Content uploaded by Josephine Baako-Amponsah
Author content
All content in this area was uploaded by Josephine Baako-Amponsah on Oct 05, 2022
Content may be subject to copyright.
Vol.:(0123456789)
Child Indicators Research
https://doi.org/10.1007/s12187-022-09980-y
1 3
Household National Health Insurance Subscription
andLearning Outcomes ofPoor Children inGhana
RaymondElikplimKonti1 · JosephineBaako‑Amponsah2· PrinceDanso1
Accepted: 22 September 2022
© The Author(s), under exclusive licence to Springer Nature B.V. 2022
Abstract
This paper investigates the effect of poor households’ subscription to National
Health Insurance Scheme on their children’s school performance. Resource-poor
households are often vulnerable to low investment in education. This is particu-
larly the case where expenditure on health can trade off educational spending and
compromise children’s academic performance. This paper argues that poor house-
holds’ subscription to National Health Insurance Scheme could reduce their health
expenditure and consequentially increase educational spending to improve their
children’s school performance. This proposition was investigated using data from
the seventh round of the Ghana Living Standards Survey. The Instrumental Variable
method was employed to address possible endogeneity problems between health
insurance subscription and children’s learning outcomes. Additionally, the propen-
sity score matching technique was used to validate the results. The results show that
poor households’ subscription to National Health Insurance Scheme improves their
children’s learning outcomes in Ghana. The results, therefore, imply that universal
health coverage among the poor could enhance human capital development in devel-
oping countries.
Keywords Universal health coverage· Ghana· Human capital development·
Learning outcomes· National health insurance scheme· I130
JEL Codes I130
* Raymond Elikplim Kofinti
rkofinti@ucc.edu.gh
Josephine Baako-Amponsah
baakoamponsah@gmail.com
Prince Danso
princedansoabeam@gmail.com
1 Department ofData Science andEconomic Policy, School ofEconomics, University ofCape
Coast, CapeCoast, Ghana
2 Department ofEconomics, University ofGhana, Accra, Ghana
R.E.Kofinti et al.
1 3
1 Introduction
Health shock is one of the prevailing income shocks and a significant reason why
households fall into poverty. Globally, close to a billion people spend at least 10
per cent of their household budget on health expenses, and for almost 100 mil-
lion, these are ample expenses to push them into extreme poverty (WHO, 2017).
Health shocks potentially affect households due to the income uncertainties it
brings. Geng etal. (2018) revealed that households in low- and middle-income
countries pay a large share of health expenditures as out of pocket (OOP) and
that these households rely on precautionary savings, adjustments in labour sup-
ply, borrowing and informal credit, and informal transfers in the form of gifts
and remittances to cope with such health expenditures. Also, the literature on the
economic effect of health shocks in developing countries shows mixed results.
One strand of literature (e.g. Atake, 2018; Wagstaff & Lindelow, 2014) found
that health shocks significantly reduce household consumption, while other stud-
ies (e.g. Islam & Maitra, 2012) found no such effects. The latter evidence con-
cluded that households might be able to sustain consumption by adopting coping
strategies in response to such shocks, at least in the short run. Thus, households
in developing countries facing health shocks substitute consumer and production
spending for health care in the short run and decrease the net flows of investment
in productive activities in the long run. Akazili etal. (2017b) revealed that 10.7
per cent of Ghanaian households spent more than 10 per cent of their non-food
consumption expenditure on OOP healthcare payments, and about 2.6 per cent
of households spent over 20 per cent of their total household income on OOP
healthcare. These payments contributed to a relative increase in poverty head-
count by 9.4 per cent and 3.8 per cent using the $1.25/day and $2.5/day poverty
thresholds, respectively. According to He and Zhou (2022) and Ouadika (2020),
this may lead households to poverty or make them even poorer. Thus, health
shocks substantially increase health expenditure, reduce household income and
consumption and subsequently alter household economic decisions for resource-
constrained households.
The effect of unpredictable health shocks may be particularly devastating for
the poor, especially in the absence of affordable insurance. Atake (2018) provided
evidence that in sub-Saharan Africa, lack of health insurance coverage increases
the incidence of welfare loss and that the degree of vulnerability to poverty
largely depends on whether the household has health insurance or experiences
a reduction in user fees of health care at the point of service. Health insurance is
primarily realised to lessen the adverse economic impact of health shock (Finkel-
stein etal., 2018; Okoroh etal., 2018; Palas etal., 2017). The absence of insur-
ance coverage among the poor often implies significant out-of-pocket expenditure
as the primary source of healthcare. This primarily translates into diverting lim-
ited resources meant for other household livelihoods to cater for health shock.
Poor households without insurance pay a substantial share of their income for
healthcare which pushes them further into poverty with catastrophic conse-
quences (Finkelstein etal., 2012; Wagstaff & Lindelow, 2014).
1 3
Household National Health Insurance Subscription andLearning…
With the dynamic dual interlink between health and poverty, most house-
holds are unable to escape the trap of ill health and poverty once encountered.
Vulnerable households without health insurance mostly rely on coping strat-
egies such as the sale of assets, borrowing, and dissaving, among others, to
mitigate the effect of such shocks and to smoothen household consumption.
In most cases, the burden of health shocks is transferred to the education of
children (Lim, 2020). Some households may decide to take their children out
of school or send them to work for additional income (Dillon, 2008; Fabre &
Pallage, 2015; Lim, 2020). Time spent in school and available for the child to
develop numeracy and literacy skills reduce if the child works to complement
household income. Irregularities in school attendance vis-a-vis unavailabil-
ity of required resources to facilitate child education due to limited household
resources may affect a child’s academic achievement and basic school enrol-
ment. Thus, diverting resources meant for education investment may affect a
child’s cognitive growth and readiness for school. Evidence reveals that chil-
dren from poor households do not have the required literacy and numeracy
skills and may struggle to meet the proficiency cut-off point for English and
Mathematics, given the obstruction in their education. This is partly because of
poor investment in child’s education- access to books, pencils, notebooks, etc.
(Lim, 2020; Alcaraz etal., 2017; Liu, 2016; Haile & Haile, 2012); child work
(Fabre & Pallage, 2015; Johnson & Reynolds, 2013; Lim, 2020) due to health
expenditure. Understanding the effect of health expenditure on education deci-
sions and the channels thereof could help design policies that protect education
investments among Ghana’s vulnerable households.
Health insurance coverage improves general health status as a result of access
to affordable healthcare and subsequently ensures better health status among chil-
dren (Alhassan etal., 2016; Amo-Adjei etal., 2016; Wang etal., 2017). Subse-
quently, good health status influences children’s physical and emotional develop-
ment and their capacity to reach their full potential. Children in poor health may
have poor learning outcomes; hence, easing their health problems with access to
quality healthcare could eliminate barriers that hinder their success in the class-
room. Suhrcke and de Paz Nieves (2011) revealed that improving children’s health
will improve their performance in the classroom and that healthier children mean
greater future economic returns from schooling. The benefits of health insurance
coverage extend beyond improvement in health status to include reducing out-
of-pocket and catastrophic health expenditures. A reduction in health expendi-
ture implies that households would have extra resources that may be invested in
human capital development. Given that households are the leading investors in
both education and health in developing countries, households are likely to ease
the negative effect of health shock on child education significantly with health
insurance subscriptions (Lim, 2020). Exemplifying this, access to health insur-
ance may lessen budget constraints on education and improve educational out-
comes by increasing the resources available for educational expenses. It eludes
the need to take children out of school or send them to work to complement
household income; and increases disposable educational expenditure to facilitate
R.E.Kofinti et al.
1 3
access to books, pencils, and notebooks, among other resources, to improve child
learning outcomes.
Although conceptual links can be drawn to explain the potential impact
of health insurance coverage on child learning outcomes, given the different
strands of literature, limited evidence exists on how a subscription to health
insurance affects child learning outcomes among poor households in Ghana.
Based on the identified gaps in the literature, we seek to answer two research
questions: (1) does health insurance coverage affect the learning outcomes of
poor children? and (2) do poor households who are subscribed to the National
Health Insurance Scheme (NHIS) spend less on health expenditure and con-
sequentially spend more on their children’s education? We investigate these
questions by estimating the effect of poor households’ NHIS subscription on
their children’s learning outcomes and tease out the potential channels from
the perspectives of health and education expenditures using data from the sev-
enth round of the Ghana Living Standards Survey (GLSS7). The gender dif-
ferences in the link between NHIS subscription and child learning outcomes
are explored by estimating subsampled models for male–female children. We
empirically examine whether participating in NHIS reduces health expendi-
ture and potentially translates available resources to non-health spending that
may be directed towards human capital development. Thus, we examine such
channels to show how NHIS could affect the learning outcomes of children liv-
ing in poor households in Ghana. The study contributes to the literature in two
main ways. Foremost, we proffer empirical evidence on the role of health insur-
ance coverage as a viable policy tool in achieving Targets 4.2, 4.4 and 4.6 of
SDG4, which collectively seek to promote inclusive and equitable quality edu-
cation and engender lifelong learning opportunities for all by 2030. Second,
we resolve one of the critical methodological gaps in the literature by address-
ing the endogeneity problem associated with health insurance subscription and
children’s learning outcomes. We do this using instrumental variable estima-
tion in which the proportion of neighbours who are subscribed to NHIS is used
as an instrument. In addition to the endogeneity corrected results, Propensity
Score Matching is exploited as a robustness check.
