The effect of health insurance on the use of regular and alternative healing methods by Latinos in the US

Abstract

Purpose: We analyze the effect of health insurance coverage on the utilization of services to heal an illness or injury, ranging from visits to conventional health care providers in general and psychiatrists in particular, to the use of folk healers and prayers.
Design: The data are from the Hispanic Healthcare Survey conducted in 2007 among Latin Americans living in the US. We account for potential endogeneity of insurance, acknowledging that confounding factors may affect both insurance purchase and services use; however, endogeneity does not appear to be problematic.
Findings: Results confirm that Latin Americans with insurance coverage are more likely to visit conventional health care providers, an effect that is also found for psychiatrists. We find a reduction of the probability to visit a folk healer, suggestion substitution of alternative for regular methods. An increase of the probability that others prayed for one’s recovery is found, suggesting complementarity.
Implications: Improved access to health insurance and thereby to health care services can improve immigrants’ health.
Originality: Combining the analysis of the impact of health insurance on the usage of both conventional, traditional, and religious healing methods in a particularly vulnerable population using one representative data source.

Full text

1
The effect of health insurance on the use of regular and alternative
healing methods by Latinos in the US
Edwin van Gameren
a, *
Maritza Caicedo
b
- - - unfinished revision - - -
this version: March 27, 2015
Abstract
Purpose: We analyze the effect of health insurance coverage on the utilization of
services to heal an illness or injury, ranging from visits to conventional health care
providers in general and psychiatrists in particular, to the use of folk healers and
prayers.
Design: The data are from the Hispanic Healthcare Survey conducted in 2007
among Latin Americans living in the US. We account for potential endogeneity of
insurance, acknowledging that confounding factors may affect both insurance
purchase and services use; however, endogeneity does not appear to be
problematic.
Findings: Results confirm that Latin Americans with insurance coverage are more
likely to visit conventional health care providers, an effect that is also found for
psychiatrists. We find a reduction of the probability to visit a folk healer, suggestion
substitution of alternative for regular methods. An increase of the probability that
others prayed for one’s recovery is found, suggesting complementarity.
Implications: Improved access to health insurance and thereby to health care
services can improve immigrants’ health.
Originality: Combining the analysis of the impact of health insurance on the usage of
both conventional, traditional, and religious healing methods in a particularly
vulnerable population using one representative data source.
Keywords: Health insurance; Health care use; Conventional care services; Folk
healers; Prayers; Latin Americans in the US
a Centro de Estudios Económicos, El Colegio de México, Mexico City, Mexico
b Instituto de Investigaciones Sociales, UNAM, Mexico City, Mexico.
* Corresponding author. Email: egameren@colmex.mx.
Acknowledgements: We are grateful for comments on earlier versions of the paper from participants at
the Sociedad Mexicana de Demografía (SOMEDE; Aguascalientes, Mexico) and the European
Conference on Health Economics (Zürich, Switzerland). Obviously, all remaining errors are ours.

2
Introduction
For a variety of reasons, health insurance is far from universal for Latinos living in the US,
restraining their access to regular health care services. Cheaper but not scientifically proven
healing methods may be the only option within reach of the uninsured. Furthermore, Latinos
often come from cultures where the practice of traditional folk healers such as a curandero or a
shaman is still widespread, while at the same time they are often strongly religious. While there
is an extensive literature that analyzes the impact of health insurance coverage on the use of
conventional, formally approved, medical care, the evidence regarding the impact on the use of
traditional, unproven, alternative services is more scattered, and there is little evidence on the
effect on use of religion as a means to improve health. Given the substantial and continued
immigration of Latinos in the US, it is important to understand if access to health insurance is
able to break the often disadvantaged position of Latinos in the society in general, and in
particular if it will lead to a change in the mix of services used in case of medical needs.
Our aim is to illustrate, using data from one source, if acquiring health insurance not only
increases the incentives to use conventional medical services to recover from an illness or injury,
but also reduces the relevance of traditional and religious services for healing purposes, among
immigrants from Latin American countries and their descendants. Regarding the use of
conventional health services, we look at the visits to regular doctors and health care providers in
general, and psychiatrists in particular, while for alternative care we consider visits to folk
healers, and prayers made by the respondents themselves. The survey, we use data from the
Hispanic Healthcare Survey conducted in 2007 by the Pew Hispanic Center, is representative for
the Latinos in the US, and asks specifically if the prayers are used to be healed of an illness or
injury. In addition to that specificity, the advantage of using a single data source is that it enables
comparison of the size of the effects. We account for the potential endogeneity of insurance,
taking into account that confounding factors may affect both insurance purchase and use of
services.
The next section briefly discusses the health insurance and health care context of Latinos,
as well as the literature regarding the use of health care services in response to health insurance
coverage. It is followed by the presentation of the data and the empirical methodology, and a
section that presents the results. The paper concludes with a discussion of some implications.

3
Background
Insurance status among Latinos in the US
Differences in access to health services according to place of origin, race, and immigration status
have been well-documented (Carrasquillo et al., 2000; Berk et al. 2000; Antecol and Bedard,
2006; Pitkin Derose et al., 2009). According to Carrasquillo et al. (2000), in 1997 34.3% of
immigrants in the United States had no health insurance, while only 14.2% of the US-born
population was in this situation. Among immigrants without US-citizenship, the proportion
without health insurance is higher (43.6%). Brown et al. (2000) note that, among all ethnic
groups in the United States, Latinos show the lowest health insurance rates; nearly 4 out of 10
are not insured. Especially Latinos with low income, without US-citizenship, and with limited
English proficiency are more likely to be short of health insurance and are less likely to use
health services (Ku and Waidmann, 2003).
In an analysis of the labor conditions of Latin American and Caribbean immigrants in the
United States, Caicedo (2010) reports that 61.4% of these workers lack an employer-based health
insurance. She highlights differences by country of origin and notes that Mexicans and Central
Americans are the groups with the highest percentages of workers without employer-based
insurance (66.1% and 62.6% respectively), shares that are much higher than among native non-
Hispanic whites (38.9%), and African Americans (41.2%). Brown et al. (2000) argue that there
are two key factors that reduce the access of Latinos to job-based health coverage. First,
disadvantaged by their relatively low levels of education, they are more likely than other groups
of workers to insert in the type of jobs that do not offer health insurance. Second, their low
income does not permit them to make the required contributions to obtain adequate health
insurance even when it is offered.
The considerable number of undocumented immigrants among Latinos is an aspect that
helps explain the high percentage of uninsured.
1
In order to reduce illegal immigration, in 1994
California passed Proposition 187, through which health care services are denied to
undocumented immigrants. Proponents of the law argue that such measures reduce the incentives
1
Estimates based on the Current Population Survey 2004 count about 10.3 million undocumented immigrants, who
in majority come from Mexico (57.0%), 24.0% from other countries in Latin America, while the remainder breaks
up between Asia, Europe, Canada and Africa (Passel, 2005).

