g,di

rect

and

indi

rect

pulp

capp

ing,

fixat

ion

orsp

lintin

gof

the

toot

hw

ithco

mpo

site

,Gin

give

ctom

ylo

caliz

edex

trac

tion

of

reta

ined

piec

e,en

ucle

atio

nor

mar

supi

aliz

atio

n.

3ch

ecks

/yea

r

32R

ehab

ilita

tion

care

Reh

abili

tatio

nof

frac

ture

orsp

rain

inL

evel

Icar

e.L

evel

IIca

reon

lyfo

renr

olle

d

popu

latio

n.N

olim

it

AD

MIN

IST

RA

TIV

ESE

RVIC

ES

33E

mer

genc

yTr

ansf

erD

ispl

acem

entr

oad

orai

racc

ordi

ngto

reco

gniz

edm

edic

alin

dica

tion.

Unr

ecog

nize

d

exam

fort

hedi

agno

sis

inou

tpat

ient

s.N

olim

it

34Fo

odA

lloca

tion

Rec

ogni

zed

until

ada

yaf

terd

isch

arge

.N

olim

it

35B

uria

lN

iche

serv

ices

,cas

ket,

shro

ud,f

uner

alch

apel

and

tran

sfer

ofth

ede

ceas

edto

the

cem

eter

y."D

irec

tiva

deSe

pelio

s"

Not

es:1

/.V

aria

bles

obse

rved

inth

eE

NA

HO

.2/.

Lev

elI:

Hea

lthfa

cilit

y&

heal

thce

nter

.Lev

elII

:Doc

torV

isits

(DV

);M

edic

alA

ttent

ion

(MA

);M

edic

ines

(M);

Ana

lysi

s(A

);K

ids

Che

ck(K

C);

Iron

(I);

Plan

ning

(P);

Oth

erte

sts

(OT

);Su

rger

y(S

);Pr

egna

ncy

Car

e(P

C);

Bir

thC

ontr

ol(B

C);

Chi

ldB

irth

(CB

);D

enta

l(D

);O

phth

alm

olog

ical

care

(O);

Hos

pita

l(H

).Se

eR

.M.N

°24

0-20

09/M

INSA

form

ore

deta

ils.

A2

BVa

riab

leD

efini

tions

Tabl

e9:

Var

iabl

eD

efini

tions

Var

iabl

eD

efini

tion

Part

icip

atio

nH

ealth

Insu

ranc

eA

reyo

uen

rolle

din

SIS?

Dem

ogra

phic

sW

oman

Age

Yea

rsof

educ

atio

nN

umbe

rhou

seho

ldm

embe

rsW

oman

head

ofho

useh

old

Ann

ualh

ouse

hold

inco

me

(tho

usan

dSo

les)

Mor

ege

nera

lfor

ms

ofca

reus

ually

prov

ided

byhe

alth

care

cent

erD

octo

rvis

itsH

ave

you

visi

ted

the

doct

orin

the

last

4w

eeks

?M

edic

ines

Hav

eyo

ure

ceiv

edm

edic

ines

inth

ela

st4

wee

ks?

Ana

lysi

sH

ave

you

had

anal

ysis

inth

ela

st4

wee

ks?

X-r

ays

Hav

eyo

uha

dX

-ray

sin

the

last

4w

eeks

?O

ther

test

sH

ave

you

had

othe

rtes

tsin

the

last

4w

eeks

?G

lass

esH

ave

you

boug

htgl

asse

sin

the

last

3m

onth

s?O

ther

trea

tmen

tsH

ave

you

rece

ived

othe

rcar

ein

the

last

3m

onth

s?(i

.e.o

rtho

pedi

cse

rvic

es)

Car

epr

ovid

edby

hosp

ital

Hos

pita

lH

ave

you

been

hosp

italiz

edin

the

last

12m

onth

s?Su

rger

yH

ave

you

had

asu

rger

yin

the

last

12m

onth

s?C

hild

birt

hH

ave

you

had

child

birt

hca

rein

the

last

12m

onth

s?D

enta

lcar

eH

ave

you

had

dent

alca

rein

the

last

3m

onth

s?O

phth

alm

olog

ical

care

Hav

eyo

uha

dop

htha

lmol

ogic

alca

rein

the

last

3m

onth

s?Li

kely

prev

entiv

eca

reV

acci

nes

Hav

eyo

ure

ceiv

edva

ccin

esin

the

last

3m

onth

s?B

irth

cont

rol

Hav

eyo

ure

ceiv

edbi

rth

cont

rolp

rodu

cts

inth

ela

st3

mon

ths?

Preg

nanc

yca

reH

ave

you

rece

ived

preg

nanc

yca

rein

the

last

12m

onth

s?Pl

anni

ngH

ave

you

had

prev

entiv

edo

ctor

visi

tfor

fam

ilypl

anni

ngin

the

last

3m

onth

s?Ir

onH

ave

you

had

prev

entiv

edo

ctor

visi

tfor

rece

ivin

gir

onsu

pple

men

tsin

the

last

3m

onth

s?K

ids

chec

kH

ave

you

had

prev

entiv

edo

ctor

visi

tfor

chec

king

your

child

’she

alth

inth

ela

st3

mon

ths?

Prev

entiv

eca

mpa

ign

Hav

eyo

uha

dpr

even

tive

doct

orvi

sitf

orre

ceiv

ing

info

rmat

ion

onpr

even

tion

ofsi

ckne

ssin

the

last

3m

onth

s?C

urat

ive

Exp

erie

nce

heal

thpr

oble

man

dse

ekm

edic

alat

tent

ion

Hav

eyo

uex

peri

ence

dan

ysy

mpt

om,i

llnes

s,re

laps

eor

acci

dent

and

soug

htm

edic

alat

tent

ion

inth

ela

st4

wee

ks?

Exp

erie

nce

heal

thpr

oble

man

ddo

ctor

visi

ts1/

.H

ave

you

expe

rien

ced

any

sym

ptom

,illn

ess,

rela

pse

orac

cide

ntan

dso

ught

ado

ctor

inth

ela

st4

wee

ks?

Exp

erie

nce

heal

thpr

oble

man

dm

edic

ines

1/.

Hav

eyo

uex

peri

ence

dan

ysy

mpt

om,i

llnes

s,re

laps

eor

acci

dent

and

rece

ived

med

icin

esin

the

last

4w

eeks

?E

xper

ienc

ehe

alth

prob

lem

and

anal

ysis

1/.

Hav

eyo

uex

peri

ence

dan

ysy

mpt

om,i

llnes

s,re

laps

eor

acci

dent

and

rece

ived

anal

ysis

inth

ela

st4

wee

ks?