The remaining sections of the paper are arranged as follows. Section2 provides
an overview of the NHIS and educational system in Ghana, whiles Sect.3 discusses
the methodology which covers data and variable definitions, and empirical model
specifications. Section4 presents the descriptive statistics, results and discussions.
Finally, Sect.5 concludes and provides recommendations.
2 Overview oftheNational Health Insurance Scheme
andtheeducational system ofGhana
2.1 Overview ofNational Health Insurance Scheme inGhana
The drive towards universal health coverage (UHC) across the globe has wit-
nessed diverse interventions to ensure fair and equal access to health. Regional
1 3
Household National Health Insurance Subscription andLearning…
and country-driven efforts to advance UHC on policy agendas have occurred,
both building upon long-established UHC programmes such as those in Japan and
many countries in Latin America and stimulating newer commitments to UHC
implementation in countries such as India, Kenya, and South Africa (Lozano
et al., 2020). Most of these policies target vulnerable populations such as the
poor, women and children, involving prepayment. Prepayment in healthcare
financing is characterised by a tax-financed government scheme and health insur-
ance where the latter takes the form of social health insurance, private compul-
sory health insurance and voluntary health insurance. Many developing countries
opted to set up tax-financed government schemes which offer comprehensive cov-
erage to the whole population, raising revenue from a broad base of tax and non-
tax sources and containing costs through vertical integration (Barasa etal., 2021).
Countries such as Gabon, Ghana, Mali, Nigeria, Senegal, South Africa, Tanzania
and Uganda have taken steps towards universal coverage by adopting risk pooling
systems to provide financial protection and thus, established health insurance sys-
tems to reduce financial barriers to healthcare.
Ghana implemented a NHIS as a step towards achieving UHC. Prior to the
introduction of NHIS, there existed financial catastrophe, and Akazili et al.
(2017a) revealed that, on average, OOP for healthcare accounted for 2.7 per cent
of total household expenditure, with poverty headcount increasing by 1.6 per
cent. Also, an exemption package that was in place to cushion the effects of user
fees on vulnerable populations was poorly implemented due to a lack of clarity
and understanding of its operation (Akazili etal., 2017a). Evidence showed that
many who were eligible for exemption were not exempted from paying and were
made to pay for services.
In 2004, Ghana instigated the NHIS to ensure equitable and universal access
to health care by removing such financial barriers imposed by user fees, and
thus, substituting the cash and carry system which was restricting access to
timely healthcare, especially by the poor. The scheme is financed with 2.5
per cent deductions from the Social Security and National Insurance Trust
(SSNIT), a 2.5 per cent insurance levy as Valued Added Tax (VAT) on goods
and services, annual premiums paid by subscribers who are 18years and above
and voluntary contributions, donations, gifts, grants, investments, and mon-
etary allocations made to the Health Insurance Fund by parliament (Govern-
ment of Ghana, 2004; National Health Insurance Authority, 2010). According
to NHIS (2018), NHIS covers 95 per cent of health conditions in the country,
including inpatient and outpatient services, maternity care (antenatal care, nor-
mal or assisted deliveries), and access to surgical care, emergency care, and
obstetrics. To enrol, one must pay a registration fee and annual premium based
on income and ability to pay. Children under 18years of age, pensioners with
SSNIT, people aged 70years and above, pregnant women, the indigent, and
Livelihood Empowerment Against Poverty (LEAP) beneficiaries are exempted
from payment of the annual premiums.
Although the recent population and housing census reported a significant pro-
portion of the Ghanaian population (68%) is covered by the NHIS (Ghana Sta-
tistical Service, 2021), the goal of universal coverage is far from being achieved
R.E.Kofinti et al.
1 3
by 2030. According to the Ministry of Health (2018), Out Patient Department
attendance per capita continues to decline from 1.16 in 2013 to 0.98 in 2017 and
regional supervised delivery distributions range from 43.3% to 75.5%, with some
regions worse off.
2.2 Overview oftheGhanaianEducational System
This section briefly describes the educational system of Ghana as of the Sev-
enth round of the GLSS collected in the 2016/2017 survey periods. Ghana has
three layers of the educational system: basic, secondary and tertiary education
(Adu-Gyamfi etal., 2016; Aziabah, 2018). The basic level of education consists
of 6years of Primary School and three years of Junior Secondary/High School
education. Primary School education has six (6) classes ranging from Primary
1 to Primary 6. Primary 1 to 3 is sub-categorized as lower primary, whereas
primary 4 to 6 qualify as upper primary. The accepted entry age to Primary 1 is
six (6) years; by 12years, a child is expected to have completed Primary School
education. Currently, a two (2) year kindergarten education in Ghana precedes
Primary School entry to smoothen the transition from home to Primary School
education. The content of Primary School education focuses on subjects such
as English Language (Reading, Writing, Comprehension, Dictation), Ghanaian
Language (Reading, Writing, Comprehension, Dictation), Mathematics, Inte-
grated Science, Introduction to Information Communication Technology (ICT),
Religious and Moral Education, Citizenship Education and Creative Arts. The
Junior Secondary/High School education consists of three classes, grouped into
Junior Secondary/High School forms 1, 2 and 3. Students get to Form 1 by their
13th birthday and complete Junior Secondary/High School education when they
are 15years by taking the Basic Education Certificate Examination (BECE) at
the end of Form 3. However, due to late enrolment and repetition, some chil-
dren complete Basic School education by 17years. The subsequent stages of
the country’s educational system are the secondary and tertiary levels of educa-
tion corresponding to three (3) years of Senior Secondary/High School and four
(4) years of University Bachelor’s Degree. The tertiary level also entails three
years of post-secondary diploma programmes from Teachers/Nurses training
colleges and High National Diploma programmes from Technical Universities.
3 Methodology
3.1 Data andVariable Definitions
Our primary data source emanates from the seventh round of the GLSS collected by
the Ghana Statistical Service (GSS) between 2016/2017. The survey applied a two-
stage sampling procedure. In the first stage, enumeration areas were selected prem-
ised on the 2010 Population and Housing Census with probability proportional to size
(number of households) (GSS, 2019). Consequently, 15 households were selected at
1 3
Household National Health Insurance Subscription andLearning…
the second stage by employing a systematic sampling method within each selected EA
(GSS, 2019). Overall, the primary sampling units in the GLSS 7 were 1000 (43.8% in
rural areas and 56.2% in urban areas). And the number of households was 15,000 (8430
in rural areas and 6570 in urban areas). The survey response rate was 93.4%, resulting
in 14,009 households and 59,864 individuals. The survey has information on children
and their households’ socioeconomic conditions. Further, the GLSS 7 has information
on the reading, writing and mathematics proficiency of children 11 years and older.
Based on these questions, we measure five dependent variables of learning outcomes.
The first is the child’s ability to read English or French, measured as a binary outcome
variable denoting whether the child could correctly read a simple phrase in English or
French on a flashcard. The second binary outcome variable indicates the child’s abil-
ity to write a simple phrase in English or French. The third and fourth binary/dummy
variables measure whether the child can read and write a Ghanaian language. The
last binary outcome variable assigns one to those who could solve simple mathemat-
ics questions and zero if otherwise. Available studies have measured these variables
accordingly (Afoakwah & Koomson, 2021; Frempong etal., 2021). Furthermore, as a
robustness check, we used the principal component analysis and additive index to gen-
erate a composite and additive measure of academic performance from these five indi-
cators. The GSS identifies (consumption expenditure) poor households in the aggregate
poverty file of the GLSS 7. Poor households are defined as households with consump-
tion expenditure below an upper poverty line of GH¢ 1,760.8 per adult equivalent per
year (GSS, 2018). We focus on poor households with children aged 11 to 17years who
are in basic school (from Primary 5 to Junior High School form 3). Children within this
age cohort have completed lower primary education, hence should read/write simple
phrases in English/French and any of their respective Ghanaian Languages. Also, they
should be able to solve simple mathematics questions. After cleaning the data, we have
a final sample of 2,755 children aged 11 to 17years living in poor households. It is
worth mentioning that though the completion age of basic schooling is largely 15years
in Ghana, the likelihood of late enrolment and class repetition sometimes delay com-
pletion. Hence, it is logical to include children who are 17years old in our sample.
The leading independent variable of the study is household NHIS subscription sta-
tus. Households’ subscription to NHIS is measured as a dichotomous variable, with
one (1) denoting the case where all household members are subscribed to the scheme
and zero (0) if otherwise. This definition caters for the risk-sharing element of insur-
ance. Thus, if all household members are not insured and there is a shock of ill-health
to an uninsured member, the other members will pool resources together to offset the
medical cost, which could affect the income and vulnerability of the entire household.
Hence, this measure emphasises the complete subscription of all the individual mem-
bers of the household.
In terms of the remaining control variables, their choices are dictated by their iden-
tification as predictors of children’s learning outcomes in existing studies (Afoakwah
et al., 2020; Afoakwah & Koomson, 2021; Frempong et al., 2021; Koomson &
Afoakwah, 2022) and the potential channel analyses carried out from the perspectives
of correlates of health and education expenditures (Donkoh & Amikuzuno, 2011; Huy,
2012; Nghiem & Connelly, 2017; Yetim etal., 2021). Table1 presents the dependent
and independent variables and their measurements.