4
for immigrants to enter the country irregularly. Opponents argue that this law violates the
fundamental human rights, and also point out that immigrants do not come to the US in search of
social services but with the purpose of entering the labor market to improve their living
conditions in general and their families in their countries of origin (Berk et al. 2000; Massey,
2005). In fact, health care expenditures of immigrants are generally lower than the expenditures
of US-born people (Mohanty, 2005), and it is often found that Latinos report the same or even a
better health status than the native white population (Turra and Goldman, 2007; Alegría et al.,
2008; Vega et al., 2009), a phenomenon coined as the Hispanic Health Paradox (Markides and
Coreil, 1986), although there is discussion how general the paradox is (Teruya and Bazargan-
Hejazi, 2013).
Insurance and care usage
It is well-documented in the international economics literature that an improved access to health
insurance, and therefore to the health care services covered by the insurance, increases the use of
these services. Zweifel and Manning (2000) document strong evidence for (static ex-post) moral
hazard, that is to say, an increase in the demand for medical care of a given technology due to
improved accessibility. The RAND Health Insurance Experiment (Newhouse et al., 1993) has
shown that a more generous health insurance, with lower co-insurance rates, induces more
demand for care services. More recently also the Oregon Health Insurance Experiment
(Finkelstein et al., 2012) has shown that people who were randomly assigned to enroll Medicaid
had higher medical consumption than other who were denied to enroll. Also for Latinos,
multivariate analyses have established that obtaining health insurance positively affects
utilization of conventional health care services (Wagner and Guendelman, 2000; Choi, 2006;
Ortega et al., 2007; Nandi et al., 2008). The effects are larger among (undocumented)
immigrants than among US-born people with a Latin origin (Lara et al., 2005). See also the
references in Pitkin Derose et al. (2007).
Less clear is the effect of extended health insurance on the use of non-covered traditional
or other alternative healing methods. Alternative services could be replaced by conventional care
when obtaining access through health insurance coverage (substitution), but it is also possible
that people continue to use the alternative options along with the insurance-covered services
(complements). Lack of insurance has been identified as a reason to use traditional or alternative

5
services (Ransford et al., 2010; Nahin et al., 2010; Iniguez and Palinkas, 2003), though in
general the analyses are based on small samples in specific locations. In a review of the
literature, Favazza Titus (2014) concludes that affordability and Spanish literacy are the main
reasons to seek help from traditional healers, while reasons as immigration status, culture, and
dissatisfaction with conventional (‘Western’) medicine were less common.
2
For example Reyes-
Ortiz et al. (2009) point at confusion regarding the information given at medical encounters and
perceptions about the quality of medical care as factors that led to the use of spiritual healing or
other alternative services. Nevertheless, Lopez (2005) concludes that traditional, indigenous
beliefs and practices persist even among young and high assimilated Mexican-American women.
Empirical findings for Mexicans living in Mexico imply a substitution away from alternative and
traditional medicine (Van Gameren, 2010) toward conventional health care services (Wong and
Díaz, 2007; Pagán, Puig and Soldo, 2007). We expect to find the same for Mexicans and other
Latinos in the US.
Ransford et al. (2010) not only highlight the seeking of traditional medicine as an
alternative, but also address individual prayer and faith as part of the alternative healing system
used by Latinos, positing that religion helps Latinos to gain control over health when faced with
the health system’s barriers. Respondents of their (small and non-representative) sample express
the importance of religion for prevention and cure, but emphasize that it should be combined
with seeking traditional or conventional medical attention. On the other hand, the church was not
viewed as a place to receive health support. Others such as Seybold and Hill (2001) argue that
positive effects of religious and spiritual experiences on health are based on the assumption that
the experience itself is positive and healthy. Lujan and Campbell (2006) argue that religious
practices can go together with other care services, but also review examples where faith
increases risk-taking (for example with HIV). In general, we may expect to find complementarity
between religious faith and regular medicine.
2
Several studies indicate that hospitals on the Mexican side of the US-Mexico border witness migrants returning to
Mexico in search of health care services (Wallace et al., 2009; González-Block and De la Sierra-de la Vega, 2011).

6
Methods
Data
We use data from the Hispanic Healthcare Survey, collected by the Pew Hispanic Center in 2007
through a survey among immigrants and other people from Hispanic or Latino origin or descent
in the US aged 18 years or older. The sample of 4,013 respondents, of which 1,625 were born in
the US (including Puerto Rico) and 2,378 in other countries, can be considered nationally
representative for the Latinos in the US (PHC, 2008). The questionnaire inquires about the
respondent’s socio-demographic situation, health status, the usage of health care services and
other sources of support, and the current labor status and income level. Respondents could
respond the telephone interview in their preferred language (Spanish or English).
Dependent and explanatory variables
Five different measures of conventional and alternative healing methods are used as dependent
variables in our empirical model. The use of conventional medical services is measured by two
indicators. The first indicator is derived from the survey question, ‘About how long has it been
since you last saw a doctor or another health care provider about your health?’. The five response
categories are transformed into a binary variable that indicates whether the respondent has
visited a doctor or other care provider during the previous 12 months. The first row in Table 1
shows that 80.2% of the respondents have seen a care provider, but that this share is lower
among those without insurance. The second measure, the use of mental health care, is derived
from the survey question, ‘During the past 12 months, have you seen or talked to a mental health
professional such as a psychiatrist, psychologist, psychiatric nurse, or social worker about your
health?’, with response categories Yes and No.
Furthermore, we construct three indicators for the use of alternative healing services. The
first is whether the respondent used to visit folk healers to recover from an illness, an indicator
derived from the survey question ‘Do you ever seek help from a curandero, a shaman or
someone else with special powers to heal the sick?’. The third row of Table 1 shows that folk
healers are visited by about 5.9% of the respondents. The next measure indicates whether the
respondent used to pray in case of illness, and is derived from the survey question ‘Have you
ever prayed specifically to be healed of an illness or injury?’. The fourth row in Table 1 shows