Exp

erie

nce

heal

thpr

oble

man

dX

-ray

s1/

.H

ave

you

expe

rien

ced

any

sym

ptom

,illn

ess,

rela

pse

orac

cide

ntan

dre

ceiv

edX

-ray

sin

the

last

4w

eeks

?E

xper

ienc

ehe

alth

prob

lem

and

othe

rtes

ts1/

.H

ave

you

expe

rien

ced

any

sym

ptom

,illn

ess,

rela

pse

orac

cide

ntan

dre

ceiv

edot

hert

ests

inth

ela

st4

wee

ks?

Not

es:

Tabl

e1

inth

em

ain

text

repo

rts

sum

mar

yst

atis

tics.

1/.

Indi

vidu

als

are

aske

dto

resp

ond

thes

equ

estio

nson

lyif

they

have

had

answ

ered

“yes

”to

the

prev

ious

ques

tion

rela

ted

tose

ekat

tent

ion

afte

rhav

ing

expe

rien

ced

ahe

alth

prob

lem

(“H

ave

you

expe

rien

ced

any

sym

ptom

,illn

ess,

rela

pse

orac

cide

ntan

dso

ught

med

ical

atte

ntio

nin

the

last

4w

eeks

?”).

A3

Tabl

e10

:Var

iabl

eD

efini

tions

(con

tinue

d)

Var

iabl

esD

efini

tion

Hea

lthre

port

Sym

ptom

Hav

eyo

uha

da

sym

ptom

orhe

alth

prob

lem

inth

ela

st4

wee

ks?

(i.e

.hea

dach

e,fe

ver,

naus

ea)

Illn

ess

Hav

eyo

ube

enill

inth

ela

st4

wee

ks?

(i.e

.flu,

colit

is)

Chr

onic

illne

ssD

oyo

uha

vea

chro

nic

illne

ssor

heal

thpr

oble

m?

(i.e

.art

hriti

s,hy

pert

ensi

on,d

iabe

tes)

Rel

apse

Hav

eyo

uha

da

rela

pse

inyo

urch

roni

cill

ness

inth

ela

st4

wee

ks?

(i.e

.art

hriti

s,hy

pert

ensi

on,d

iabe

tes)

Acc

iden

tH

ave

you

had

anac

cide

ntin

the

last

4w

eeks

?N

um.d

ays

with

sym

ptom

Num

bero

fday

syo

uco

uld

notp

erfo

rmno

rmal

activ

ities

beca

use

ofa

sym

ptom

Num

.day

sw

ithill

ness

Num

bero

fday

syo

uco

uld

notp

erfo

rmno

rmal

activ

ities

beca

use

ofan

illne

ssN

um.d

ays

with

rela

pse

Num

bero

fday

syo

uco

uld

notp

erfo

rmno

rmal

activ

ities

beca

use

ofa

rela

pse

Num

.day

sw

ithac

cide

ntN

umbe

rofd

ays

you

coul

dno

tper

form

norm

alac

tiviti

esbe

caus

eof

anac

cide

ntO

ut-o

f-Po

cket

Spen

ding

Any

heal

thex

pend

iture

sA

nyan

nual

heal

thex

pend

iture

sH

ealth

expe

nditu

res

Ann

ualh

ealth

expe

nditu

res

Abs

olut

ede

viat

ion

heal

thex

pend

iture

sV

aria

bilit

yof

annu

alhe

alth

expe

nditu

res

Abs

olut

eva

lue

resi

dual

expe

nditu

res

Abs

olut

eva

lue

ofre

sidu

alfr

omre

gres

sion

sfo

rann

ualh

ealth

expe

nditu

res

Squa

rere

sidu

alex

pend

iture

sSq

uare

ofre

sidu

alfr

omre

gres

sion

sfo

rann

ualh

ealth

expe

nditu

res

Exp

endi

ture

sex

ceed

med

ian

Ann

ualh

ealth

expe

nditu

res

exce

ed50

thpe

rcen

tile

(med

ian)

Exp

endi

ture

sex

ceed

75th

perc

entil

eA

nnua

lhea

lthex

pend

iture

sex

ceed

75th

perc

entil

e(t

hird

quar

tile)

Hea

lthex

pend

iture

sas

shar

eof

inco

me

Shar

eex

pend

iture

sSh

are

annu

alhe

alth

expe

nditu

res

ofan

nual

hous

ehol

dpe

rcap

itain

com

eA

bsol

ute

devi

atio

nsh

are

Var

iabi

lity

ofSh

are

annu

alhe

alth

expe

nditu

res

ofan

nual

hous

ehol

dpe

rcap

itain

com

eA

bsol

ute

valu

ere

sidu

alsh

are

Abs

olut

eva

lue

ofre

sidu

als

from

regr

essi

onfo

rSha

rean

nual

heal

thex

pend

iture

sof

annu

alho

useh

old

perc

apita

inco

me

Squa

rere

sidu

alex

pend

iture

sSq

uare

ofre

sidu

als

from

regr

essi

onfo

rsha

reof

annu

alhe

alth

expe

nditu

res

onan

nual

hous

ehol

dpe

rcap

itain

com

eE

xpen

ditu

res

exce

edm

edia

nSh

are

annu

alhe

alth

expe

nditu

res

ofan

nual

hous

ehol

dpe

rcap

itain

com

eex

ceed

50th

perc

entil

e(m

edia

n)E

xpen

ditu

res

exce

ed75

thpe

rcen

tile

Shar

ean

nual

heal

thex

pend

iture

sof

annu

alho

useh

old

perc

apita

inco

me

exce

ed75

thpe

rcen

tile

(thi

rdqu

artil

e)C

atas

thro

phic

heal

thex

pend

iture

sex

ceed

s5%

ofpe

rca

pita

hous

ehol

din

com

eSh

are

annu

alhe

alth

expe

nditu

res

ofan

nual

hous

ehol

dpe

rcap

itain

com

eex

ceed

5%ex

ceed

s10

%of

per

capi

taho

useh

old

inco

me

Shar

ean

nual

heal

thex

pend

iture

sof

annu

alho

useh

old

perc

apita

inco

me

exce

ed10

%ex

ceed

s15

%of

per

capi

taho

useh

old

inco

me

Shar

ean

nual

heal

thex

pend

iture

sof

annu

alho

useh

old

perc

apita

inco

me

exce

ed15

%ex

ceed

s20

%of

per

capi

taho

useh

old

inco

me

Shar

ean

nual

heal

thex

pend

iture

sof

annu

alho

useh

old

perc

apita

inco

me

exce

ed20

%ex

ceed

s25

%of

per

capi

taho

useh

old

inco

me

Shar

ean

nual

heal

thex

pend

iture

sof

annu

alho

useh

old

perc

apita

inco

me

exce

ed25

%

Not

es:S

eeal

soSe

ctio

n6.