R.E.Kofinti et al.
1 3
Table 1 Measurement of variables
Variable Measurement
Dependent Variables
Written calculation Binary variable equals 1 if the child can solve sim-
ple mathematics questions and zero, otherwise
Read English/French Binary variable equals 1 if child can correctly
read a simple phrase in English or French and 0,
otherwise
Write English/French Binary variable equals 1 if child can correctly
write a simple phrase in English or French and 0,
otherwise
Read a Ghanaian language Binary variable equals 1 if a child can correctly read
a simple phrase in a Ghanaian Language and 0,
otherwise
Write a Ghanaian language Binary variable equals 1 if a child can correctly read
a simple phrase in a Ghanaian Language and 0,
otherwise
Log of education expenditure Continuous
Log of household expenditure on basic school Continuous
Log of health expenditure Continuous
Explanatory Variables
Child variables
Age of child Continuous
Female child Dummy variable equals 1 if child is female and 0,
otherwise
Son/daughter to household head Dummy variable equals 1 if child is son/daughter to
household head and 0, otherwise
Grandchild to household head Binary variable equals 1 if child is grandchild to
household head
Child suffered from illness Binary variable equals 1 if child suffered from a
recent illness
Travel to school on foot Binary variable equals 1 if child commutes to
school on foot
Household variables
NHIS Household Dummy variable equals 1 if all household members
are subscribed to the National Health Insurance
Scheme or 0 otherwise
Age of head Continuous
Female-headed home Binary variable equals 1 if household head is female
Account ownership by head Binary variable equals 1 if household owns a sav-
ings account
Rural household Binary variable equals 1 if household resides in
rural locality
Household head is married Binary variable equals 1 if household head is mar-
ried
Household head is Separated/Divorced/Wid-
owed
Binary variable equals 1 if household head is Sepa-
rated/Divorced/Widowed
Number of household members Continuous
1 3
Household National Health Insurance Subscription andLearning…
3.2 Empirical Model
To estimate the link between NHIS subscription and children’s learning outcomes, we
use a linear probability model (LPM) as our baseline model. Two reasons underscore
our choice of LPM as the main model. First is the ease of interpretation of its mar-
ginal effects as observable in other related studies (Afoakwah etal., 2020; Afoakwah
& Koomson, 2021; Frempong etal., 2021; Koomson & Afoakwah, 2022). Second, the
coefficients produced by LPM are directly comparable with those of the two-stage least
squares (TSLS) procedure which is used to resolve the endogeneity problem associated
with NHIS subscription. To corroborate our baseline model, it is pertinent to note that
we estimate a probit model, and the results are presented in Table8 in the Appendix.
Table 1 (continued)
Variable Measurement
Household operates a farm Binary variable equals 1 if household operates a
farm
Household income per capita Continuous
Non-food price index Continuous
Household connected to National Electricity
Grid
Binary variable equals 1 if household is connected
to National Electricity Grid
Presence of elderly (65 +) Binary variable equals 1 if household has an older
adult beyond 65years
Presence of disability Binary variable equals 1 if household has a member
with a disability
Household member hospitalised Binary variable equals 1 if household has a member
hospitalised
Household member received a scholarship/
bursary
Binary variable equals 1 if any household member
has received a scholarship/bursary
Ecological zone-Coastal Binary variable equals 1 if household resides in
Coastal zone
Ecological zone-Forest Binary variable equals 1 if household resides in
Forest zone
Ecological zone-Savannah Binary variable equals 1 if household resides in
Savannah zone
District variables
District Development Scores Continuous
Number of teachers in the district Continuous
English pass rate BECE (2016/2017) Continuous
Mathematics pass rate BECE (2016/2017) Continuous
Science pass rate BECE (2016/2017) Continuous
Number of schools with electricity infrastruc-
ture (2016/2017)
Continuous
Number of schools with toilet infrastructure
(2016/2017)
Continuous
R.E.Kofinti et al.
1 3
Hence, for our baseline model, we estimate the learning outcomes model using LPM
for basic school students in poor households. The formal notation of the model follows
the works of Afoakwah and Koomson (2021), Frempong etal. (2021) and Koomson
and Afoakwah (2022).
For every child, i, the vector,
Learning
, contains the five learning outcomes (
j
=
do simple calculations, read English or French, write English or French, read a Gha-
naian language and write a Ghanaian language). Each of these variables is estimated
separately with the same set of control variables.
NHIS
is the dummy variable show-
ing households that are subscribed to the health insurance scheme. The vector,
HH
, contains household level control variables: age and sex of the household head,
household size and education status of the household head. In addition, we control
for the child’s age, sex and relationship to the household head in the vector
CHILD
.
Additionally,
DISTRICT
is a vector of district information: the number of teach-
ers in basic schools, pass rates (English language, mathematics and integrated
science) in the nationwide Basic Education Certificate Examination, number of
schools with electricity and toilet infrastructure from the Ghana Education Manage-
ment Information Systems (GEMIS) Data and district development score from the
UNICEF District League Table (UNICEF, 2017). It is worth noting that the district
information, particularly infrastructure and pass rates, control for the supply-side
effect of quality education on learning outcomes.
3.3 Identification
The coefficient of
NHIS
in Eq. 1 may be biased by possible endogeneity
between learning outcomes and
NHIS
subscription. For example, the parents’
unobservable inherent cognitive abilities could dictate their children’s academic
abilities. At the same time, their cognitive abilities could also influence par-
ents’ evaluation of the NHIS and whether to subscribe or otherwise. The incon-
sistency of the coefficient of NHIS households in our baseline model, LPM,
will therefore depend on the correlation between NHIS subscription and the
error term, which includes parents’ unobservable inherent cognitive abilities.
According to Wooldridge (2015), the bias in the coefficient of the NHIS house-
holds is positive if the correlation between NHIS subscription and the error
term is positive, and the reverse is true for the negative correlation. Unfor-
tunately, we cannot determine the direction of the bias (a priori) because the
error term is unobserved. Furthermore, the self-selectivity problem may also be
inherent in the subscription to health insurance and could bias the estimated β
due to nonrandomness in our sample (Fiestas Navarrete etal., 2019; Hellinger
& Wong, 2000; Liang et al., 2004). We circumvent these problems with the
instrumental variable (IV) technique drawing on the approach employed by Lu
etal. (2012) to construct a measure of cluster NHIS insurance prevalence rate
as an IV to reflect a source of variation in insurance subscription, which is
(1)
Learning𝐢𝐣
=
𝛼
+
𝛽
lNHIS +CHILD
𝐢
�
𝜸+HH
�
𝐢
𝚿+DISTRICT
�
𝐢
𝚪+
𝜖i
1 3
Household National Health Insurance Subscription andLearning…
exogenous in order to produce consistent results. We termed the IV as the pro-
portion of neighbours who are subscribed to NHIS. We used the cluster variable
having a number of households in the GLSS7 ranging from eight (8) to fifteen
(15) to capture neighbours. We believe that this variable meets the relevance
and exclusion restriction conditions necessary for instrument validity (Cameron
& Trivedi, 2010; Greene, 2003; Stock & Yogo, 2002; Wooldridge, 2015). In
terms of relevance, we argue that the proportion of neighbours who are sub-
scribed to NHIS is positively associated with the likelihood that a household
subscribes to NHIS. Whether a household participates in NHIS partly depends
on information, which hinges on neighbours who have subscribed to the scheme
(Lu etal., 2012; Peng & Conley, 2016). Concerning exclusion restriction, we
do not expect the proportion of neighbours who are subscribed to NHIS to have
a direct influence on child learning unless it indirectly affects the child through
the household’s NHIS subscription. We then estimate the effect of NHIS on
learning outcomes in a two-stage instrumental variables model. In the first
stage, Eq.(2) states NHIS subscription as a function of the proportion of neigh-
bours’ who have subscribed to NHIS and the control variables in Eq.(1).
We estimate Eq.(2) with the ordinary least square (OLS) model and predict
NHIS
. In the second stage, we replace
NHIS
in Eq.(1) with its prediction from
Eq.(2) and estimate a linear probability model of learning outcomes in Eq.(3).
A causal interpretation of
𝜃
requires that
NHIS_NEIGHBOURS
has a positive
association with NHIS subscription and can only influence learning outcomes
through NHIS subscription.
In addition to the instrumental variable estimations, the study exploits the
propensity score matching (PSM) method to address potential endogeneity as
observed in other studies (Gertler etal., 2016; Khandker etal., 2010). The treat-
ment variable in this study is NHIS subscription and is used to estimate the aver-
age treatment effect on child learning outcomes. The technique produces an esti-
mate to measure the counterfactual impact of NHIS subscription on child learning
outcomes. We use five matching techniques to subject our findings to sensitivity
tests: nearest neighbour (1), nearest neighbour (5), radius, kernel and local linear
regression matching methods.
3.4 Potential Channels
The literature suggests that participating in national health insurance schemes
reduces health expenditure in general and out-of-pocket health expenditure in
particular (Finkelstein et al., 2012; Islam & Maitra, 2012; Palas et al., 2017;
Wagstaff & Lindelow, 2014). This potentially translates into available resources
(2)
NHISi
=
𝜑
+
𝛿
NHIS_NEIGHBOURS +CHILD
�
i
𝝈+HH
�
𝐢𝜃
+DISTRICT
�
𝐢
𝚿+v
i
(3)
Learning𝐢𝐣
=𝜂+𝜃
NHIS +CHILD�
𝐢
𝚫+HH�
𝐢
𝛟+DISTRICT�
𝐢
𝛚+𝜉
i
R.E.Kofinti et al.
