World Development Vol. xx, pp.

xxx–xxx, 2017
0305-750X/Ó 2017 Elsevier Ltd. All rights reserved.

www.elsevier.com/locate/worlddev
http://dx.doi.org/10.1016/j.worlddev.2017.05.007

New Evidence on the Impact of Large-scale Conditional Cash

Transfers on Child Vaccination Rates: The Case of a
Clustered-Randomized Trial in Indonesia
DIAN KUSUMA a,b, HASBULLAH THABRANY b, BUDI HIDAYAT b, MARGARET McCONNELL a,
PETER BERMAN a and JESSICA COHEN a,*
a
Harvard T.H. Chan School of Public Health, Boston, USA
b
Faculty of Public Health Universitas Indonesia, Indonesia
Summary. — Despite recent progress, millions of children still die every year from vaccine-preventable diseases. One strategy is Condi-
tional Cash Transfers (CCTs), which provide cash payments to poor households in exchange for compliance with health-related condi-
tionalities including child vaccination. Using a randomized trial, we provide new evidence on the impact of large-scale CCTs on child
vaccination rates in Indonesia by investigating the Program Keluarga Harapan (PKH) with a sample of over four thousand children
under two years old. After two years of implementation from 2007 to 2009, difference-in-differences (DID) estimates show that PKH
significantly increases child vaccination rates for all basic vaccine types by up to 30% compared to the control group means among
children aged less than 12 months old but PKH shows modest effects among children aged 12–23 months old. There is also evidence
that PKH is equity enhancing by increasing child vaccination rates for most vaccine types by up to 52% among children aged less than
12 months old living with less educated mothers (below six years). All this underscores the ability of cash transfers to reach poor children
for whom health systems supply-side-oriented strategies have been less successful.
Ó 2017 Elsevier Ltd. All rights reserved.

Key words — Indonesia, Conditional Cash Transfers (CCT), household, child vaccination, clustered-randomized trials