7
that about 60% of the respondents prayed to be healed. The final variable indicates whether
others prayed for the respondent’s healing, derived from the question ‘Have you EVER asked
OTHERS to pray specifically for you to be healed of an illness or injury?’. The last row in Table
1 shows that about 48% have asked others to pray. It is clear from these numbers that religion is
much more important among Latinos than traditional folk healers.
Our main explanatory variable is whether the respondents have some kind of health
insurance. With a health insurance, access to professional services at low or zero costs is
obtained, reducing the incentives to rely on less expensive but unproven methods such as
religion or folk healers. The indicator for health insurance is obtained from the question, ‘Are
you, yourself, now covered by any form of health insurance or health plan?’. In the explanation
added to the question it is emphasized that this includes privately bought insurance plans, private
insurance through the employer, and government programs such as Medicare or Medicaid. The
last line of Table 1 shows that about two-thirds of the sample has some form of health insurance,
while one-third does not.
Table 1 Dependent variables by health insurance
a
variable
full
sample
without
insurance
with
insurance
Visited doctor or other health care provider 80.2% 68.1% 86.7%
Nu
mber of observations 3408 1199 2209
Visited psychiatrist
8.7%
6.3%
10.0%
Number of observations
3413 1199 2214
Help from a folk healer
5.9%
6.5%
5.5%
Number of observations
3411 1200 2211
Prayed to be healed 60.5% 58.2% 59.7%
Number of observations
3410 1201 2209
Others prayed for my healing 48.6% 46.7% 49.6%
Number of observations
3393 1196 2197
total
100%
35.2% 64.8%
a Percentage of people in the respective categories (full sample, without insurance, with
insurance) who show the behavior indicated in the column “variable”.
Table 2 presents the descriptive statistics of the other explanatory variables used in the
analysis. Following the conceptual framework of Andersen and Newman (1973) three categories
of factors explaining the care utilization can be distinguished: the need for services, the available
resources (enabling factors), and the social conditions (predisposing variables). The primary
determinant of the need for services is expected to be one’s health. The first health indicator
included is the self-assessed health status, a variable obtained from the question, ‘In general, how

8
would you describe your own health?’, with five possible responses (excellent, very good, good,
fair, poor). In addition, we include more objective indicators, in particular if someone was
diagnosed (by a doctor or care provider) to suffer from diabetes, hypertension, asthma, or a
depression.
3
These are chronic problems with recurrent expenses for which the conventional
medical services offer a treatment but no cure; therefore we can expect that people seek other
ways to improve their life satisfaction, such as praying or alternative or traditional medicine.
Likewise, we include a binary variable that indicates if the respondent had to stay in bed for
more than half a day or was absent from work due to illness or injury during the previous 12
months.
The incidence of a health problem is not sufficient to generate usage of medical services;
required is their availability and accessibility. Enabling factors in the model, in addition to the
health insurance indicator, are the income (included as categorical variables: below 30,000 USD
(reference category), between 30,000 and 50,000 USD, more than 50,000 USD, and a category
that indicates that the income has not been reported). Information about the distance to a clinic,
probably a relevant measure, is unavailable. In addition, we include indicators of whether the
respondent is employed by someone else or whether he or she is self-employed (reference
category: not working), a sign for the potential availability of an employer-based health
insurance or other financial means to obtain insurance. Also the legal status and the length of
stay in the US can be expected to be relevant for the access to health care services, given that for
undocumented immigrants or people with a short time in the US it is more difficult to visit a
doctor (Ku and Matani, 2001).
Among the predisposing factors we include whether someone is living in a rural or
suburban area, with urban areas as the excluded (reference) category. Table 2 shows that the
Latinos are concentrated in urban (47.3%) and suburban (43.4%) zones, and that only a minority
of 9.3% is found in the rural zones. The specific ethnicity of the respondent, that it, the country
from which the respondent or their ancestors migrated, may be an important measure of cultural
factors that guide the decision to use traditional medicine or religious rituals. A majority of the
3
Questions about hypertension and asthma are only asked if preceding questions are responded negatively. Hence,
the variable “hypertension” indicates hypertension given no diabetes (self nor among family/friends), and “asthma”
indicates asthma given no diabetes (self/others) and no hypertension. In the Codebook a variable “chronic illness” is
proposed, defined as the sum of diabetes, hypertension, and asthma. Although we obtain essentially identical results
with that combined indicator, we decided to include the three variables individually, thereby accounting better for
differences in effects of various chronic diseases, although we should be careful not to interpret them separately.