1 3
for non-health spending in poor households. Non-health spending frees up
resources that may be directed towards other investments in basic physical assets
and human capital development. This consequently increases non-health expendi-
ture as that of education expenditure which could affect learning outcomes.
Therefore, we examine these channels to show how NHIS could potentially affect
the learning outcomes of children living in poor households in Ghana. Empiri-
cally, we employ the OLS to examine the effect of NHIS subscription on health
and education expenditures.
4 Results
4.1 Descriptive Statistics
Table2 summarises the main variables and tests their differences between NHIS
and non-NHIS households. Fifty-one per cent of children in poor households live
in homes where every member is subscribed to NHIS. On average, fewer females
and older children are found in NHIS households than in Non-NHIS house-
holds. We find that basic school children in NHIS households are better at read-
ing and writing English or French. Also, we find that these children performed
better in reading and writing Ghanaian languages. They also perform better on
the calculation questions. We further observed that households that subscribed to
NHIS spend more on their children’s basic education and overall education than
Non-NHIS households. On the contrary, NHIS households spend less on health
expenditure (Table2).
4.2 NHIS Subscription inGhana
We precede the regression analysis with the distribution of household NHIS sub-
scriptions and household expenditure on health and education. Figure1, based on
the last two rounds of the GLSS, shows household NHIS subscriptions from 2013
to 2017. In 2012/13, 50.62% of Ghanaian households subscribed to NHIS, with
a poor and non-poor household distribution of 38.50% and 52.99%, respectively.
Though NHIS households generally increased, the proportion of poor households
that have subscribed increased by less than 10 per cent to 46.81% during the
2016/17 round of the survey. Unlike the poor households, the two survey peri-
ods show consistently that more than half of the non-poor households have sub-
scribed to NHIS.
Figure2 explores the relationship between NHIS subscription and total house-
hold health and out-of-pocket expenditures, as shown in panels 2a and 2b, respec-
tively. Figure2 shows that Ghana’s poor households who subscribed to NHIS
consistently spent less on total health expenditure and out-of-pocket expenditure
over the survey periods.
1 3
Household National Health Insurance Subscription andLearning…
Table 2 Descriptive statistics
Dependent and independent variables (1) (2) (3) (4)
Non-NHIS household NHIS household Difference t < p
n = 1330 n = 1425
(48.28%) (51.72%)
Child variables
Written calculation 0.576 0.724 -0.148 0.000
Read English/French 0.496 0.636 -0.141 0.000
Write English/French 0.474 0.594 -0.120 0.000
Read a Ghanaian language 0.195 0.222 -0.026 0.088
Write a Ghanaian language 0.172 0.198 -0.026 0.081
Age of child 13.800 13.815 -0.015 0.847
Female child 0.441 0.444 -0.002 0.896
Son/daughter to household head 0.784 0.782 0.002 0.906
Grandchild to household head 0.092 0.098 -0.006 0.598
Child suffered from illness 0.077 0.069 0.008 0.423
Travel to school on foot 0.980 0.971 0.010 0.093
Household variables
Age of head 50.646 51.389 -0.743 0.163
Female-headed home 0.186 0.274 -0.087 0.000
Educated headed homes 0.365 0.404 -0.039 0.035
Account ownership by head 0.203 0.350 -0.147 0.000
Rural household 0.901 0.914 -0.014 0.219
Household head is married 0.802 0.744 0.059 0.000
Household head is Separated/Divorced/Widowed 0.131 0.213 -0.081 0.000
Number of household members 8.600 7.389 1.211 0.000
Household operates a farm 0.922 0.914 0.008 0.435
R.E.Kofinti et al.
1 3
Source: Computed from GLSS 7(2016/2017), GEMIS Data 2016/2017 and UNICEF 2017 District League Table
Table 2 (continued)
Dependent and independent variables (1) (2) (3) (4)
Non-NHIS household NHIS household Difference t < p
n = 1330 n = 1425
(48.28%) (51.72%)
Household income per capita 1225.435 1136.115 89.320 0.384
Non-food price index 0.982 0.973 0.009 0.000
Household connected to National Electricity Grid 0.439 0.485 -0.046 0.015
Log of education expenditure 5.348 5.770 -0.422 0.000
Log of health expenditure 1.965 1.556 0.409 0.000
Log of household expenditure on basic school 5.061 5.405 -0.344 0.000
Presence of elderly (65 +) 0.246 0.307 -0.061 0.000
Presence of disability 0.107 0.074 0.033 0.002
Household member hospitalised 0.243 0.242 0.001 0.972
Household member received a scholarship/bursary 0.005 0.002 0.002 0.269
Ecological zone-Coastal 0.157 0.066 0.091 0.000
Ecological zone-Forest 0.211 0.179 0.032 0.033
District variables
District Development Scores 64.801 65.797 -0.996 0.000
Number of teachers in the district 1199.965 1277.289 -77.324 0.035
English pass rate BECE (2016/2017) 59.084 57.494 1.590 0.020
Mathematics pass rate BECE (2016/2017) 66.462 63.451 3.011 0.000
Science pass rate BECE (2016/2017) 63.753 61.763 1.989 0.005
Number of schools with electricity infrastructure (2016/2017) 59.522 62.321 -2.799 0.245
Number of schools with toilet infrastructure (2016/2017) 92.117 94.485 -2.368 0.248
1 3
Household National Health Insurance Subscription andLearning…
Figure3, on the other hand, explores the relationship between NHIS subscrip-
tion and total household education expenditure and basic education expenditure,
as shown in panels 3a and 3b, respectively. Figure3a shows that poor households
in Ghana who are subscribed to NHIS consistently spent more on their total edu-
cational expenditure across the two surveys in 2012/2013 and 2016/2017, respec-
tively. In Fig.3b, although NHIS households spend less on basic schooling in
the 2012/2013 survey year compared to Non-NHIS households, the opposite is
the case in the 2016/17 survey periods, where NHIS households expend more on
basic schooling than their non-NHIS counterparts.
4.3 Model Results
This section presents the baseline and the endogeneity corrected results using
LPM and TSLS techniques, respectively.
4.3.1 Effect ofNHIS Subscriptionon Learning Outcomes inGhana (Baseline Results‑LPM)
Table 3 presents a surmised version of baseline results for the association
between NHIS subscription and children’s learning outcomes using the LPM.
The full results with estimates for the confounding variables, including probit
estimates, are presented in Table8 in the Appendix. In Column 1 of Table3, we
observe that NHIS subscription is associated with an improvement in children’s
written calculation by 12.3 per cent. In Columns 2 and 3, NHIS subscription is
associated with 13.4 and 12.4 per cent improvement in children’s ability to read
and write basic words and phrases in English/French, respectively.
The results of the control variables presented in Table8 of the Appendix are
largely intuitive and consistent with our a priori expectation. For instance, we
find that older children have more improved learning outcomes. Also, financially
included households from the perspective of saving account ownership, house-
holds connected to the National Electricity Grid and educated household heads
have better children’s learning outcomes. Children in districts having higher num-
ber ofschools with electricity and toilet facilities have better learning outcomes.
4.3.2 Effect ofNHISSubscription onLearning Outcomes inGhana (IV Results)
In this section, we report TSLS results because LPM produces inconsistent
estimates in the existence of endogeneity. Table 4 presents the endogeneity
corrected estimates of the effect of NHIS on the children’s mathematics and
language (English and French) proficiencies. Statistically, the F-statistic of the
TSLS models shows that the instrument strongly identifies our endogenous
regressor (Greene, 2003; Stock & Yogo, 2002), and the first stage results indi-
cate the validity of our instrument given by a positive relationship between
the proportion of neighbours who have subscribed to the scheme (instru-
ment) and the NHIS subscription [endogenous regressor] (Cameron & Trivedi,
2005; Wooldridge, 2015). In Columns 1 to 3, NHIS households have a positive
R.E.Kofinti et al.
1 3
Fig. 1 NHIS subscription in Ghana (2012–2017). [Source: Computed from GLSS 6 & GLSS 7]
Fig. 2 Average health expenditure by household type. [Source: Computed from GLSS 6 & GLSS 7]
1 3
Household National Health Insurance Subscription andLearning…
association with children undertaking written calculation, reading English/
French and writing English/French by 35.6, 36.1 and 30.6 percentage points,
respectively. A scrutiny of the TSLS results revealed larger estimates compared
to that of the LPM, indicating a downward bias in the LPM estimates. The val-
ues of the coefficients signify that NHIS Households have a positive associa-
tion with all three learning outcomes.
The results of the control variables presented in Table9 of the Appendix are
intuitive and largely corroborate our a priori expectations. Older children have
better learning outcomes (written calculation, reading English/French and writ-
ing English/French), households with educated heads, biological children of the
head of household and children in districts with higher number of schools with
electricity infrastructure have better learning outcomes.