3fo

rfur

ther

expl

anat

ions

.Tab

le2

inth

em

ain

text

repo

rts

sum

mar

yst

atis

tics.

A4

C Further Details on Eligibility for SIS

C.1 Variables and Weights for IFH Construction

The IFH index is constructed such that it takes on values between 0 and 100. Higher valuesindicate better living conditions. In the following, we explain how the index is calculated.SISFOH (2010) provides more details.

First, the ENAHO for the year 2009 was used to determine the set of variables that entersinto the IFH computation. The Sommers test was used to identify correlation between candidateexplanatory variables and a measure of poverty. Then, significant variables were selected anda Principal Component analysis for discrete variables was applied to reduce dimensions and tofocus on those variables that mainly explain the variability of the data. The weights that areused to construct the index correspond to the contribution of the respective variables to the firstprincipal component. This was done separately for three geographic areas, the Lima Province,the other urban areas, and all rural areas.

Table 11 and 12 show the variables, the mutually exclusive alternatives and the correspond-ing weights. There are three independent sets of weight that correspond to households livingin different geographic areas. For instance, consider a household from Lima that cooks withcarbon, uses water from outside the house and lives in a house with brick walls. Then, the firstthree addends of the IFH index are -0.33, -0.35 and 0.10.1

Using those weights a raw index i f hi j is calculated as a linear combination of householdcharacteristics with cluster-specific weights. Then, it is standardized so that it lies between 0and 100 in each cluster. The standardized index is

i f h′i j = 100∗i f hi j− i f hmin

j

i f hmaxj − i f hmin

j,

where i f h′i j is the adjusted IFH that lies in the interval [0,100] and i f hminj and i f hmax

j are theminimum and the maximum values of the original IFH index in cluster j, respectively.

1Importantly, the number of members of the household with health insurance does not include those with SIS.This is important because otherwise, our third assumption in Section 5, the exclusion restriction, would likely beviolated.

A5

Table 11: Variables and weights for IFH construction

Variables MetropolitanLima

remaining urban areas rural areas

Fuel used to cookDo not cook -0.49 -0.67 -0.76Other -0.40 -0.50 -0.38Firewood -0.37 -0.33 0.05Carbon -0.33 -0.22 0.36Kerosine -0.29 -0.19 0.37Gas 0.02 0.12 0.52Electricity 0.43 0.69 0.52

Water supply in the homeOther -0.78 -0.58River -0.65 -0.42Well -0.62 -0.37Water tanker -0.51 -0.34Pipe -0.41 -0.32Outside -0.35 -0.25Inside 0.10 0.12

Wall materialOther -0.70 -0.80Wood or mat -0.48 -0.55Stone with mud -0.44 -0.46Rushes covered with mud -0.41 -0.43Clay -0.39 -0.38Sun-dried brick or adobe -0.37 -0.20Stones, lime or concrete -0.33 -0.07Brick 0.10 0.25

Type of drainageNone -0.89 -0.68River -0.75 -0.49Sinkhole -0.59 -0.40Septic tank -0.46 -0.30Drainage system outside the house -0.39 -0.21Drainage system inside the house 0.10 0.20

Number of members with health insurance 1/.None -0.26 -0.25 -0.10One -0.04 0.06 0.50Two 0.06 0.17 0.59Three 0.14 0.27 0.66More than three 0.32 0.48 0.86

Goods that identify household wealthNone -0.47 -0.35 -0.11One -0.17 0.05 0.64Two 0.02 0.25 0.83Three 0.15 0.40 0.90Four 0.25 0.52 1.09Five 0.47 0.75 1.09

Has fixed phoneYes -0.32No 0.20

Notes: Taken from SISFOH (2010).1/. The number of members of the household with health insurance does not include those with SIS. This is important becauseotherwise, our third assumption in Section 5, the exclusion restriction, would likely be violated.

A6

Table 12: Variables and weights for IFH construction (Continued)

Variables MetropolitanLima

remaining urban areas rural areas

Roof materialOther -0.86 -0.90Straw -0.74 -0.72Mat -0.67 -0.62Woven cane -0.38 -0.23Tiles -0.23 0.03Wood or mat -0.21 0.07Concrete 0.17 0.32

Education of the Household headNone -0.51 -0.57 -0.59Preschool -0.43 -0.25 -0.08Primary -0.28 0.01 0.35Secondary -0.06 0.19 0.59Vocational education (VET) 0.10 0.33 0.68Undergraduate 0.22 0.55 0.88Postgraduate 0.40 0.55 0.88

Floor materialOther -0.97 -1.12Land -0.60 -0.47Concrete -0.16 -0.01Wood 0.08 0.30Tiles 0.16 0.40Vinyl sheets 0.28 0.51Parquet 0.51 0.71

OvercrowdingMore than six -0.68Between four and six -0.51Between two and four -0.31Between one and two -0.07Less than one 0.24

Highest level of education in the houseNone -0.35Primary 0.11Secondary 0.41Vocational education (VET) 0.62Undergraduate 0.83

ElectricityNo -0.29Yes 0.22

Floor made of earthYes -0.17No 0.47

Notes: Taken from SISFOH (2010).

A7

C.2 Thresholds for IFH by Cluster

Table 13: Eligibility Thresholds by Cluster

Cluster Threshold Population Per capitaincome 1/.

Per capitaspending

1/.

Povertystatus

1 33 208,101 2,184 1,815 0.51592 36 1,907,122 2,116 1,697 0.59943 34 2,284,876 2,332 1,937 0.54044 38 2,646,680 2,282 1,916 0.53895 35 634,472 2,067 1,595 0.64106 34 212,723 5,941 4,045 0.26067 52 2,544,448 5,141 4,260 0.25658 42 2,134,993 5,667 4,428 0.23979 44 3,740,611 6,403 5,050 0.135210 50 2,229,638 5,997 4,673 0.162011 44 490,207 5,498 4,015 0.272512 43 101,993 8,632 4,638 0.164513 43 1,636,740 5,045 4,024 0.211614 33 93,527 8,961 6,178 0.026115 55 9,342,700 8,712 6,612 0.1546

Peru - 30,208,831 5,793 4,501 0.2764Notes: Based on SISFOH (2010), own calculations using the ENAHO 2011. There is no threshold at the nationallevel. 1/. Soles.