9
respondents, 64.3%, report Mexican roots.
4
Cultural factors could also be reflected in the
respondents’ language proficiency. The majority report that Spanish is their dominant language
(48.8%, our default category, while for 16.5% is the primary language; the remaining 34.6%
consider themselves bilingual. Language proficiency, in particular a lack of English knowledge,
may result in communication problems with care providers in the US. Other predisposing
variables that enter the analysis are gender, age, marital status, the number of children in the
household, and the level of education.
Table 2 Other explanatory variables
a
variable mean
st.dev.
variable mean
st.dev.
Need factors (health status)
marital/household status
single, divorced, widowed 0.353
0.478
self-assessed health
b
2.053
1.142
no. of children <18 in household
c
1.179
3.155
diabetes, high blood sugar level 0.154
0.361
degree of urbanization
hypertension, heart disease 0.065
0.246
urb.: urban {reference} 0.473
0.499
not asked about hypertension 0.595
0.491
urb.: suburban 0.434
0.496
asthma, emphysema, chr. bronchitis 0.019
0.136
urb.: rural 0.093
0.290
depression 0.170
0.376
country of origin
bed-days, job absentee 0.377
0.485
Mexican {reference} 0.666
0.472
Puerto Rican 0.085
0.280
Enabling factors
Cuban 0.039
0.194
income
Dominican 0.042
0.201
income below $30,000 {reference} 0.521
0.500
Salvadoran 0.041
0.198
income $30,000 - $50,000 0.172
0.378
other Central American 0.062
0.242
income $50,000 or more 0.161
0.367
other South American 0.064
0.245
no information about income 0.146
0.353
primary language
labor status
Spanish {reference} 0.500
0.500
work, employed/self-employed 0.632
0.482
English 0.154
0.361
migratory status
bBilingual 0.347
0.476
permanent resident or US citizen 0.819
0.385
education
US gov’t picture ID (or more) 0.925
0.264
no/primary education {reference}
0.273
0.445
no. years in (continental) US
c
23.669
17.219
high school incomplete 0.190
0.392
high school completed 0.253
0.435
Predisposing variables
business/technical/vocational 0.038
0.191
college incomplete 0.114
0.317
female 0.500
0.500
college graduated 0.103
0.303
age
c
42.92
15.33
some postgraduate schooling 0.030
0.171
a Binary (0/1) variables, unless indicated otherwise
b Five-point scale from 0 (poor) to 4 (excellent)
c Continuous variable
4
The question asks where the respondent and the ancestors stem from, but adds a phrase asking with which country
or region the respondent identifies him/herself. We eliminate those who respond US, Europe, or Other, because they
do not consider their Latino descent very important.

10
Empirical model
A complication in the construction of the empirical model is that the survey asks if folk healers
and prayers have ever been used, while insurance is reported at the moment of the survey. Hence
it is possible that people have changed insurance status after praying or visiting folk healers.
Moreover, the loss of the insurance (e.g. after the termination of a job) can be caused by the
health incidences that led to the usage of services. The same issues hold, to a lesser extent, for
the visits to formal care providers, for which the questions refer to the use during the previous 12
months. Therefore, insurance status is suspect to be an endogenous variable in the econometric
model. Furthermore, it is possible that insurance status depends on care usage: for those with a
high health care demand it may be more difficult to contract an insurance, because they will be
expensive clients for insurers. This bidirectional relation may not hold directly for the usage of
folk healers or prayers (which are not covered by insurances); nevertheless it becomes important
if these services are substitutes for regular (insurance-covered) health care. Besides, those with
strong religious feelings may be less interested in buying insurance, under the allegation that
‘God determines my life’.
5
In order to account for the potential endogeneity of insurance, we perform an
instrumental variable analysis. Due to the discrete nature of both the outcomes (use of services,
S
i
) and the endogenous explanatory (health insurance, HI
i
), the model boils down to a bivariate
probit model (Wooldridge, 2010):
S
i*
= λ HI
i
+ β
S
’x
i
+ u
i
, (1)
HI
i*
= β
I
’x
i
+ γ’z
i
+ v
i
, (2)
where S
i*
and HI
i*
are unobserved latent variables. Observed is whether people visit a regular
doctor, go to a folk healer, or pray – each variable is analyzed in a separate model – S
i
, which equals
1 if S
i*
>0, and zero otherwise, and the endogenous health insurance, HI
i
=1 if HI
i*
>0 and zero
otherwise. The two equations are jointly estimated, together with the correlation (ρ) between the
error terms u
i
and v
i
. The vector of exogenous variables is x
i
, while the vector z
i
contains the
instrumental variables that correct for the endogeneity of health insurance. The parameter λ
measures the (causal) effect of health insurance on the use of prayers and folk healers, in the models
5
Using feelings about medical information and quality as independent variables, as in Reyes-Ortiz et al. (2009),
potentially suffers the same endogeneity issue and hampers the interpretation as a causal relation.

11
for the respective dependent variables. If tests demonstrate that endogeneity is not a problem, it
suffices to estimate only the first equation.
Our proposal is to use the individual’s citizenship status as an instrument, along with the
usage of internet in the household, and an indicator of the average insurance rate in the region
per ethnic group (see Appendix, Table A1). Legal status, and in particular whether someone is a
US citizen, should qualify as a valid instrument because there is no reason to expect that US
citizenship has a direct effect on the variables of interest (the use of the different types of healing
services), while it has a direct effect on the (potentially endogenous) insurance coverage, because
the US nationality implies easier access to job-based health insurance or otherwise earn enough
to buy insurance as well as access to government-sponsored health insurance programs such as
Medicare and Medicaid. Internet access tells something about the connectivity of the household
with the information needed to obtain a health insurance, while it is unlikely to have a direct
effect on the use of the various services. Furthermore, we use the regional gender-specific
probability of health insurance, an aggregate measure that is less affected by each individual
observation, while we can expect that the aggregate level tells us something about the likelihood
that an individual has insurance. The measure is based on the number of employed, retired, and
disabled people, because these statuses have a higher probability of access to health insurance –
either job-based or through Medicare/Medicaid. In order to minimize the impact of individual
observations, we calculate the average that applies for each individual respondent excluding the
specific respondent itself (but only the others in the same region and gender).
Results
Endogeneity of health insurance
Before we discuss the effect of insurance on the use of healing services, we address the quality of
the instruments and the relevance of the endogeneity correction. Testing the instruments in a
bivariate probit set-up is not straightforward. Preliminary tests in a linear probability model show
that the instruments contribute to the identification of the availability of health insurance and that
the instruments are valid (see Appendix, Table A2): they do explain health insurance (tests 1 &
2, rejection of H
0
) and they can be excluded from the main equation (test 3, H
0
not rejected).
However, the instruments may be considered somewhat weak (test 4, H
0
rejected at 10%). Test 5