4.3.3 Effect ofNHIS Subscription onLearning Outcomes – Girls andBoys (IV Results)
In this subsection, we report results for girls and boys to unravel any unob-
servable gender heterogeneities associated with NHIS subscription and chil-
dren’s learning outcomes using TSLS. In Table5, the estimates for girls are
presented in columns (1)–(3), while those for boys are in columns (4)–(6).
The coefficients show that NHIS households have a positive association with
boys’ performance mainly. This is not far-fetched given that in poor homes,
Fig. 3 Average education expenditure by household type. [Source: Computed from GLSS 6 & GLSS 7]
R.E.Kofinti et al.
1 3
girls’ education is usually compromised with less resource allocation than boys.
Hence, we may expect the NHIS households to have higher positive effects on
boys than girls. The control variables are largely consistent with the a priori
expectation and intuitive, as in the case of Tables3 (baseline results) and 4
(IV-results).
Table 3 Effect of NHISSubscription on Learning Outcomes in Ghana (Baseline LPM Results)
Robust standard errors in parentheses, *** p < 0.01, ** p < 0.05, * p < 0.1 LPM: Linear Probability
Model. See Table8 of the Appendix for estimates of the control variables
(1) (2) (3)
Written calculation Read English/French Write
English/
French
NHIS Household 0.123*** 0.134*** 0.124***
(0.026) (0.030) (0.029)
Child Characteristics Yes Yes Yes
Household Characteristics Yes Ye s Yes
District Characteristics Yes Ye s Yes
Observations 2755 2755 2755
R-squared 0.167 0.164 0.169
Table 4 Effect of NHISSubscription on Learning Outcomes in Ghana (IV Results)
Robust standard errors in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01. See Table 9 of the Appendix
for estimates of the control variables
(1) (2) (3)
Written calculation Read English/French Write
English/
French
NHIS Household 0.356*** 0.361*** 0.306***
(0.093) (0.090) (0.093)
Child Characteristics Yes Ye s Yes
Household Characteristics Yes Yes Yes
District Characteristics Yes Yes Yes
Kleibergen-Paap rk LM statistic 51.437 51.437 51.437
[0.000] [0.000] [0.000]
Kleibergen-Paap rk Wald F statistic 87.327 87.327 87.327
First Stage
Proportion of neighbours subscribed to NHIS 0.007*** 0.007*** 0.007***
(0.001) (0.001) (0.001)
Observations 2755 2755 2755
R-Squared 0.112 0.117 0.139
1 3
Household National Health Insurance Subscription andLearning…
4.4 Potential Channels
4.4.1 Effect ofNHIS Subscription onHealth andEducational Expenditures
Table6 provides evidence of potential channels through which NHIS subscrip-
tion could affect the learning outcomes of children living in poor Ghanaian
households. We use the OLS to examine the effect of NHIS subscription on
health, education and basic schooling expenditures, given the continuous nature
of the variables. The analyses in this section will empirically assess whether
poor householdswho are subscribed to the NHIS spend less on health expendi-
ture and consequentially spend more on their children’s education. To deal with
the outliers in health, education and basic schooling expenditures, we used the
logarithm of these variables. These variables are obtained from the aggregate
dataset of the GLSS 7 survey, hence the need to minimise the variability across
households. We hypothesise that NHIS subscription would reduce household
health expenditure and consequentially increase total household education and
basic schooling expenditures. Column 1 shows the effect of NHIS subscrip-
tion on health expenditure, whereas Columns 1 and 2 depict the association
between NHIS subscription and total educationexpenditure, and basic school-
ing expenditurerespectively. The results in Table6 indicate that households
who have subscribed to NHIS reduce health expenditure by 43.6 per cent, and
increase their education and basic schooling expenditures by 72.6 and 67.5 per-
centage points, respectively.
The full results for the control variables for Column 1 are presented in
Table10, whereas that of Columns 2 and 3 are shown in Table11 in the Appen-
dix. The results are generally intuitive for both tables. Regarding Appendix
Table 10, we found that households spend less health expenditure on older
children. In contrast, married household heads, larger household sizes, pres-
ence of disability and households where a member has been hospitalised in the
past 12months have higher health expenditure. Concerning Appendix Table11,
from the perspective of total education expenditure, households spend less
education expenditure on children in rural areas and those that travel to school
on foot. On the other hand, older household heads, educated household heads,
financially included (savings account ownership) households, and those
with larger membership and connected to the national electricity grid incur
highereducation expenditure.
4.5 Robustness/Sensitivity Checks
4.5.1 Effect ofNHIS Subscription onAcademic Performance (Composite Index
andAdditive Indices)
We test the robustness of the findings using different specifications of the dependent
variables to re-estimate the effect of NHIS on learning outcomes in Table7. Using the
R.E.Kofinti et al.
1 3
Table 5 Effect of NHIS Subscriptionon learning outcomes- girls and boys (IV Results)
Robust Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01
(1) (2) (3) (4) (5) (6)
Girls Boys
Written calculation Read English/French Write Eng-
lish/French
Written calculation Read English/French Write
English/
French
NHIS Household 0.257 0.132 0.096 0.234*** 0.309*** 0.244***
(0.188) (0.180) (0.174) (0.083) (0.079) (0.085)
Child Characteristics Yes Yes Yes Yes Ye s Yes
Household Characteristics Yes Ye s Yes Yes Ye s Yes
District Characteristics Yes Ye s Yes Yes Ye s Yes
Kleibergen-Paap rk LM statistic 51.430 51.430 51.430 21.775 21.775 21.775
[0.000] [0.000] [0.000] [0.000] [0.000] [0.000]
Kleibergen-Paap rk Wald F statistic 114.494 114.494 114.494 29.006 29.006 29.006
First Stage
Proportion of neighbours subscribed to NHIS 0.005*** 0.005*** 0.005*** 0.008*** 0.008*** 0.008***
(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
Observations 1219 1219 1219 1536 1536 1536
R-Squared 0.215 0.259 0.255 0.211 0.175 0.208
1 3
Household National Health Insurance Subscription andLearning…
principal component analysis, we generate a composite measure of the child’s academic
performance using the five learning outcome variables (written calculation, read English/
French, write English/French, read Ghanaian Language and write Ghanaian Language)
and re-estimate the models with this variable in Column 1. Column 2 presents the addi-
tive index of the five learning outcomes variables. Results from the different models in
Table 7 show that NHIS subscription positively affects the academic performance of
basic school children living in poor households in the country. The control variables are
consistent with the a priori expectation and intuitive, as in the case of Table4.
4.5.2 Additional Robustness Check using thePSM
The PSM results provide an additional robustness check regarding the relation-
ship between NHIS subscription and child learning outcomes. We exploit sev-
eral matching techniques—nearest neighbour, radius, kernel and local linear
regression matching methods—to ensure robustness in our findings consistent
with other studies (Churchill & Marisetty, 2020; Churchill etal., 2020; Gertler
etal., 2016; Kofinti etal., 2022; Koomson & Danquah, 2021). The results of
the PSM are presented in Table13 of the Appendix, and the estimates corrobo-
rate our TSLS results.
5 Conclusion
This paper investigated the effect of households’ subscription to NHIS on the
learning outcomes of children living in poor households. The study postulates
that expenditure on health constraints educational expenditure in poor house-
holds and consequently compromises their children’s academic performance.
However, by subscribing to NHIS, poor households are protected from unex-
pected health expenditure, which relaxes non-health expenditure to be invested
Table 6 Effect of NHIS subscription on education and health expenditures (OLS Results)
Standard errors in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01. See Appendix Table10 for control vari-
ables estimates for Column 1 (health expenditure) in the Appendix; See Appendix Table11 for control
variables estimates for Columns 2 and 3 (total household education and basic schoolingexpenditures) in
the Appendix
(1) (2) (3)
Log (health
expenditure)
Log (total household edu-
cation expenditure)
Log (household
expenditure on basic
school)
NHIS Household -0.436** 0.726*** 0.675***
(0.179) (0.110) (0.104)
Child Characteristics Yes Yes Yes
Household Characteristics Yes Ye s Ye s
Observations 2755 2755 2755
R-Squared 0.107 0.264 0.262
R.E.Kofinti et al.
1 3
in children’s educational outcomes. Unlike most developing countries, Gov-
ernment of Ghana has a health financing policy spanning over a decade to
increase universal health coverage, especially among the poor. Predicated on
this background, the country provides a fertile ground to examine the prospects
of households’ subscription to NHIS on the educational performance of chil-
dren living in consumption expenditure poor households in Ghana. Nationally
representative data was employed to address the research propositions. The
paper corrected the potential endogeneity between NHIS and learning out-
comes using the instrumental variable technique. We find that children perform
better in mathematics, reading and writing when all the household members
are subscribed to NHIS. Further analyses indicate that improved learning out-
comes could be attributed to a decrease in health expenditure and the conse-
quent increase in educational expenditure among poor households who have
subscribed to the NHIS. Overall, these results suggest that promoting health
insurance coverage among poor households could have human capital develop-
mental effects on their children besides formal health care utilisation. Based on
our findings, policymakers and governments should encourage universal health
coverage instruments among poor households in developing countries to engen-
der human capital development. A limitation of this study is that our analyses
do not cater for the data’s hierarchical structure at student/children, school and
district levels, given that the GLSS does not have school characteristics infor-
mation on household members. Hence, we recommend that prospective studies
with appropriate data should fill this gap.