To determine eligibility, there are thresholds for the IFH index by cluster. Individuals andhouseholds with an index below or equal to the threshold are eligible for SIS. Table 13 shows thethresholds by cluster. The 15 clusters were defined by identifying areas with similar monetarypoverty in the year 2009. In general, each cluster includes several unconnected geographicareas.2 As an example, consider cluster 2, which includes the rural areas of the jungle of thedepartments of Ayacucho, Junin, Loreto, Puno, San Martin and Ucayali; and also the rural areasof the northern highlands of the departments of Cajamarca and Lambayeque. The thresholdswere determined such that poor individuals, in some sense that is not spelled out, were eligible.They are conservative in the sense that they allow for type 2 errors in the sense that an individualmight be declared as eligible even though, according to the criteria that are used, is not part ofthe target population.3

Table 13 also provides some economic indicators obtained from the ENAHO for the year2011. The variation of the thresholds across clusters reflects the variation in income within Peru.The lower thresholds correspond to the poorer clusters, that is, those with the lowest levels of

2Only clusters 1, 14 and 15 include connected geographic areas.3The thresholds are set such that the marginal benefit of expanding five percentage points of coverage of the

eligible population generated a marginal increase of one percentage point in that error.

A8

Figure 3: Relationship between IFH and Expenditures Per Capita in Lima Province

56

78

9Lo

g E

xpen

ditu

re p

er c

apita

0 20 40 60 80 100IFH index

020

4060

8010

0ifh

Extremely poor Nonextremely Poor NonPoor

Notes: Based on ENAHO data for the year 2011, See Appendix C for details on how the IFH index is computed.The right figure shows box plots for expenditures by poverty status.

per capita monetary income (or spending) and the highest proportions of poor individuals.4 TheLima Province, the city under analysis, is in cluster 15, with a cutoff of 55.

We use ENAHO data for the year 2011 as well as the actual weights to re-compute the IFHindex for individuals and households in our sample. We cannot directly assess how stronglyour index is correlated with the one that was used by the government to determine eligibility.However, as an informal test, we re-produced figures illustrating the correlations between theIFH index and expenditures per capita in SISFOH (2010). Figure 3 shows the figure we obtainfor the Lima Province. Generally speaking, the reproduced figures resemble the official ones.

C.3 Further Details on Eligibility and Enrollment

The main eligibility criterion for the sample of individuals in this paper is the IFH index. Wecontrol in our analysis for the fact that individuals do not become eligible when crossing thethreshold when water and electricity expenditures are both above their respective thresholds.Here we provide more details on eligibility for SIS and on the enrollment procedure.

First, if no information for water and electricity expenditures is available, then a household

4Cluster 14, which corresponds to the urban areas of the jungle of Madre de Dios, is an interesting exception.Informal mining and illegal drugs production have increased income levels during the past years.

A9

is eligible if its IFH index is below the threshold.Second, in case one of the household members works in the formal sector, then eligibility

is related to income.5 In particular, if the monthly wage is greater than 1,500 Soles, or 544.5U.S. dollars, then the household is not eligible for a social program, unless either water orelectricity service expenditures are below their thresholds.

If eligible, individuals have the possibility to enroll into SIS at a number of places, includ-ing facilities by the Ministry of Health (MINSA). They are covered as soon as eligibility isconfirmed, which is usually a matter of days. Then, they receive the health services that areoffered at MINSA facilities and that are part of the benefit package.

Besides the main plan targeted to the poor, SIS offers two additional plans for self-employedindividuals and to employees of small firms, respectively. The latter are not seen as dependentemployees and therefore do not have to be enrolled in EsSalud, which is the public insurerprovider for formal employees. Both plans are not free of charge, but involve enrollment at arate below the actual cost. Moreover, they involve a slightly different benefit package. However,these two additional plans are not important in practice. Administrative statistics from SIS showthat the main plan targeted to the poor reaches 12.7 million individuals, or 99.8% of the entireSIS population.6 In this paper, we focus on the effects of the first plan and refer to it simply asthe SIS plan.

5One may wonder whether the SIS program creates incentives to transit from formal and informal status. In thiscontext, it is important to note that informality in Peru is not per se illegal. We have investigated this possibilityin two ways and do not find evidence for this. First, following Card and Shore-Sheppard. (2004), we exploredwhether there is evidence for SIS crowding out insurance offered by private providers or EsSalud, which would beconsistent to the idea of incentives to leave formality due to SIS. Second, we analyzed whether SIS enrollment hasa significant effect on informality. We did not find evidence in favor of either of the two.

6See SIS Statistic Report, available at http://www.sis.gob.pe/Portal/estadisticas/index.html, accessed September2013.

A10

D Sensitivity Analysis

Figure 4: McCrary Test for Discontinuity of the Density of the IFH Index at the Threshold

0.0

1.0

2.0

3.0

4.0

5

−50 0 50

Notes: The figure shows an estimate of the density of the IFH index and allows for a discontinuity at the threshold.Based on 1,129 observations at the household level. Bin size of 0.25 and bandwidth of 5.0. The size of thediscontinuity is estimated to be 0.226, with a standard error of 0..331.

A11

Table 14: Results for Subsamples of Eligibles and Ineligibles using IFH index with HypotheticalThreshold equal to Respective Median 1/2

IFH index < threshold IFH index > thresholdEstimates Ste. Estimates Ste.

More general forms of care usually provided by health care centerDoctor visits -0.0379 (0.0455) 0.0540 (0.0450)Medicines -0.0367 (0.0470) 0.0187 (0.0478)Analysis -0.0152 (0.0231) -0.0355 (0.0247)X-rays -0.0135 (0.0181) -0.0174 (0.0195)Other tests 0.0002 (0.0103) 0.0077 (0.0104)Glasses -0.0242* (0.0141) -0.0172 (0.0231)Other treatments -0.0772** (0.0391) 0.0312 (0.0415)

Care provided by hospitalHospital -0.0303 (0.0222) -0.0067 (0.0233)Surgery -0.0310* (0.0185) 0.0021 (0.0188)Child birth -0.0197 (0.0302) 0.0073 (0.0208)Dental care -0.0071 (0.0283) -0.0515 (0.0338)Ophthalmological care -0.0279* (0.0143) -0.0465* (0.0255)

Likely preventive careVaccines -0.0669** (0.0310) 0.0752*** (0.0275)Birth control 0.0243 (0.0216) 0.0278 (0.0247)Pregnancy care -0.0670 (0.0476) 0.0353 (0.0327)Planning 0.0495 (0.0589) -0.0175 (0.0625)Iron -0.1512 (0.1103) -0.0971 (0.1230)Kids check -0.0548 (0.0786) 0.1154 (0.1180)Preventive campaign -0.0108 (0.0202) 0.0229 (0.0170)

CurativeExperience health problem and seek medical attention -0.0218 (0.0382) 0.0153 (0.0382)Experience health problem and doctor visits -0.0292 (0.0381) 0.0177 (0.0382)Experience health problem and medicines -0.0334 (0.0377) 0.0003 (0.0374)Experience health problem and analysis -0.0129 (0.0222) -0.0281 (0.0235)Experience health problem and X-rays -0.0121 (0.0178) -0.0100 (0.0187)Experience health problem and other tests 0.0023 (0.0069) 0.0066 (0.0084)