12
indicates that there is no problem with endogeneity: the null hypothesis of exogenous insurance
access is not rejected. This implies that there is no need to correct for endogeneity, and that the
single probit estimation (eq. 1) should be preferred over the bivariate probit model (eq. 1 and 2).
6
Additionally, in the bivariate probit models, the correlation (ρ) between the insurance status and
the respective dependent variables is insignificant, another result that suggests that the decisions
are independent and that a single probit is preferable.
7
Health care usage
Table 3 presents the results of the estimations, explaining the effect of health insurance on the
probability that the respondent visited a doctor or health care provider (column 1), the
probability of a visit to a psychiatrist (column 2), whether the respondent visited folk healers
such as a curandero or a shaman as a treatment method (column 3), whether the respondent
prayed to get healed (column 4), and whether others used to pray for the healing of the
respondent (column 5).
We find that health insurance coverage strongly increases the use of regular services
offered by doctors and other care providers, and that also visits to psychiatrists are more likely
among people with health insurance, controlling for a broad set of needs indicators, enabling and
predisposing factors. In addition, we see that insurance reduces the use of folk healers,
suggesting some substitution away from the traditional services toward conventional medical
care. At the same time we find that insurance has a small but (marginally) significant positive
effect on the probability that others prayed for the respondent’s health. The results suggest that
(at some level) conventional care and praying by others are complementary; maybe this occurs
because more doctor visits imply more detection of problems and more active treatments.
However, we do not find any effect of insurance on one’s own praying activities.
6
The variables legal permanent residency in the US and whether one has an official identification document issued
by any US government agency, do not pass the overidentification test and are therefore included in the main
equation.
7
With the bivariate probit model we find essentially the same effects of insurance on care use as in the single
probits, although less precisely estimate and generally insignificant due to the loss of efficiency when instrumental
variables are applied.

13
Table 3 Probit models for use of conventional health care and alternative services
a
[1]
Visited doctor/other
health care provider
[2]
Visited
psychiatrist
[3]
Help from a
folk healer
[4]
Prayed to be healed
[5]
Others prayed for
my healing
with health insurance 0.536
***
(0.060) 0.248
*** (0.081) -0.151
* (0.083) 0.035
(0.051) 0.084
* (0.051)
Need factors (health status)
self-assessed health -0.148
***
(0.027) -0.068
** (0.033) -0.031
(0.037) -0.016
(0.022) -0.056
** (0.022)
diabetes, high blood sugar 0.492
***
(0.101) 0.150
(0.095) 0.004
(0.105) 0.060
(0.071) 0.075
(0.069)
hypertension, heart disease 0.417
***
(0.145) 0.011
(0.141) -0.034
(0.166) 0.139
(0.100) 0.158
(0.099)
not asked about hypertension -0.006
(0.060) -0.108
(0.084) 0.159
* (0.086) 0.162
*** (0.053) 0.166
*** (0.053)
asthma, emphys., chr.bronch. 0.302
(0.216) 0.358
(0.236) -0.017
(0.287) 0.078
(0.161) -0.094
(0.167)
depression 0.264
***
(0.090) 0.933
*** (0.074) 0.272
*** (0.090) 0.161
** (0.064) 0.218
*** (0.063)
bed-days, job absentee 0.349
***
(0.060) 0.092
(0.070) 0.139
* (0.075) 0.270
*** (0.048) 0.204
*** (0.048)
Enabling factors
income $30,000 - $50,000 -0.032
(0.076) -0.029
(0.100) -0.099
(0.107) -0.071
(0.065) -0.215
*** (0.065)
income $50,000 or more 0.162
(0.103) 0.063
(0.114) -0.027
(0.125) -0.110
(0.079) -0.395
*** (0.079)
no information about income 0.046
(0.084) -0.131
(0.100) 0.199
** (0.098) 0.124
* (0.068) 0.019
(0.067)
work, employed/self-empl. -0.167
** (0.069) -0.377
*** (0.079) 0.055
(0.086) -0.113
** (0.055) -0.040
(0.054)
permanent resident or US
citizen 0.252
***
(0.086) -0.002
(0.125) 0.041
(0.131) -0.081
(0.078) -0.176
** (0.077)
US gov’t picture ID (or more) -0.007
(0.114) -0.143
(0.166) 0.154
(0.177) 0.112
(0.105) 0.209
** (0.105)
no. years in (continental) US 0.005
(0.003) 0.002
(0.003) 0.002
(0.003) 0.001
(0.002) 0.004
** (0.002)
Predisposing variables
female 0.559
***
(0.058) -0.000
(0.071) -0.227
*** (0.075) 0.171
*** (0.048) 0.169
*** (0.047)
age (x10) -0.242
** (0.119) 0.116
(0.132) -0.038
(0.131) 0.163
* (0.084) 0.267
*** (0.084)
age squared (x100) 0.027
** (0.013) -0.026
* (0.014) -0.003
(0.014) -0.012
(0.009) -0.025
*** (0.009)
single, divorced, widowed -0.050
(0.062) 0.245
*** (0.073) 0.006
(0.080) -0.070
(0.050) -0.024
(0.050)
no. of children <18 in househ. 0.013
(0.015) 0.010
(0.007) -0.011
(0.018) -0.024
*** (0.008) -0.013
** (0.006)
urb.: suburban 0.020
(0.057) 0.075
(0.071) -0.054
(0.075) 0.083
* (0.047) 0.017
(0.047)
urb.: rural -0.092
(0.096) 0.030
(0.115) 0.041
(0.124) 0.145
* (0.081) 0.080
(0.080)
Ethnic group
Puerto Rican 0.131
(0.117) 0.230
** (0.110) -0.155
(0.139) 0.083
(0.085) 0.265
*** (0.084)
Cuban 0.194
(0.164) 0.196
(0.162) 0.587
*** (0.155) 0.117
(0.121) 0.159
(0.122)
Dominican 0.363
** (0.161) 0.223
(0.167) 0.136
(0.175) -0.001
(0.112) 0.245
** (0.113)
Salvadoran 0.039
(0.130) 0.276
* (0.164) -0.426
* (0.246) 0.309
*** (0.117) 0.272
** (0.113)
other Central American 0.168
(0.114) 0.076
(0.146) -0.201
(0.168) 0.208
** (0.095) 0.348
*** (0.093)
other South American 0.069
(0.116) 0.005
(0.149) 0.193
(0.140) 0.127
(0.096) 0.171
* (0.096)
primary language: English -0.085
(0.105) 0.124
(0.123) 0.305
** (0.131) -0.150
* (0.084) -0.139
* (0.084)
primary language: bilingual -0.060
(0.069) 0.045
(0.087) 0.247
*** (0.089) -0.080
(0.057) -0.057
(0.056)
high school incomplete 0.159
** (0.081) -0.099
(0.102) -0.241
** (0.110) -0.148
** (0.069) -0.004
(0.068)
high school completed 0.085
(0.078) -0.040
(0.100) -0.260
** (0.104) 0.074
(0.067) 0.098
(0.066)
business/technical/vocational -0.181
(0.144) -0.647
** (0.253) -0.292
(0.210) -0.040
(0.123) -0.105
(0.123)
college incomplete 0.208
* (0.109) -0.058
(0.125) -0.268
** (0.132) 0.001
(0.089) -0.073
(0.088)
college graduated 0.293
** (0.122) -0.182
(0.142) -0.060
(0.135) -0.093
(0.094) -0.106
(0.094)
some postgraduate schooling 0.017
(0.184) 0.265
(0.191) -0.222
(0.214) -0.294
** (0.142) -0.217
(0.146)
constant 0.555
** (0.276) -1.486
*** (0.344) -1.545
*** (0.352) -0.452
** (0.217) -0.989
*** (0.216)
pseudo-R
2
0.172
0.157
0.058
0.044
0.059
number of observations 3408
3413
3411
3410
3393
Wald test of constant-only
model 461.4
279.8
99.2
197.2
263.5
p-value Wald χ
2
test (d.f.=37) 0.000
0.000
0.000
0.000
0.000
LogLikelihood -1406.0
-850.4
-716.9
-2199.0
-2212.3
a We present the parameter estimates from single probit models for each dependent variable, because the tests (Table A2)
indicate that health insurance does not suffer from endogeneity. Robust standard errors. Significance levels: ***, **, *:
significant at 1%, 5%, 10%.