Table 7 Effect of NHISSubscription on Academic Performance in Ghana
Robust Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01
(1) (2)
Academic Performance
(Composite)
Academic
Performance
(Additive)
NHIS Household 0.122*** 0.126***
(0.036) (0.036)
Child Characteristics Yes Yes
Household Characteristics Ye s Yes
District Characteristics Yes Ye s
Kleibergen-Paap rk LM statistic 50.378 50.378
[0.000] [0.000]
Kleibergen-Paap rk Wald F statistic 96.122 96.122
First Stage
Proportion of neighbours subscribed to NHIS 0.007*** 0.007***
(0.001) (0.001)
Observations 2755 2755
R-Squared 0.217 0.215
1 3
Household National Health Insurance Subscription andLearning…
Table 8 Effect of NHIS on learning outcomes in Ghana (full results: baseline)
Explanatory variables (1) (2) (3) (4) (5) (6)
Written calculation Read English/French Write English/French
Probit(ME) LPM Probit(ME) LPM Probit(ME) LPM
NHIS Household 0.121*** 0.123*** 0.132*** 0.134*** 0.121*** 0.124***
(0.020) (0.026) (0.023) (0.030) (0.023) (0.029)
Age of child 0.034*** 0.036*** 0.046*** 0.047*** 0.046*** 0.048***
(0.005) (0.005) (0.005) (0.006) (0.005) (0.006)
Female child -0.000 -0.005 0.022 0.017 0.007 0.002
(0.020) (0.022) (0.021) (0.023) (0.022) (0.022)
Son/daughter to
household head
0.052 0.055*0.095** 0.094** 0.086** 0.086**
(0.036) (0.033) (0.040) (0.039) (0.040) (0.041)
Grandchild to house-
hold head
0.012 0.030 0.107** 0.123** 0.070 0.084
(0.050) (0.054) (0.052) (0.061) (0.055) (0.062)
Child suffered from
illness
-0.019 -0.016 0.020 0.022 0.015 0.016
(0.035) (0.041) (0.038) (0.040) (0.038) (0.041)
Age of head 0.002*0.002 0.000 0.000 0.000 0.000
(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
Female-headed home 0.142*** 0.139*** 0.063 0.065 0.036 0.037
(0.034) (0.037) (0.041) (0.044) (0.041) (0.046)
Educated headed
homes
0.082*** 0.093*** 0.040 0.048 0.019 0.027
(0.028) (0.032) (0.029) (0.033) (0.029) (0.034)
Account ownership
by head
0.094*** 0.092*** 0.078*** 0.077** 0.074*** 0.071**
(0.023) (0.028) (0.026) (0.031) (0.026) (0.030)
Appendix
R.E.Kofinti et al.
1 3
Table 8 (continued)
Explanatory variables (1) (2) (3) (4) (5) (6)
Written calculation Read English/French Write English/French
Probit(ME) LPM Probit(ME) LPM Probit(ME) LPM
Rural household -0.036 -0.033 -0.022 -0.021 -0.060 -0.060
(0.041) (0.057) (0.043) (0.058) (0.043) (0.056)
Household head is
married
0.099** 0.089*0.047 0.043 0.043 0.037
(0.047) (0.053) (0.050) (0.054) (0.050) (0.054)
Household head is
Separated/Divorced/
Widowed House-
hold
-0.034 -0.028 -0.042 -0.044 -0.022 -0.025
(0.059) (0.061) (0.065) (0.069) (0.064) (0.072)
Number of household
members
-0.003 -0.005 -0.005*-0.007 -0.005 -0.006
(0.003) (0.004) (0.003) (0.004) (0.003) (0.004)
Household operates
a farm
-0.017 -0.009 -0.039 -0.038 -0.030 -0.029
(0.041) (0.040) (0.046) (0.051) (0.047) (0.052)
Household income per
capita
0.000 0.000 0.000 0.000 0.000 0.000
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
Non-food price index -0.324 -0.349 -0.368 -0.359 -0.565** -0.558
(0.250) (0.399) (0.273) (0.430) (0.276) (0.411)
Household connected
to National Electric-
ity Grid
0.035*0.049 0.058** 0.065** 0.055** 0.064**
(0.021) (0.032) (0.023) (0.032) (0.023) (0.032)
Coastal Zone 0.084** 0.101*0.070*0.092 0.116*** 0.148**
(0.039) (0.058) (0.041) (0.061) (0.041) (0.062)
Forest Zone 0.107*** 0.112** 0.155*** 0.156*** 0.213*** 0.219***
(0.036) (0.053) (0.038) (0.056) (0.037) (0.054)
1 3
Household National Health Insurance Subscription andLearning…
Table 8 (continued)
Explanatory variables (1) (2) (3) (4) (5) (6)
Written calculation Read English/French Write English/French
Probit(ME) LPM Probit(ME) LPM Probit(ME) LPM
District Development
Scores
0.003 0.002 0.003 0.002 0.004** 0.004
(0.002) (0.003) (0.002) (0.003) (0.002) (0.003)
number of teachers in
the district
0.000 -0.000 -0.000 -0.000 -0.000 -0.000
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
English pass rate
BECE
0.002 0.003 0.002 0.003 0.001 0.002
(0.001) (0.002) (0.001) (0.002) (0.001) (0.002)
Mathematics pass rate
BECE
-0.001 -0.001 -0.001 -0.002 -0.001 -0.002
(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
Science pass rate
BECE
0.000 0.000 0.001 0.001 0.002 0.002
(0.001) (0.002) (0.002) (0.002) (0.002) (0.002)
Number of schools
with electricity
infrastructure
0.002** 0.001 0.002*** 0.002 0.002** 0.001
(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
Number of schools
with toilet infra-
structure
-0.001 -0.001 -0.001 -0.001 -0.000 -0.000
(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
Constant -0.090 -0.171 -0.073
(0.419) (0.421) (0.421)
Observations 2755 2755 2755 2755 2755 2755
R-Squared 0.167 0.164 0.169
Robust standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1; LPM Linear probability model; ME Marginal effects
R.E.Kofinti et al.
1 3
Table 9 Effect of NHIS on learning outcomes in Ghana (full results: IV)
Explanatory Variables (1) (2) (3)
Written calculation Read English/French Write English/French
NHIS Household 0.356*** 0.361*** 0.306***
(0.093) (0.090) (0.093)
Age of child 0.036*** 0.048*** 0.048***
(0.005) (0.006) (0.006)
Female child -0.005 0.017 0.002
(0.023) (0.023) (0.022)
Son/daughter to household head 0.046 0.086** 0.080*
(0.034) (0.040) (0.041)
Grandchild to household head 0.035 0.127*0.088
(0.059) (0.065) (0.065)
Child suffered from illness -0.011 0.027 0.020
(0.042) (0.041) (0.042)
Age of head 0.001 -0.000 -0.000
(0.001) (0.001) (0.001)
Female-headed home 0.106** 0.032 0.011
(0.042) (0.049) (0.050)
Educated headed homes 0.070** 0.026 0.010
(0.035) (0.034) (0.035)
Account ownership by head 0.055 0.040 0.042
(0.037) (0.038) (0.037)
Rural household -0.044 -0.032 -0.069
(0.058) (0.058) (0.056)
1 3
Household National Health Insurance Subscription andLearning…
Table 9 (continued)
Explanatory Variables (1) (2) (3)
Written calculation Read English/French Write English/French
Household head is married 0.057 0.011 0.012
(0.055) (0.056) (0.055)
Household head is Separated/Divorced/
Widowed Household
-0.057 -0.072 -0.047
(0.058) (0.064) (0.067)
Number of household members -0.000 -0.002 -0.002
(0.004) (0.004) (0.004)
Household operates a farm -0.015 -0.044 -0.034
(0.043) (0.051) (0.051)
Household income per capita 0.000 0.000 0.000
(0.000) (0.000) (0.000)
Non-food price index -0.383 -0.392 -0.585
(0.412) (0.449) (0.427)
Household connected to National Elec-
tricity Grid
0.035 0.051 0.053
(0.033) (0.033) (0.033)
Coastal Zone 0.187*** 0.176*** 0.216***
(0.064) (0.068) (0.068)
Forest Zone 0.152*** 0.195*** 0.251***
(0.054) (0.059) (0.056)
District Development Scores 0.002 0.002 0.003
(0.003) (0.003) (0.003)
number of teachers in the district -0.000 -0.000 -0.000
(0.000) (0.000) (0.000)
R.E.Kofinti et al.