Health reportSymptom 0.0328 (0.0471) -0.0100 (0.0469)Illness -0.0036 (0.0354) 0.0520 (0.0332)Chronic illness -0.0071 (0.0421) 0.0116 (0.0451)Relapse 0.0215 (0.0252) -0.0418 (0.0315)Accident -0.0224 (0.0158) 0.0044 (0.0122)Num. days with symptom -0.0355 (0.0860) 0.0613 (0.0951)Num. days with illness 0.1509 (0.1234) -0.0574 (0.0950)Num. days with relapse 0.0454 (0.1204) -0.3444 (0.2341)Num. days with accident -0.1035 (0.1048) -0.0762 (0.0703)

Notes: Eligibles: N=1,786 (total); N=363 (kids check); N=532 (pregnancy care); N=532 (child birth); N=532(planning); N=177 (iron). Ineligibles: N=2,375 (total); N=286 (kids check); N=650 (pregnancy care); N=650(child birth); N=649 (planning); N=166 (iron). Standard errors are denoted by Ste. and reported in parentheses. "*p<0.10 ** p<0.05 *** p<0.01.

.

A12

Table 15: Effect of Eligibility using IFH on Health Care Utilization (No Controls and ReducedSamples)

Full sample no controls Sample 75% Sample 50%Estimates Ste. Estimates Ste. Estimates Ste.

More general forms of care usually provided by health care centerDoctor visits 0.0765** (0.0307) 0.0866** (0.0364) 0.0474 (0.0487)Medicines 0.1039*** (0.0324) 0.1375*** (0.0383) 0.1098** (0.0508)Analysis 0.0300* (0.0160) 0.0431** (0.0186) 0.0183 (0.0245)X-rays 0.0202 (0.0126) 0.0353** (0.0152) 0.0236 (0.0193)Other tests 0.0100 (0.0076) 0.0027 (0.0092) 0.0040 (0.0140)Glasses -0.0008 (0.0123) 0.0030 (0.0138) -0.0093 (0.0167)Other treatments 0.0465* (0.0269) 0.0051 (0.0319) -0.0412 (0.0417)

Care provided by hospitalHospital 0.0246 (0.0162) 0.0386** (0.0190) 0.0504** (0.0255)Surgery 0.0361*** (0.0131) 0.0445*** (0.0154) 0.0349* (0.0197)Child birth 0.0273 (0.0266) 0.0040 (0.0353) 0.0401 (0.0463)Dental care 0.0077 (0.0206) 0.0339 (0.0246) 0.0203 (0.0323)Ophthalmology -0.0022 (0.0147) 0.0016 (0.0170) -0.0153 (0.0222)

Likely preventive careVaccines 0.0412* (0.0214) 0.0366 (0.0256) 0.0451 (0.0356)Birth control -0.0205 (0.0159) -0.0244 (0.0190) -0.0036 (0.0253)Pregnancy care 0.0951** (0.0372) 0.0851* (0.0471) 0.0896 (0.0603)Planning 0.0396 (0.0389) 0.0491 (0.0447) 0.1231** (0.0587)Iron 0.0671 (0.0782) 0.0835 (0.0913) -0.0564 (0.1143)Kids check -0.0443 (0.0783) -0.0581 (0.0691) -0.0855 (0.0960)Preventive campaign 0.0029 (0.0121) -0.0199 (0.0142) -0.0396** (0.0184)

CurativeExperience health problem and seek medical attention 0.0824*** (0.0263) 0.0856*** (0.0311) 0.0459 (0.0422)Experience health problem and doctor visits 0.0796*** (0.0263) 0.0820*** (0.0310) 0.0391 (0.0420)Experience health problem and medicines 0.0765*** (0.0259) 0.0835*** (0.0306) 0.0400 (0.0413)Experience health problem and analysis 0.0289* (0.0153) 0.0388** (0.0177) 0.0125 (0.0232)Experience health problem and X-rays 0.0152 (0.0119) 0.0275* (0.0141) 0.0139 (0.0178)Experience health problem and other tests 0.0089 (0.0058) 0.0016 (0.0069) 0.0057 (0.0100)

Health reportAny symptom 0.0240 (0.0320) 0.0191 (0.0384) 0.0104 (0.0508)Illness 0.0612*** (0.0228) 0.0638** (0.0269) 0.0393 (0.0358)Chronic illness 0.0315 (0.0318) 0.0603* (0.0354) 0.0401 (0.0471)Relapse -0.0091 (0.0197) -0.0030 (0.0231) 0.0100 (0.0314)Accident 0.0145 (0.0104) 0.0280** (0.0125) 0.0155 (0.0168)Num. days with symptom 0.0122 (0.0846) 0.0732 (0.1075) 0.0861 (0.1581)Num. days with illness 0.1358** (0.0688) 0.1062 (0.0684) 0.0865 (0.0882)Num. days with relapse 0.1409 (0.1451) 0.2049 (0.1724) 0.3706 (0.2521)Num. days with accident 0.0732 (0.0647) 0.1768** (0.0896) 0.1129 (0.1089)

Notes: 1/. Full: N=4,161 (total); N=649 (kids check); N=1.182 (pregnancy care); N=1,182 (child birth). 2/. Sample 75: N=3,124 (total);N=499 (kids check); N=892 (pregnancy care) N=892 (child birth). 3/. Sample 50: N=2,078 (total); N=347 (kids check); N=618 (pregnancycare) N=618 (child birth). 5/. Standard errors in parentheses * p<0.10 ** p<0.05 *** p<0.01.

A13

Tabl

e16

:Eff

ecto

fElig

ibili

tyus

ing

IFH

onE

xpen

ditu

res

(No

Con

trol

san

dR

educ

edSa

mpl

es)

Full

sam

ple

noco

ntro

lsSa

mpl

e75

%Sa

mpl

e50

%E

stim

ates

Ste.

Est

imat

esSt

e.E

stim

ates

Ste.

Any

heal

thex

pend

iture

s0.

0455

(0.0

322)

0.06

33*

(0.0

384)

0.01

97(0

.050

4)

Hea

lthca

reex

pend

iture

sH

ealth

expe

nditu

res

158.

5857

**(7

1.99

43)

163.

1471

**(7

6.26

21)

144.

9099

(93.

1877

)A

bsol

ute

devi

atio

nex

pend

iture

s-4

6.57

42(6

4.46

04)

-72.

5885

(67.

1493

)-4

3.02

28(8

1.90

61)

Abs

olut

eva

lue

resi

dual

expe

nditu

res

105.

1096

*(6

3.43

04)

105.

4242

(67.

1828

)13

0.60

63(8

2.25

40)

Squa

red

resi

dual

expe

nditu

res

1.36

3e+0

6*(7

.845

e+05

)7.