14
Other individual characteristics have the expected effects on the utilization of services. A
more favorable (subjective) evaluation of the own health has a strong negative effect on doctor
visits in general, and to a lesser extent on visits to psychiatrists. Also praying by others is
significantly smaller among people who evaluate their health as good. The other services are less
affected by the general health status. The objective health indicators show that being diagnosed
with a (chronic) disease increases the use of health care services. Mental health problems
(depression) increase the use of all services, while other health problems have more specific
effects on only one or a few services.
Women are more inclined to visit regular health care providers, other things equal,
however, no difference is found with regard to psychiatric visits. Also healing through praying is
sought more frequently by women, but they are less likely to visit folk healers. Visits to regular
health care providers start to increase after the age of 44, while the use of other’s praying for
healing is increasing with age until about 54 years and then starts to decrease. The probability of
psychiatric visits is decreasing with age over the complete relevant age range. The degree of
urbanization and the country of origin are of relatively minor importance for most dependent
variables. Cubans are more likely to use folk healers, probably for cultural reasons (Brandon,
1991), while Dominicans make more doctor visits. The legal status, in particular having at least
the permanent residency status, has a positive effect on the number of visits to regular health care
providers in general. College education appears to increase doctor visits, while folk healer visits
are more likely among those with only primary education or less. Knowledge about the services
and interest in taking care of one’s health may explain these findings. High income strongly
reduces praying for better health.
Discussion
We have analyzed the effect of health insurance coverage on the use of conventional and
traditional healing services in a representative sample of Latinos, and found that the availability
of health insurance coverage substantially increases the use of regular services, of doctors in
general, and also of the specific services provided by psychiatrists. These findings are in line
with the literature regarding health insurance effects on health care usage in general as well as
with earlier findings in various groups of Latinos in the US. In addition, we find that insurance

15
reduces the use of folk healers. The strongly significant positive effect of health insurance on the
use of conventional methods, and the negative effect on the use of alternative therapists imply
that the price reduction for conventional services inherited from health insurance coverage
results in a substitution away from the use of traditional services. The latter’s price is unaltered
by the health insurance access, since those methods are usually not covered by a health insurance
plan. Moreover we find a (small) increase of the praying by others for one’s health, suggesting
that religious faith can be considered as complementary to the use of medical services; perhaps
because more doctor visits imply the detection of more problems and hence more to worry about.
At the same time we see no effect at all on the own praying.
A shift toward conventional methods implies a large increase in the health care costs at
the community level. However, the increased costs that will arise when health insurance
coverage is expanded, for example by authorizing immigrants’ admittance to Medicaid, are
likely to have positive effects on immigrant’s health and living conditions. Poor health reduces
the capacity to work and has substantive negative effects on wages and labor force participation,
while access to health insurance and health care services have important effects on both labor
force participation and job choice (Currie and Madrian, 1999). Despite the potential beneficial
health impacts of alternative traditional healing methods and religious faith, these services
cannot replace conventional medicine and can at best be of supplementary relief. Moreover, our
results indicate that their use will not be eradicated when Latinos obtain better access to health
insurance and henceforth to regular medical services. In addition, not everyone with a right to
enroll Medicaid or Medicare does so, for example due to a lack of confidence in the authorities
in general or the services in particular, conditions that could also be related to cultural factors or
language problems. The apparent confidence in religious faith, shown through its
complementarity with medical services, can be used to provide counselling and stimulate the use
of regular health services.
Furthermore, many Latino immigrants have children born in the US, who by implication
obtain the US citizenship. Although through their citizenship they may be entitled to Medicaid, it
is difficult for them to exercise their rights if undocumented parents are reluctant to visit the
regular care services with their children for fear of deportation. Those children will have a ‘bad
start’, while it has been shown that initial childhood conditions have long-term effects (Currie,
2009). Berk et al. (2000) show that excluding undocumented immigrants from receiving