1 3
Table 9 (continued)
Explanatory Variables (1) (2) (3)
Written calculation Read English/French Write English/French
English pass rate BECE 0.003 0.002 0.002
(0.002) (0.002) (0.002)
Mathematics pass rate BECE -0.001 -0.002 -0.002
(0.001) (0.001) (0.001)
Science pass rate BECE 0.001 0.001 0.002
(0.002) (0.002) (0.002)
Number of schools with electricity
infrastructure
0.001 0.002 0.002
(0.001) (0.001) (0.001)
Number of schools with toilet infra-
structure
-0.001 -0.000 -0.000
(0.001) (0.001) (0.001)
Constant -0.063 -0.144 -0.052
(0.424) (0.428) (0.423)
Observations 2755 2755 2755
R-Squared 0.112 0.117 0.139
Robust Standard errors in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01
1 3
Household National Health Insurance Subscription andLearning…
Table 10 Effect of NHIS subscription on health expenditure (Potential Chanel-OLS)
Explanatory Variables (1)
Logof health expenditure
NHIS Household -0.436**
(0.179)
Age of child -0.038*
(0.023)
Female child 0.189
(0.117)
Son/daughter to household head 0.170
(0.224)
Grandchild to household head -0.091
(0.317)
Child suffered from illness 0.294
(0.227)
Age of head -0.011
(0.008)
Female-headed home 0.084
(0.298)
Educated headed homes -0.229
(0.184)
Account ownership by head -0.280
(0.192)
Rural household 0.339
(0.281)
Household head is married 0.795**
(0.382)
Household head is Separated/Divorced/Widowed 0.745
(0.463)
Household operates a farm -0.334
(0.332)
Household income per capita -0.000
(0.000)
Non-food price index -0.068
(2.663)
Household connected to National Electricity Grid -0.051
(0.207)
Number of household members 0.084***
(0.032)
Presence of elderly(65 +) 0.032
(0.227)
Presence of disability 0.628*
(0.321)
R.E.Kofinti et al.
1 3
Table 10 (continued)
Explanatory Variables (1)
Logof health expenditure
Household member hospitalised 0.702***
(0.219)
Coastal -0.502
(0.365)
Forest -0.125
(0.313)
Constant 1.550
(2.679)
Observations 2755
R-Squared 0.107
Robust Standard errors in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01
1 3
Household National Health Insurance Subscription andLearning…
Table 11 Effect of NHIS subscription on educational expenditure(Potential Chanel-OLS)
Explanatory Variables (1) (2)
Log of total household education expenditure Log of household expenditure on basic school
NHIS Household 0.726*** 0.675***
(0.110) (0.104)
Age of child 0.022 -0.003
(0.017) (0.016)
Female child -0.040 0.003
(0.074) (0.071)
Son/daughter to household head 0.146 0.223
(0.169) (0.160)
Grandchild to household head 0.198 0.477*
(0.268) (0.262)
Child suffered from illness 0.239*0.233*
(0.122) (0.124)
Age of head 0.010** 0.004
(0.005) (0.005)
Female-headed home 0.365** 0.275
(0.164) (0.169)
Educated headed homes 0.444*** 0.527***
(0.133) (0.126)
Account ownership by head 0.300*** 0.228*
(0.114) (0.116)
Rural household -0.379** -0.182
(0.150) (0.173)
R.E.Kofinti et al.
1 3
Table 11 (continued)
Explanatory Variables (1) (2)
Log of total household education expenditure Log of household expenditure on basic school
Household head is married -0.073 -0.104
(0.238) (0.237)
Household head is Separated/Divorced/Widowed -0.268 -0.230
(0.243) (0.241)
Household operates a farm 0.202 0.103
(0.190) (0.197)
Household income per capita 0.000 0.000
(0.000) (0.000)
Non-food price index -0.063 0.229
(1.237) (1.212)
Household connected to National Electricity Grid 0.506*** 0.383***
(0.130) (0.124)
Coastal 1.329*** 1.474***
(0.179) (0.188)
Forest 0.791*** 0.890***
(0.186) (0.180)
Number of household members 0.175*** 0.141***
(0.020) (0.017)
Travel to school on foot -0.929*** -0.946***
(0.169) (0.157)
Received a scholarship/bursary -0.546 -0.374
(0.378) (0.363)
1 3
Household National Health Insurance Subscription andLearning…
Table 11 (continued)
Explanatory Variables (1) (2)
Log of total household education expenditure Log of household expenditure on basic school
Constant 3.239*** 3.467***
(1.215) (1.190)
Observations 2755 2755
R-Squared 0.264 0.262
Robust Standard errors in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01
R.E.Kofinti et al.
1 3
Table8
Table9
Table10
Table11
Table 12 Effect of NHIS on learning outcomes in Ghana (Ghanaian/Native language)
Robust Standard errors in parentheses * p < 0.1, ** p < 0.05
(1) (1)
Read a Ghanaian language Write a Ghanaian language
NHIS Household 0.136 0.100
(0.094) (0.096)
Child Characteristics Yes Yes
Household Characteristics Yes Yes
District Characteristics Yes Ye s
Kleibergen-Paap rk LM statistic 50.378 50.378
[0.000] [0.000]
Kleibergen-Paap rk Wald F statistic 96.122 96.122
First Stage
Proportion of neighbours subscribed to NHIS 0.007*** 0.007***
(0.001) (0.001)
Observations 2755 2755
R-Squared 0.200 0.193
1 3
Household National Health Insurance Subscription andLearning…
Table 13 Propensity score matching using different matching methods
Bootstrap standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1
Variables Written calculation Read English/French Write English/French
1 – Nearest Neighbour (one-to-one) 0.058* 0.060* 0.034
(0.032) (0.032) (0.028)
5 – Nearest Neighbour 0.079*** 0.084*** 0.075***
(0.021) (0.028) (0.025)
Radius 0.079*** 0.141*** 0.120***
(0.021) (0.020) (0.018)
Kernel 0.069*** 0.080*** 0.063***
(0.017) (0.022) (0.019)
Local linear regression 0.070*** 0.082*** 0.062***
(0.019) (0.019) (0.024)
Observations 2,755 2,755 2,755
Table12
Table13
Acknowledgements We are indebted to the Ghana Statistical Service for granting us the permission to
use the GLSS7 data for this study. The access link for GLSS7 datasets is: https:// www2. stats ghana. gov.
gh/ nada/ index. php/ catal og/ 97/ study- descr iption
Additionally, we are also grateful to Mrs Patricia Anna Mensah, English Tutor at the Accra Academy
Senior High School of Ghana for proofreading the revised version of the Manuscript.
Author Contributions All authors contributed to the conception and design of the study. Material prep-
aration and analysis were performed by Raymond Elikplim Kofinti, Josephine Baako- Amponsah and
Prince Danso. The first draft of the manuscript was written by Raymond Elikplim Kofinti and all authors
commented on previous versions of the manuscript. All authors read and approved the final manuscript.
Data Availability Data and materials are available upon request to the Authors.
Declarations
The authors have no relevant financial or non-financial interests to disclose.
Ethics Approval Not applicable.
Consent to Participate Not applicable.
Consent for Publication Not applicable.
Conflict of Interest Raymond Elikplim Kofinti, Josephine Baako- Amponsah and Prince Danso declare
that they have no conflict of interest.
References
Adu-Gyamfi, S., Donkoh, W. J., & Addo, A. A. (2016). Educational reforms in Ghana: Past and present.
Journal of Education and Human Development, 5(3), 158–172.
Afoakwah, C., & Koomson, I. (2021). How does school travel time impact children’s learning outcomes
in a developing country? Review of Economics of the Household, 19(4), 1077–1097.
R.E.Kofinti et al.
1 3
Afoakwah, C., Deng, X., & Onur, I. (2020). Women’s bargaining power and children’s schooling out-
comes: Evidence from Ghana. Feminist Economics, 26(3), 1–29.
Akazili, J., Ataguba, J. E. O., Kanmiki, E. W., Gyapong, J., Sankoh, O., Oduro, A., & McIntyre, D.
(2017a). Assessing the impoverishment effects of out-of-pocket healthcare payments prior to the
uptake of the national health insurance scheme in Ghana. BMC International Health and Human
Rights, 17(1), 1–8.
Akazili, J., McIntyre, D., Kanmiki, E. W., Gyapong, J., Oduro, A., Sankoh, O., & Ataguba, J. E.
(2017b). Assessing the catastrophic effects of out-of-pocket healthcare payments prior to the
uptake of a nationwide health insurance scheme in Ghana. Global Health Action, 10(1), 1289735.
Alcaraz, C., Chiquiar, D., Orraca, M. J., & Salcedo, A. (2017). The effect of publicly provided health
insurance on education outcomes in Mexico. The World Bank Economic Review, 30(Supple-
ment_1), S145–S156.
Alhassan, R. K., Nketiah-Amponsah, E., & Arhinful, D. K. (2016). A review of the National Health
Insurance Scheme in Ghana: What are the sustainability threats and prospects? PLoS ONE,
11(11), e0165151.
Amo-Adjei, J., Anku, P. J., Amo, H. F., & Effah, M. O. (2016). Perception of quality of health deliv-
ery and health insurance subscription in Ghana. BMC Health Services Research, 16(1), 1–11.
Atake, E. H. (2018). Health shocks in Sub-Saharan Africa: Are the poor and uninsured households
more vulnerable? Health Economics Review, 8, 26. https:// doi. org/ 10. 1186/ s13561- 018- 0210-x
Aziabah, M. A. (2018). Educational reforms in Ghana. In The Politics of Educational Reform in
Ghana (pp. 43–51). Springer, Cham.
Barasa, E., Kazungu, J., Nguhiu, P., etal. (2021). Examining the level and inequality in health insur-
ance coverage in 36 sub-Saharan African countries. BMJ Global Health, 6, e004712. https:// doi.
org/ 10. 1136/ bmjgh- 2020- 004712
Cameron, A. C., & Trivedi, P. K. (2005). Microeconometrics: Methods and applications. Cambridge
University Press.
Cameron, A. C., & Trivedi, P. K. (2010). Microeconometrics using Stata (vol. 2). Stata Press College
Station
Churchill, S. A., & Marisetty, V. B. (2020). Financial inclusion and poverty: A tale of forty-five thou-
sand households. Applied Economics, 52(16), 1777–1788.