277e

+05

(6.0

60e+

05)

9.26

5e+0

5(7

.559

e+05

)E

xpen

ses

exce

edm

edia

n0.

1701

***

(0.0

323)

0.19

97**

*(0

.038

4)0.

1498

***

(0.0

507)

Exp

ense

sex

ceed

75th

perc

entil

e0.

1377

***

(0.0

274)

0.15

03**

*(0

.032

2)0.

1274

***

(0.0

422)

Hea

lthex

pend

iture

sas

ash

are

ofin

com

eSh

are

expe

nditu

res

0.01

30(0

.012

6)0.

0246

(0.0

156)

0.04

02*

(0.0

226)

Abs

olut

ede

viat

ion

shar

e0.

0023

(0.0

116)

0.00

89(0

.014

5)0.

0265

(0.0

211)

Abs

olut

eva

lue

resi

dual

shar

e0.

0073

(0.0

115)

0.01

33(0

.014

4)0.

0312

(0.0

210)

Squa

red

resi

dual

expe

nditu

res

0.03

08(0

.037

0)0.

0419

(0.0

468)

0.09

13(0

.074

0)Sh

are

exce

eds

med

ian

0.11

64**

*(0

.032

4)0.

1586

***

(0.0

387)

0.11

40**

(0.0

510)

Shar

eex

ceed

s75

thpe

rcen

tile

0.07

72**

*(0

.027

7)0.

1098

***

(0.0

331)

0.11

43**

*(0

.044

0)

Cat

astr

ophi

che

alth

expe

nditu

res

exce

eds

5%of

per

capi

taho

useh

old

inco

me

0.06

37**

(0.0

270)

0.09

54**

*(0

.032

2)0.

1102

**(0

.042

8)ex

ceed

s10

%of

per

capi

taho

useh

old

inco

me

0.01

94(0

.022

1)0.

0397

(0.0

262)

0.02

91(0

.034

4)ex

ceed

s15

%of

per

capi

taho

useh

old

inco

me

0.01

83(0

.018

7)0.

0445

**(0

.022

3)0.

0502

*(0

.029

4)ex

ceed

s20

%of

per

capi

taho

useh

old

inco

me

0.01

99(0

.016

2)0.

0437

**(0

.019

6)0.

0390

(0.0

262)

exce

eds

25%

ofpe

rca

pita

hous

ehol

din

com

e0.

0074

(0.0

142)

0.03

07*

(0.0

172)

0.03

50(0

.023

9)

Not

es:.

1/.F

ull:

N=4

,161

(tot

al);

N=6

49(k

ids

chec

k);N

=1.1

82(p

regn

ancy

care

);N

=1,1

82(c

hild

birt

h).2

/.Sa

mpl

e75

:N=3

,124

(tot

al);

N=4

99(k

ids

chec

k);N

=892

(pre

gnan

cyca

re)N

=892

(chi

ldbi

rth)

.3/.

Sam

ple

50:

N=2

,078

(tot

al);

N=3

47(k

ids

chec

k);N

=618

(pre

gnan

cyca

re)N

=618

(chi

ldbi

rth)

.5/.

Stan

dard

erro

rsin

pare

nthe

ses

*p<

0.10

**p<

0.05

***

p<0.

01.

A14

Tabl

e17

:Eff

ecto

fElig

ibili

tyus

ing

IFH

onE

xpec

tatio

nof

Cov

aria

tes

(1)

(2)

(3)

Full

sam

ple

Sam

ple

75%

Sam

ple

50%

Est

imat

esSt

e.N

Est

imat

esSt

e.N

Est

imat

esSt

e.N

Wom

an1/

.-0

.004

4(0

.032

6)41

61-0

.026

9(0

.039

3)31

240.

0159

(0.0

517)

2078

Age

1/.

-1.7

523

(1.4

376)

4161

0.40

63(1

.702

7)31

24-1

.409

2(2

.237

1)20

78Y

ears

ofed

ucat

ion

1/.

0.20

92(0

.295

7)41

61-0

.065

6(0

.356

9)31

24-0

.779

1*(0

.470

2)20

78N

umbe

rhou

seho

ldm

embe

rs2/

.-0

.084

8(0

.213

6)11

29-0

.351

0(0

.259

5)84

6-0

.193

6(0

.349

2)56

9W

omen

ashe

adof

hous

ehol

d2/

.-0

.006

5(0

.056

0)11

290.

0592

(0.0

692)

846

0.10

25(0

.089

3)56

9L

ogof

hous

ehol

din

com

e2/

.0.

0198

(0.1

231)

1115

-0.2

449

(0.1

494)

839

-0.4

329*

*(0

.209

0)56

5L

ogof

hous

ehol

dex

pend

iture

2/.

0.02

75(0

.059

0)11

29-0

.100

9(0

.070

3)84

6-0

.154

1(0

.096

3)56

9L

ogof

wat

erco

nsum

ptio

n2/

.0.

0407

(0.0

690)

985

-0.0

224

(0.0

785)

739

0.02

79(0

.098

5)50

3L

ogof

elec

tric

ityco

nsum

ptio

n2/

.0.

0095

(0.0

832)

1004

-0.0

274

(0.0

985)

750

-0.0

167

(0.1

347)

504

Not

es:

N=

4,16

1.St

anda

rder

rors

inpa

rent

hese

s.*

p<0.

10**

p<0.

05**

*p<

0.00

.1/

.In

divi

dual

leve

l.2/

.H

ouse

hold

leve

l.Z

ero

hous

ehol

din

com

e,ho

useh

old

expe

nditu

re,w

ater

and

elec

tric

ityco

nsum

ptio

nw

ere

code

das

mis

sing

.

A15

Tabl

e18

:Loc

alL

inea

rReg

ress

ions

1/2

Cor

resp

ondi

ngre

sult

Loc

alL

inea

rL

ocal

Qua

drat

ic10

2050

100

1020

5010

0

Tabl

e3

Doc

torv

isits

0.06

31*

0.06

87**

*0.

0687

**0.

0689

***

0.05

430.

0823

**0.

0729

**0.

0720

**(0

.033

7)(0

.026

5)(0

.030

5)(0

.023

9)(0

.056

4)(0

.035

0)(0

.036

3)(0

.036

7)Ta

ble

3M

edic

ines

0.08

63**

0.08

07**

*0.

0709

***

0.07

01**

*0.

0668

0.10

13**

0.10

08**

*0.

1006

***

(0.0

383)

(0.0

272)

(0.0

261)

(0.0

266)

(0.0

558)

(0.0

400)

(0.0

358)

(0.0

382)

Tabl

e3

Ana

lysi

s0.

0277

*0.

0285

*0.