16
government-funded health care services will not reduce the immigration from Latinos to the US,
but that it is likely it will affect the well-being of their (US-citizen) children. A good health
protection system is therefore not only in the immigrants’ interest, but can create circumstances
with positive payoffs that reach beyond the directly affected people.
References
Alegria, M., G. Canino, P.E. Shrout, M. Woo, N. Duan, D. Vila, M. Torres, C. Chen C, and X.L.
Meng (2008). “Prevalence of mental illness in immigrant and non-immigrant U.S. Latino
groups”. American Journal of Psychiatry, Vol. 165 No. 3, pp. 359-369.
Andersen, R., and J.F. Newman (1973). “Societal and Individual Determinants of Medical Care
Utilization in the United States.” The Milbank Memorial Fund Quarterly. Health and Society,
Vol. 51, pp. 95-124.
Antecol, H., and K. Bedard (2006). “Unhealthy Assimilation: Why Do Immigrants Converge to
American Health Status Levels?” Demography, Vol. 43 No. 2, pp. 337-360.
Berk, M.L., C.L. Schur, L.R. Chavez, and M. Frankel (2000). “Health care use among undocumented
Latino immigrants.” Health Affairs, Vol. 19, pp. 51-64.
Brandon, G. (1991). “The Uses of Plants in Healing in an Afro-Cuban Religion, Santeria.” Journal of
Black Studies, Vol. 22 No. 1, pp. 55-76.
Brown, E.R., V.D. Ojeda, R. Wyn, and R. Levan (2000). Racial and Ethnic Disparities in Access to
Health Insurance and Health Care. UCLA Center for Health Policy Research/Henry J.
Kaiser Family Foundation, Los Angeles, CA.
Caicedo, M. (2010). Migración, Trabajo y Desigualdad. Los Inmigrantes latinoamericanos y
caribeños en Estados Unidos. El Colegio de México, Mexico City.
Carrasquillo, O., A.I. Carrasquillo, and S. Shea (2000). “Health insurance coverage of immigrants
living in the United States: differences by citizenship status and country of origin.” American
Journal of Public Health, Vol. 90 No. 6, pp. 917-923.
Choi, S. (2006). “Insurance Status and Health Service Utilization Among Newly-Arrived Older
Immigrants.” Journal of Immigrant and Minority Health, Vol. 8 No. 2, pp. 149-161.
Currie, J. (2009). “Healthy, Wealthy, and Wise: Socioeconomic Status, Poor Health in Childhood, and
Human Capital Development.” Journal of Economic Literature, Vol. 47 No. 1, pp. 87-122.
Currie, J., and B.C. Madrian (1999). “Health, health insurance and the labor market.” In: Ashenfelter O.,
Card, D.E. (eds.), Handbook of Labor Economics, vol. 3, part 3, chapter 50, pp. 3309-3416.
Favazza Titus, S.K. (2014). “Seeking and Utilizing a Curandero in the United States: A
Literature Review.” Journal of Holistic Nursing,
Vol.
32,
pp.
189-201.
Finkelstein, A., S. Taubman, B. Wright, M. Bernstein, J. Gruber, J.P. Newhouse, H. Allen, and K.
Baicker (2012). “The Oregon Health Insurance Experiment: Evidence from the First Year.”
Quarterly Journal of Economics, Vol. 127 No. 3, pp. 1057-1106.
González-Block, M.A, and L.A. de la Sierra-de la Vega (2011). “Hospital utilization by Mexican
migrants returning to Mexico due to health needs.” BMC Public Health, Vol. 11, Art. no. 241
(http://www.biomedcentral.com/1471-2458/11/241)
Iniguez, E., and L.A. Palinkas (2003). “Varieties of health services utilization by underserved
Mexican American women.” Journal of Health Care for the Poor and Underserved, Vol. 14
No. 1, pp. 52-69.

17
Ku, L., and S. Matani (2001). “Left-Out: Immigrants’ Access To Health Care And Insurance.”
Health Affairs, Vol. 20 No. 1, pp. 247-256.
Ku, L., and T. Waidmann (2003). How Race/Ethnicity, Immigration Status and Language Affect
Health Insurance Coverage, Access to Care and Quality of Care Among the Low-Income
Population. Commission on Medicaid and the Uninsured. Kaiser Family Foundation,
Washington, DC.
Lara, M., C. Gamboa, M.I. Kahramanian, L.S. Morales, and D.E. Hayes Bautista (2005). “Acculturation
and Latino health in the United States: A review of the literature and its sociopolitical context.”
Annual Review of Public Health, Vol. 26, pp. 367–397.
Lopez, R.A. (2005). “Use of Alternative Folk Medicine by Mexican American Women.” Journal of
Immigrant Health, Vol. 7 No. 1, pp. 23-31.
Lujan, J., and H.B. Campbell (2006). “The Role of Religion on the Health Practices of Mexican
Americans.” Journal of Religion and Health, Vol. 45 No. 2, pp. 183-195.
Markides, K.S., and J. Coreil (1986). “The health of Hispanics in the Southwestern: an epidemiologic
paradox.” Public Health Reports, Vol. 101 No. 3, pp. 253-265.
Massey, D.S. (2005). “Five Myths about immigration: Common misconceptions underlying U.S.
border- enforcement policy.” Immigration Policy in Focus, Vol. 4 No. 6, pp. 1-11.
Mohanty, S.A., S. Woolhandler, D.U. Himmelstein, S. Pati, O. Carrasquillo, and D.H. Bor (2005).
“Health Care Expenditures of Immigrants in the United States: A Nationally Representative
Analysis.” American Journal of Public Health, Vol. 95, pp. 1431-1438.
Nahin, R.L, J.M. Dahlhamer, and B.J. Stussman (2010). “Health need and the use of alternative
medicine among adults who do not use conventional medicine.” BMC Health Services
Research. Vol. 10, Art. no. 220 (http://www.biomedcentral.com/1472-6963/10/220).
Nandi, A., S. Galea, G. Lopez, V. Nandi, S. Strongarone, and D.C. Ompad (2008). “Access to and use of
health services among undocumented Mexican immigrants in a US urban area.” American
Journal of Public Health, Vol. 98 No. 11, pp. 2011-2020.
Newhouse, J.P., and the Insurance Experiment Group (1993). Free for All? Lessons from the RAND
Health Insurance Experiment. Harvard University Press, Cambridge, MA.
Ortega, A.N., H. Fang, V.H. Perez, J.A. Rizzo, O. Carter-Pokras, S.P. Wallace, and L. Gelberg (2007).
“Health Care Access, Use of Services, and Experiences Among Undocumented Mexicans and
Other Latinos.” Archives of Internal Medicine, Vol. 167 No. 21, pp. 2354-2360.
Pagán, J.A., A. Puig, and B.J. Soldo (2007). “Health insurance coverage and the use of preventive
services by Mexican adults.” Health Economics, Vol. 16, pp. 1359-1369.
Passel, J.S. (2005). Unauthorized migrants: numbers and characteristics. Washington, DC: Pew
Hispanic Center (http://pewhispanic.org/files/reports/46.pdf, published online June 14, 2005).
PHC (2008). Pew Hispanic Center 2007 Hispanic Healthcare Survey. Methodology Report. Pew
Research Center, Washington DC.
(http://pewhispanic.org/datasets/signup.php?DatasetID=10, published online August 13,
2008)
Pitkin Derose, K., J.J. Escarce, and N. Lurie (2007). “Immigrants and Health Care: Sources of
Vulnerability.” Health Affairs, Vol. 26 No. 5, pp. 1258-1268.
Pitkin Derose K, B.W. Bahney, N. Lurie, and J.J. Escarce (2009). “Review: immigrants and health
care access, quality, and cost.” Medical Care Research and Review. Vol. 66 No. 4, pp. 355-
408.
Ransford, H.E., F.R. Carrillo, Y. Rivera (2010). “Health care-seeking among Latino immigrants:
blocked access, use of traditional medicine, and the role of religion.” Journal of Health Care for
the Poor and Underserved, Vol. 21 No. 3, pp. 862-878.