Churchill, S. A., Smyth, R., & Farrell, L. (2020). Fuel poverty and subjective wellbeing. Energy Eco-
nomics, 86, 104650.
Donkoh, S. A., & Amikuzuno, J. A. (2011). The determinants of household education expenditure in Ghana.
Dillon, A. (2008). Child labor and schooling responses to production and health shocks in Northern
Mali. In: IFPRI Working Paper.
Fabre, A., & Pallage, S. (2015). Child labor, idiosyncratic shocks, and social policy. Journal of Mac-
roeconomics, 45, 394–411.
Fiestas Navarrete, L., Ghislandi, S., Stuckler, D., & Tediosi, F. (2019). Inequalities in the benefits of
national health insurance on financial protection from out-of-pocket payments and access to health
services: Cross-sectional evidence from Ghana. Health Policy and Planning, 34(9), 694–705.
Finkelstein, A., Mahoney, N., & Notowidigdo, M. J. (2018). What does (Formal) health insurance do, and
for whom? Annual Review of Economics, 10(1), 261–328.
Finkelstein, A., Taubman, S., Wright, B., Bernstein, M., Gruber, J., Newhouse, J. P., ..., & Oregon Health
Study Group. (2012). The Oregon health insurance experiment: evidence from the first year.The Quar-
terly journal of economics,127(3), 1057-1106.
Frempong, R. B., Orkoh, E., & Kofinti, R. E. (2021). Household’s use of cooking gas and Children’s learning
outcomes in rural Ghana. Energy Economics, 103, 105617.
Geng, X., Janssens, W., Kramer, B., & van der List, M. (2018). Health insurance, a friend in need? Impacts
of formal insurance and crowding out of informal insurance. World Development, 111, 196–210.
Gertler, P. J., Martinez, S., Premand, P., Rawlings, L. B., & Vermeersch, C. M. (2016). Impact evaluation in
practice. World Bank Publications.
Greene, W. H. (2003). Econometric analysis. Pearson Education India.
Ghana Statistical Service (GSS). (2018). The Ghana living standards survey round seven: Poverty trends in
Ghana-2005–2017. Accra. Retrieved on 3rd January 2022 from https:// www2. stats ghana. gov. gh/ docfi
les/ publi catio ns/ GLSS7/ Pover ty% 20Pro file% 20Rep ort_ 2005% 20-% 202017. pdf
Ghana Statistical Service (GSS). (2019). The Ghana living standards survey round seven main report. Accra.
Retrieved on 8th February 2022 from https:// www. stats ghana. gov. gh/ gssma in/ fileU pload/ press relea se/
GLSS7% 20MAIN% 20REP ORT_ FINAL. pdf
1 3
Household National Health Insurance Subscription andLearning…
Ghana Statistical Service. (2021). General Report Volume 3C: Background Characteristics on 2021 Popula-
tion and Housing Census.
Government of Ghana. (2004). National Health Insurance Regulations, 2004 (L.I. 1809). Accra.
Haile, G., & Haile, B. (2012). Child labour and child schooling in rural Ethiopia: Nature and trade-off. Edu-
cation Economics, 20(4), 365–385.
He, L., & Zhou, S. (2022). Household financial vulnerability to income and medical expenditure shocks:
Measurement and determinants. International Journal of Environmental Research and Public Health.
https:// doi. org/ 10. 3390/ ijerp h1908 4480
Hellinger, F. J., & Wong, H. S. (2000). Selection bias in HMOs: A review of the evidence. Medical Care
Research and Review, 57(4), 405–439.
Huy, V. Q. (2012). Determinants of educational expenditure in Vietnam. International Journal of Applied
Economics, 9(1), 59–72.
Islam, A., & Maitra, P. (2012). Health shocks and consumption smoothing in rural households: Does micro-
credit have a role to play? Journal of Development Economics, 97(2), 232–243.
Johnson, E., & Reynolds, C. L. (2013). The effect of household hospitalizations on the educational attain-
ment of youth. Economics of Education Review., 37, 165e182.
Khandker, S. R., Barnes, D. F., & Samad, H. A. (2010). Energy poverty in rural and urban India: Are the
energy poor also income poor? The World Bank.
Kofinti, R. E., Asmah, E. E., & Ameyaw, E. K. (2022). Comparative study of the effect of National Health
Insurance Scheme on use of delivery and antenatal care services between rural and urban women in
Ghana. Health Economics Review, 12(1), 13. https:// doi. org/ 10. 1186/ s13561- 022- 00357-z
Koomson, I., & Afoakwah, C. (2022). Can financial inclusion improve children’s learning outcomes and late
school enrolment in a developing country? Applied Economics, 1–18.
Koomson, I., & Danquah, M. (2021). Financial inclusion and energy poverty: Empirical evidence from
Ghana. Energy Economics, 94, 105085.
Liang, S.-Y., Phillips, K. A., & Wang, H. (2004). Selection bias into health plans with specific characteristics:
A case study of endogeneity of gatekeeper requirements and mammography utilization. Health Services
and Outcomes Research Methodology, 5(2), 103–118.
Lim, S. S. (2020). Parental chronic illness and child education: Evidence from children in Indonesia. Interna-
tional Journal of Educational Development, 73, 102099.
Liu, K. (2016). Insuring against health shocks: Health insurance and household choices. Journal of Health
Economics, 46, 16–32.
Lozano, R., Fullman, N., Mumford, J. E., Knight, M., Barthelemy, C. M., Abbafati, C., ..., & Cárdenas, R.
(2020). Measuring universal health coverage based on an index of effective coverage of health services
in 204 countries and territories, 1990–2019: A systematic analysis for the Global Burden of Disease
Study 2019. The Lancet, 396(10258), 1250-1284.
Lu, C., Chin, B., Lewandowski, J. L., Basinga, P., Hirschhorn, L. R., Hill, K., Murray, M., & Binagwaho, A.
(2012). Towards universal health coverage: An evaluation of Rwanda Mutuelles in its first eight years.
PLoS One, 7(6), e39282.
Ministry of Health. (2018). Holistic Assessment of 2017 health sector programme of work. Accra.
National Health Insurance Authority. (2010). National Health Insurance Scheme, Annual Report 2009.
Accra.
National Health Insurance Scheme. (2018). Benefits Package. http:// www. nhis. gov. gh/ benefi ts. aspx.
Accessed 16 April 2018.
Nghiem, S. H., & Connelly, L. B. (2017). Convergence and determinants of health expenditures in OECD
countries. Health Economics Review, 7(1), 1–11.
Okoroh, J., Essoun, S., Seddoh, A., etal. (2018). Evaluating the impact of the national health insurance
scheme of Ghana on out-of-pocket expenditures: A systematic review. BMC Health Services Research,
18, 426. https:// doi. org/ 10. 1186/ s12913- 018- 3249-9
Ouadika, S. A. B. (2020). Health shocks and vulnerability to poverty in Congo. Humanity Soc Sci Commun,
7, 182. https:// doi. org/ 10. 1057/ s41599- 020- 00674-w
Palas, M. J. U., Ashraf, M., & Ray, P. K. (2017). Financing universal health coverage: A systematic review.
The International Technology Management Review, 6(4), 133–148.
Peng, X., & Conley, D. (2016). The implication of health insurance for child development and maternal
nutrition: Evidence from China. The European Journal of Health Economics, 17(5), 521–534.
Stock, J. H., & Yogo, M. (2002). Testing for weak instruments in linear IV regression. National Bureau of
Economic Research Cambridge.
R.E.Kofinti et al.
1 3
Suhrcke, M., & de Paz Nieves, C. (2011). The impact of health and health behaviours on educational out-
comes in high-income countries: A review of the evidence. WHO Regional Office for Europe.
UNICEF Ghana. (2017). Ghana’s District League Table2017. Retrieved on 2nd February 2022 from https://
afric aopen data. org/ datas et/ ghana-s- distr ict- league- table- 2017/ resou rce/ 973b0 d74- a00f- 4342- b6ab-
68073 4ff38 27? inner_ span= True
Wagstaff, A., & Lindelow, M. (2014). Are health shocks different? Evidence from a multi-shock survey in
Laos. Health Economics, 23(6), 706–718.
Wang, W., Temsah, G., & Mallick, L. (2017). The impact of health insurance on maternal health care utiliza-
tion: Evidence from Ghana, Indonesia and Rwanda. Health Policy and Planning, 32(3), 366–375.
Wooldridge, J. M. (2015). Introductory econometrics: A modern approach. Cengage Learning.
World Health Organization. (2017). Tracking Universal Health Coverage: 2017 Global Monitoring Report.
World Health Organization and the International Bank for Reconstruction and Development/The World
Bank.
Yetim, B., İlgün, G., Çilhoroz, Y., Demirci, Ş, & Konca, M. (2021). The socioeconomic determinants of
health expenditure in OECD: An examination on panel data. International Journal of Healthcare Man-
agement, 14(4), 1265–1269.
Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Springer Nature or its licensor holds exclusive rights to this article under a publishing agreement with the
author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article
is solely governed by the terms of such publishing agreement and applicable law.
A preview of this full-text is provided by Springer Nature.
Content available from Child Indicators Research
This content is subject to copyright. Terms and conditions apply.