1. INTRODUCTION While there are several rigorous studies on the impact of

Despite recent progress, millions of children still die every
year from vaccine-preventable diseases. The WHO (2016)
estimated that 19.4 million children did not receive routine
life-saving vaccinations in 2015. One explanation is a consider-
able inequality in child vaccination. Globally, countries in the
African region have most unvaccinated children and very low
national immunization coverage based on the WHO/UNICEF
report. In Indonesia, the proportion of one-year olds com-
pletely immunized among the poorest quintile is only half that
of the wealthiest—39% and 75% respectively (Utomo,
Sucahya, & Utami, 2011).
One strategy is Conditional Cash Transfers (CCTs), which
provide cash payments to poor households in exchange for
compliance with health-related conditionalities including child
vaccination (Fiszbein, Schady, & Ferreira, 2009). There are
several theoretical pathways from conditionality to improved
vaccination. Based on the human capital theory, consumers
will invest in health if the expected private benefit exceeds
the cost (Grossman, 2000). In terms of vaccination, it is harder
to estimate the benefit since it comes years in the future partic-
ularly while facing the cost such as geographical barriers (e.g.,
no health facilities nearby, vaccine stock-outs), financial barri-
ers, and social barriers (e.g., belief that vaccines are harmful)
(Adato, Roopnaraine, & Becker, 2011). The cash element of
CCT can help with financial barriers and the conditionality
element might be seen as a way to transfer health information
on the benefit of vaccination and signal the importance of vac-
cination for both households and health workers. Implemen-
tation issues, however, might jeopardize this potential
effectiveness, rendering the conditions meaningless (Kremer
& Glennerster, 2012).
1
Please cite this article in press as: Kusuma, D. et al. New Evidence on the Impact of Large-scale Conditional Cash Transfers on Child Vaccination Rates:
The Case of a Clustered-Randomized Trial in Indonesia, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.05.007
CCTs on child vaccination rates, the literature is limited in
two ways. First, evidence from randomized controlled trials
(RCTs) is still inconclusive (Johri et al., 2015; Owusu-Addo
& Cross, 2014; Ranganathan & Lagarde, 2012). Evidence
shows that CCTs have positive impact on the coverage of
measles-containing vaccine (MCV) by 11 and 3 percentage
points in Nicaragua (Red de Protección Social) and Mexico
(PROGRESA). But evidence also shows that CCTs have no
significant impact on the coverage of MCV and Bacille
Calmette-Guérin (BCG) in Honduras (Programa de Asig-
nación Familiar), Mexico, and Zimbabwe (Robertson et al.,
2013; Barham & Maluccio, 2009; Barham, 2005; Morris,
Flores, Olinto, & Medina, 2004). Secondly, large-scale evi-
dence on the impact of CCTs on child vaccination is limited
to only PROGRESA with over five thousand children under
the age of three years. Other studies in Nicaragua and Hon-
duras used about a thousand children. While the evidence
from small-scale evaluations seems to be promising, one of
the main challenges is whether the findings would be replicable
at scale. A large-scale evaluation helps establish whether the
CCT impact on vaccination is possible when taking into
account the complexity of implementation namely issues with
* The authors are especially grateful to Vivi Yulaswati (National Planning
Agency), Harapan Lumban Gaol (Ministry of Social Affairs), Jay
Rosengard (Harvard Kennedy School), and Ajay Mahal (Monash
University) for their support and many helpful discussions. We also
thank seminar participants at the Harvard T.H. Chan School of Public
Health and Harvard Kennedy School for valuable suggestions. Research
fellowship from the Harvard Kennedy School Indonesia Program is
gratefully acknowledged. Final revision accepted: May 13, 2017.
2 WORLD DEVELOPMENT
cash distribution, monitoring conditionality, and supply-side
improvements (Madon, Hofman, & Glass, 2007). 1
Taking advantage of a large clustered-randomized trial in
Indonesia, we provide new evidence on the impact of large-
scale household cash transfers (PKH) on child vaccination
rates using a sample over four thousand children under two
years old. Previous PKH study (Alatas et al., 2011) evaluated
complete vaccination among all children under three years old
and showed that PKH did not increase complete vaccination
overall. However, they showed significant effects by 3.3 per-
centage points among confirmed program beneficiaries (i.e.,
participant effect). Our study provides a more in-depth analy-
sis by investigating the PKH effects on different types of vac-
cines. There are at least two reasons: each vaccine combats a
different disease with its own prevalence and potential conse-
quences; vaccines can differ in doses, age, modality, and avail-
ability (Barham & Maluccio, 2009).
2. BACKGROUND
(a) Vaccination programs in Indonesia
Indonesia has a decentralized health system with most
responsibilities shared to district governments. In public sec-
tor, district hospitals provide secondary/tertiary care; district
health offices provide public health services; and health centers
(Puskesmas) and network provide primary care. Puskesmas
network includes supporting Puskesmas (Pustu) and village
midwife (Polindes). There are also integrated care posts
(Posyandu) run by cadres with health services provided by
Puskesmas doctors, nurses, and midwives. In private sector,
there are doctor practices, midwife practices, and hospitals.
The country adopted basic child vaccination during the
1980s. The Ministry of Health (MOH) in Indonesia has two
main vaccination strategies: (1) to provide vaccinations at
public health facilities; and (2) to hold vaccination campaigns,
which typically take vaccines to households. The first strategy
relies on individuals bringing their children to facilities, which
tends to result in incomplete coverage of the population.
Health officials assert that barriers to this strategy include
geography and social norms. The archipelagic nature of the
country makes it challenging to implement monthly vaccina-
tion in areas where facilities are lacking or less supply-ready
(MOH, 2010a). Also, negative perceptions of immunization
still exist (Judarwanto, 2012).
Prior to the PKH program, officially reported vaccination
rates in Indonesia were 84% for BCG, 82% for MCV, 72%
for OPV3, and 74% for DPT3 in 2007. These rates are lower
than those in neighboring countries such as Thailand and
the Philippines where the rates for all those vaccines are at
least 95% (WHO/UNICEF, 2012). Further, the MOH uses
an indicator called Universal Coverage of Immunization
(UCI) village. A village is considered universal if at least
80% of all children aged 12 months or younger in the village
complete basic child vaccination. Data show a constant and
flat trend in the proportion of UCI villages at 70%, 76%,
73%, and 71% during the period of 2004, 2005, 2006, 2007,
respectively, prior to PKH (MOH, 2007a).
(b) Program Keluarga Harapan (PKH)
In 2007, the government launched a large-scale pilot of
PKH, a CCT to household. The goals are to reduce poverty,
maternal and child mortality, and to ensure universal coverage
of basic education. PKH, a traditional CCT program to poor