0287

**0.

0286

**-0

.011

4**

0.03

32*

0.03

53**

0.03

54**

(0.0

168)

(0.0

149)

(0.0

137)

(0.0

125)

(0.0

222)

(0.0

182)

(0.0

163)

(0.0

152)

Tabl

e3

Surg

ery

0.04

39**

*0.

0408

***

0.03

70**

*0.

0366

***

0.04

52*

0.04

65**

*0.

0488

***

0.04

90**

*(0

.015

4)(0

.012

2)(0

.010

6)(0

.011

5)(0

.023

3)(0

.016

9)(0

.016

2)(0

.014

4)Ta

ble

3Pr

egna

ncy

care

0.06

540.

0782

**0.

0828

***

0.08

33**

*0.

0052

0.05

690.

0525

0.05

23(0

.041

3)(0

.032

5)(0

.030

2)(0

.031

6)(0

.052

4)(0

.047

9)(0

.044

6)(0

.037

8)Ta

ble

3A

nyut

iliza

tion

0.04

920.

0593

**0.

0592

**0.

0593

**0.

0292

0.06

310.

0576

*0.

0568

*(0

.039

3)(0

.024

3)(0

.023

1)(0

.024

7)(0

.049

2)(0

.038

4)(0

.034

9)(0

.030

8)

Tabl

e4

Exp

.pro

blem

and

med

ical

atte

ntio

n0.

0833

**0.

0796

***

0.07

56**

*0.

0752

***

0.05

890.

0899

**0.

0903

***

0.09

04**

*(0

.033

2)(0

.022

8)(0

.021

5)(0

.020

5)(0

.042

4)(0

.035

1)(0

.034

4)(0

.030

4)Ta

ble

4E

xp.p

robl

eman

ddo

ctor

visi

t0.

0776

**0.

0783

***

0.07

44**

*0.

0741

***

0.04

920.

0838

***

0.08

68**

*0.

0870

***

(0.0

320)

(0.0

228)

(0.0

202)

(0.0

235)

(0.0

388)

(0.0

287)

(0.0

285)

(0.0

309)

Tabl

e4

Exp

.pro

blem

and

med

icin

es0.

0720

**0.

0730

***

0.06

88**

*0.

0684

***

0.03

750.

0767

***

0.08

04**

*0.

0807

***

(0.0

283)

(0.0

228)

(0.0

206)

(0.0

211)

(0.0

438)

(0.0

279)

(0.0

305)

(0.0

301)

Tabl

e4

Exp

.pro

blem

and

anal

ysis

0.01

860.

0239

*0.

0248

**0.

0248

*-0

.018

50.

0219

0.02

57*

0.02

59*

(0.0

159)

(0.0

133)

(0.0

122)

(0.0

128)

(0.0

197)

(0.0

189)

(0.0

150)

(0.0

150)

Tabl

e4

Any

cura

tive

use

0.10

51**

*0.

1020

***

0.09

83**

*0.

0980

***

0.08

61*

0.10

82**

*0.

1096

***

0.10

96**

*(0

.033

5)(0

.025

3)(0

.023

0)(0

.023

8)(0

.048

2)(0

.039

9)(0

.031

5)(0

.033

5)

Not

e:Se

eal

sono

tes

toTa

ble

3an

dTa

ble

4.L

ocal

aver

age

trea

tmen

teff

ects

are

repo

rted

for

band

wid

ths

10,2

0,50

and

100

and

for

apr

oced

ure

base

don

loca

llin

ear

and

loca

lqua

drat

icre

gres

sion

s,re

spec

tivel

y.B

oots

trap

ped

stan

dard

erro

rsin

pare

nthe

ses.

*p<

0.10

**p<

0.05

***

p<0.

00.

A16

Tabl

e19

:Loc

alL

inea

rReg

ress

ions

2/2

Cor

resp

ondi

ngre

sult

Loc

alL

inea

rL

ocal

Qua

drat

ic10

2050

100

1020

5010

0

Tabl

e5

Hea

lthex

pend

iture

197.

1291

***

149.

1220

***

141.

4523

***

140.

7061

***

164.

4689

*18

8.23

23**

172.

5285

**17

1.63

04**

(70.

1100

)(5

1.79

58)

(48.

0938

)(5

0.40

22)

(98.

5298

)(8

6.20

46)

(77.

1523

)(8

3.56

88)

Tabl

e5

Exp

en.e

xcee

dsm

edia

n0.

1346

***

0.12

71**

*0.

1260

***

0.12

59**

*0.

1021

**0.

1464

***

0.13

72**

*0.

1365

***

(0.0

371)

(0.0

275)

(0.0

280)

(0.0

274)

(0.0

542)

(0.0

402)

(0.0

421)

(0.0

383)

Tabl

e5

Exp

en.e

xcee

ds75

thpc

tile.

0.10

68**

*0.

0871

***

0.09

49**

*0.

0955

***

0.08

88*

0.10

50**

*0.

0879

***

0.08

70**

*(0

.029

6)(0

.025

7)(0

.022

8)(0

.021

9)(0

.047

3)(0

.033

1)(0

.032

9)(0

.032

2)Ta

ble

5Sh

are

exce

eds

med

ian

0.12

02**

*0.

1031

***

0.09

56**

*0.

0950

***

0.10

38*

0.13

60**

*0.

1285

***

0.12

78**

*(0

.035

9)(0

.028

2)(0

.024

6)(0

.023

2)(0

.058

2)(0

.040

2)(0

.036

3)(0

.037

0)Ta

ble

5Sh

are

exce

eds

75th

pctil

e.0.

1021

***

0.06

58**

*0.

0617

***

0.06

13**

*0.

1000

*0.

1058

***

0.08

97**

*0.

0888

***

(0.0

288)

(0.0

240)

(0.0

229)

(0.0

219)

(0.0

549)

(0.0

331)

(0.0

313)

(0.0

364)

Tabl

e5

Exp

en.e

xcee

ds5%

inco

me

0.10

28**

*0.

0639

***

0.05

81**

0.05

75**

0.10

83**

0.10

88**

*0.

0920

***

0.09

10**

*(0

.033

6)(0

.021

6)(0

.024

8)(0

.027

0)(0

.048

4)(0

.035

1)(0

.032

0)(0

.031

3)

Tabl

e6

Illn

ess

0.03

350.

0411

*0.

0431

**0.

0434

**0.

0671

*0.

0484

*0.

0406

0.03

98(0

.029

2)(0

.021

8)(0

.018

6)(0

.021

1)(0

.040

2)(0

.027

2)(0

.026

3)(0

.028

1)Ta

ble

6N

um.d

ays

with

illne

ss0.

0624

0.08

12*

0.08

46*

0.08

48*

0.03

120.

0780

0.08

450.