18
Reyes-Ortiz, C.A., M. Rodriguez, and K.S. Markides (2009). “The Role of Spirituality Healing with
Perceptions of the Medical Encounter among Latinos.” Journal of General Internal Medicine,
Vol. 24 Suppl. 3, pp. 542–547.
Seybold, K.S., and P.C. Hill (2001). “The Role of Religion and Spirituality in Mental and Physical
Health.”, Current Directions in Psychological Science, Vol. 10 No. 1, pp. 21-24.
Teruya, S.A., and S. Bazargan-Hejazi (2013). “The Immigrant and Hispanic Paradoxes. A Systematic
Review of Their Predictions and Effects.” Hispanic Journal of Behavioral Sciences, Vol. 35 No.
4, pp. 486–509.
Turra, C.M., and N. Goldman (2007). “Socioeconomic Differences in Mortality Among U.S. Adults:
Insights Into the Hispanic Paradox.” The Journals of Gerontology: Series B, Vol. 62 No. 3, pp.
S184–S192
van Gameren, E. (2010). “Health insurance and use of alternative medicine in Mexico,” Health Policy,
Vol. 98, pp. 50-57.
Vega, W.A., M.A. Rodriguez, and E. Gruskin (2009). Health Disparities in the Latino Population.
Epidemiologic Reviews, Vol. 31 No. 1, pp. 99-112.
Wagner, T.H., and S. Guendelman (2000). “Healthcare Utilization Among Hispanics: Findings From the
1994 Minority Health Survey.” American Journal of Managed Care, Vol. 6 No. 3, pp. 355-364.
Wallace, S.P., C. Mendez-Luck, and X. Castañeda (2009). “Heading south: Why mexican immigrants in
California seek health services in Mexico.” Medical Care, Vol. 47, pp. 662-669.
Wong, R., and J.J. Díaz (2007). “Health care utilization among older Mexicans: health and
socioeconomic inequalities.” Salud Pública de México, Vol. 49, pp. S505-S514.
Wooldridge, J.M. (2010). Econometrics Analysis of Cross Section and Panel Data (2
nd
ed.). MIT Press,
Cambridge, MA.
Zweifel, P., and W.G. Manning (2000). “Moral hazard and consumer incentives in health care.” In:
Culyer, A.J., J. Newhouse (eds.), Handbook of Health Economics. vol. 1, part 1, chapter 8, pp.
409-459.

19
Appendix
Table A1 Instrumental variables
variable mean
st.dev.
US citizenship
a
0.544
0.498
Internet/email used in household
a
0.546
0.498
Region- and gender-specific health
insurance probability 0.633
0.114
a Binary (0/1) variables
Table A2 Indicative tests of the validity of the instrumental variables
a
visited (any) doctor
visited psychiatrist
visited curandero
or shaman
1 Test of excluded instruments 9.850
9.525
9.458
H
0
: excluded instruments do not explain endogenous variable
F(3, 3355): p=0.000
F(3, 3360): p=0.000
F(3, 3358): p=0.000
2 Underidentification test (Kleibergen-Paap LM Statistic) 29.164
28.227
28.024
H
0
:model is underidentified, instruments are not good
χ
2
(3): p=0.000
χ
2
(3): p=0.000
χ
2
(3): p=0.000
3 Overidentification test (Hansen J statistic) 3.501
2.362
1.836
H
0
: exclusion restrictions of instruments are valid
χ
2
(2): p=0.174
χ
2
(2): p=0.307
χ
2
(2): p=0.399
4 Weak identification test (Cragg-Donald Wald F statistic) 10.151
9.825
9.749
H
0
:weakly identified system
b
b
b
5 Endogeneity test of endogenous regressors 0.018
0.324
0.329
H
0
:variable can be considered as exogenous
χ
2
(1): p=0.894
χ
2
(1): p=0.569
χ
2
(1): p=0.566
prayed to be healed
others prayed for my
healing
1 Test of excluded instruments 9.623
9.646
H
0
: excluded instruments do not explain endogenous variable
F(3, 3357): p=0.000
F(3, 3340): p=0.000
2 Underidentification test (Kleibergen-Paap LM Statistic) 28.501
28.573
H
0
:model is underidentified, instruments are not good
χ
2
(3): p=0.000
χ
2
(3): p=0.000
3 Overidentification test (Hansen J statistic) 0.197
3.439
H
0
: exclusion restrictions of instruments are valid
χ
2
(2): p=0.906
χ
2
(2): p=0.179
4 Weak identification test (Cragg-Donald Wald F statistic) 9.935
9.912
H
0
:weakly identified system
b
b
5 Endogeneity test of endogenous regressors 1.215
0.000
H
0
:variable can be considered as exogenous
χ
2
(1): p=0.270
χ
2
(1): p=0.991
a Tests performed in the linear probability instrumental variable model, where more tests are available than in the dichotomous
models for instrumental variables. The first stage tests (1, 2, 4) differ only because of differences in the available number of
observations for the respective dependent variables.
b Stock-Yogo critical value for 5% (10%) maximal IV relative bias: 13.91 (9.08)