0849

(0.0

582)

(0.0

467)

(0.0

483)

(0.0

493)

(0.0

521)

(0.0

538)

(0.0

566)

(0.0

623)

Tabl

e6

Any

inci

dent

0.09

33**

*0.

0787

***

0.07

80**

*0.

0780

***

0.11

15**

0.09

88**

*0.

0868

**0.

0860

**(0

.034

4)(0

.027

3)(0

.026

2)(0

.025

3)(0

.047

5)(0

.034

6)(0

.039

1)(0

.038

4)Ta

ble

6N

days

with

any

inci

dent

0.62

11**

0.40

32**

0.32

32*

0.31

54*

0.44

180.

7331

***

0.66

84**

*0.

6655

***

(0.2

473)

(0.1

828)

(0.1

759)

(0.1

660)

(0.3

802)

(0.2

765)

(0.2

588)

(0.2

277)

Elig

ible

0.46

80**

*0.

5673

***

0.58

76**

*0.

5894

***

0.43

04**

*0.

4798

***

0.50

96**

*0.

5112

***

(0.0

300)

(0.0

179)

(0.0

183)

(0.0

169)

(0.0

405)

(0.0

317)

(0.0

277)

(0.0

225)

Not

e:Se

eal

sono

tes

toTa

ble

5an

dTa

ble

6.L

ocal

aver

age

trea

tmen

tef

fect

sar

ere

port

edfo

rba

ndw

idth

s10

,20

,50

and

100

and

for

apr

oced

ure

base

don

loca

llin

ear

and

loca

lqu

adra

ticre

gres

sion

s,re

spec

tivel

y.B

oots

trap

ped

stan

dard

erro

rsin

pare

nthe

ses.

*p<

0.10

**p<

0.05

***

p<0.

00.

A17

E Additional Figures

Figure 5: Expectations of Covariates

0.40

0.45

0.50

0.55

0.60

Mea

n of

wom

en

−60 −40 −20 0 20 40IFH index minus threshold

Women

20.0

030

.00

40.0

050

.00

Mea

n of

age

−60 −40 −20 0 20 40IFH index minus threshold

Age

4.00

6.00

8.00

10.0

012.0

014.0

0M

ean

of y

ears

of e

duca

tion

−60 −40 −20 0 20 40IFH index minus threshold

Years of education

2.00

3.00

4.00

5.00

6.00

Mea

n of

num

ber

of h

ouse

hold

mem

bers

−60 −40 −20 0 20 40IFH index minus threshold

Number of household members

0.00

0.20

0.40

0.60

0.80

Mea

n of

wom

en a

s he

ad o

f hou

seho

ld

−60 −40 −20 0 20 40IFH index minus threshold

Women as head of household

6.00

7.00

8.00

9.00

Mea

n of

log

of h

ouse

hold

inco

me

−60 −40 −20 0 20 40IFH index minus threshold

Household income

6.50

7.00

7.50

8.00

8.50

Mea

n of

log

of h

ouse

hold

exp

endi

ture

s

−60 −40 −20 0 20 40IFH index minus threshold

Household expenditures

2.50

3.00

3.50

4.00

Mea

n of

log

of la

st m

onth

wat

er c

onsu

mpt

ion

−60 −40 −20 0 20 40IFH index minus threshold

Water consumption

2.50

3.00

3.50

4.00

4.50

5.00

Mea

n of

log

of la

st m

onth

ele

ctric

ity c

onsu

mpt

ion

−60 −40 −20 0 20 40IFH index minus threshold

Electricity consumption

Notes: The dots denote averages, with a bin width of 3. Their size represents the number of observations. The regression lines withcorresponding 95 percent confidence intervals stem from separate linear regressions to the left and to the right of the threshold using theindividual-level data. The graphs in the top row are constructed at the individual level, the remaining graphs are constructed at the householdlevel.

A18

Figure 6: Health Expenditures for Any Doctor Visit (left) and Medicines (right)

050

01,

000

1,50

02,

000

Exp

endi

ture

s −

Doc

tor

visi

ts

0.0 0.2 0.4 0.6 0.8 1.0Quantile

Not covered

Covered

050

01,

000

1,50

02,

000

Exp

endi

ture

s −

Med

icin

es

0.0 0.2 0.4 0.6 0.8 1.0Quantile

Not covered

Covered

A19

Figu

re7:

Eff

ecto

fSIS

onH

ealth

Car

eU

tiliz

atio

n

0.150.200.250.300.350.40Mean of doctor visits

−60

−40

−20

020

40IF

H in

dex

min

us th

resh

old

Doc

tor

visi

ts

0.300.350.400.450.500.55Mean of medicines

−60

−40

−20

020

40IF

H in

dex

min

us th

resh

old

Med

icin

es

0.000.050.100.15Mean of analysis

−60

−40

−20

020

40IF

H in

dex

min

us th

resh

old

Ana

lysi

s

−0.050.000.050.100.15Mean of x−rays

−60

−40

−20

020

40IF

H in

dex

min

us th

resh

old

X−

rays

−0.020.000.020.040.06Mean of oher tests

−60

−40

−20

020

40IF

H in

dex

min

us th

resh

old

Oth

er te

sts

0.000.050.100.150.20Mean of glasses

−60

−40

−20

020

40IF

H in

dex

min

us th

resh

old

Gla

sses

0.100.200.300.400.500.60Mean of others

−60

−40

−20

020

40IF

H in

dex

min

us th

resh

old

Oth

ers

0.000.050.100.15Mean of hospital

−60

−40

−20

020

40IF

H in

dex

min

us th

resh

old

Hos

pita

l

−0.050.000.050.100.15Mean of surgery

−60

−40

−20

020

40IF

H in

dex

min

us th

resh

old

Sur

gery

−0.050.000.050.10Mean of child birth

−60

−40

−20

020

40IF

H in

dex

min

us th

resh

old

Chi

ld b

irth

0.000.050.100.150.200.25Mean of dental care

−60

−40

−20

020

40IF

H in

dex

min

us th

resh

old

Den

tal c

are

−0.050.000.050.100.150.20Mean of ophthalmological care

−60

−40

−20

020

40IF

H in

dex

min

us th

resh

old

Oph

thal

mol

ogic

al c

are

0.000.050.100.150.20Mean of vaccines

−60

−40

−20

020

40IF

H in

dex

min

us th

resh

old

Vac

cine

s

0.000.050.100.15Mean of birth control

−60

−40

−20

020

40IF

H in

dex

min

us th

resh

old

Birt

h co

ntro

l

−0.100.000.100.200.30Mean of pregnancy care

−60

−40

−20

020

40IF

H in

dex

min

us th

resh

old

Pre

gnan

cy c

are

0.000.200.400.600.80Mean of kids check

−60

−40

−20

020

40IF

H in

dex

min

us th

resh

old

Kid

s ch

eck

A20