Please cite this article in press as: Kusuma, D. et al. New Evidence on the Impact of Large-scale Conditional Cash Transfers on Child Vaccination Rates:
The Case of a Clustered-Randomized Trial in Indonesia, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.05.007
Table 1. Conditionality and target indicators for PKH
Health indicators
1. Four prenatal care visits
2. Taking iron tablets during pregnancy
3. Delivery assisted by a trained professional
4. Two postnatal care visits
5. Complete childhood immunizations
6. Adequate monthly weight increases for infants
7. Monthly weighing for children under three and biannually
for children under five
8. Vitamin A twice a year for children under five
Education indicators
9. Primary school enrollment of children 6–to-12 years old
10. Minimum attendance rate of 85% for primary school-aged
children
11. Junior secondary school enrollment of children 13–to-15
years old
12. Minimum attendance rate of 85% for junior secondary
school-aged children
Source: MOSA (2007).
households, was piloted in six provinces including West Java,
East Java, North Sulawesi, Gorontalo, East Nusa Tenggara
(ENT), and Jakarta. It was designed to achieve the target indi-
cators or conditionality provided in Table 1. For child health,
conditionality includes complete childhood immunizations,
adequate weighting, and vitamin A (Sparrow, Moeis,
Damayanti, & Herawati, 2008; Alatas et al., 2011).
There are four features of PKH: (1) cash given to mothers
quarterly; (2) conditionality and cash penalty; (3) field facilita-
tors; and (4) improvements in supply-side readiness. First, the
cash, collected by mothers through the nearest post office, ran-
ged from $60 to 220 per household per year depending on the
number and aged of children. There are no rules on how to
spend the cash. The cash is approximately 15–20% of the esti-
mated consumption of poor households. The fixed amount is
$20 per year. If a mother is pregnant and/or has children aged
0–6 years, she will receive additional $80 per year, regardless
of the number of children. If a mother has a child at primary
school, she will receive additional $40 per year. If a mother has
a child at secondary school, she will receive additional $80 per
year. The differences in cash amount for primary and sec-
ondary schools are due potentially to higher expenditures for
the latter. Second, PKH was designed to enforce conditional-
ity with a cash penalty for noncompliance: the first breach is a
warning, the second reduces the cash by 10% in the next pay-
ment, and the third is expulsion. To comply with the condi-
tionality, beneficiaries can choose various health facilities
ranging from district hospitals, Puskesmas, Polindes, and
Posyandu. Third, field facilitators are trained to advise benefi-
ciaries to comply with conditionalities, to inform them of their
rights and obligations, and to monitor eligibility. Fourth,
PKH is implemented mostly in urban areas that are supply
ready based on a statistical analysis of existing health and edu-
cation facilities in the sub-districts. Readiness is a precondition
to be part of the pilot. The threshold for readiness, however,
was set lower for off-Java island areas to ensure more inclusion
in the pilot. More details are provided elsewhere (Kusuma,
Cohen, McConnell, & Berman, 2016; Alatas et al., 2011;
Sparrow et al., 2008; Hicklin, 2008) and in Table A.1 of the
Appendix.
 NEW EVIDENCE ON THE IMPACT OF LARGE-SCALE CONDITIONAL
In practice, however, there are some deviations from the
PKH design. The enforcement to comply with conditions were
lacking due mainly to the lack of readiness of providers to fill
out verification forms and the slow establishment of manage-
ment information systems (MIS). It is important, however, to
note that enforcement to comply with conditions is just one of
the aforementioned features of PKH that were in place. In
fact, beneficiaries were likely to think that there were condi-
tionality and potential cash penalty because field facilitators
informed them. A study shows that over 90% of beneficiaries
in the spot check samples in Jakarta, West Java, and ENT
province knew about the terms and conditions of PKH
(CHR-UI, 2010; Ayala, 2010).
In addition, the government also launched the National
Program for Community Empowerment (PNPM) was
launched in 2007 toward poverty alleviation. This includes
PNPM Rural, PNPM Urban, and PNPM Generasi. PNPM
Rural (previously called Kecamatan Development Program
[KDP]) aims to empower rural communities and to provide
financial support for building infrastructure and capacity in
subdistricts. PNPM Urban has similar objective but to
empower urban communities. PNPM Generasi is an incen-
tivized community block grant program aiming mainly at
rural communities. PNPM Generasi provides annual block
grants to communities to be used for various purposes to
improve local health and education services. While PNPM
Generasi has no overlapping districts with PKH, some PNPM
Urban and PNPM Rural were implemented in some PKH
areas. In 2007, 73 (out of 180) PKH subdistricts received other
PNPM programs, as did 76 (out of 180) control subdistricts; in
2008, 107 treatment subdistricts received other PNPM pro-
grams, as did 105 control subdistricts. However, since the
other PNPMs focus on basic infrastructure, not on health or
education services, it is unlikely that they have substantial
impacts on our results (Alatas et al., 2011; Olken, Onishi, &
Wong, 2011). 2
3. METHODS
(a) Evaluation design and data
We use the impact evaluation data from the baseline
(2007) and follow-up (2009) surveys conducted by the
National Planning Agency and World Bank. The evaluation
was large scale with the overall sample of approximately
14,000 households. The baseline survey took place from
June to August 2007 and the follow-up survey was imple-
mented from October to December 2009. Ethical approval
is not required because this study uses public data without
identifiable private information. Random assignment was
implemented at the subdistrict level to allow for a rigorous
evaluation.
Sampling of PKH is as follows. The levels of geography
are ordered as follows: island, province, district, subdistrict,
village, and ward. Within the PKH provinces, 80% poorest
districts were included (based on school transition rates,
proportion of malnutrition, and proportion of poverty)
but only urban districts with no KDP subdistricts and
supply-ready subdistricts were included. KDP is a sub-
district development program that provides financial support
for infrastructure and capacities in rural areas. And then eli-
gible subdistricts were randomly allocated to treatment and
control groups with 180 subdistricts were randomly selected
in each. And then the sampling randomly selected 8 villages
per subdistrict, and 1 ward per village. Households in a
Please cite this article in press as: Kusuma, D. et al. New Evidence on the Impact of Large-scale Conditional Cash Transfers on Child Vaccination Rates:
The Case of a Clustered-Randomized Trial in Indonesia, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.05.007
3
ward were categorized into three groups: (i) households with
pregnant/lactating women, (ii) households with children
aged 6–15 years, and (iii) the remaining households. Five
households were randomly selected: two from group (i)
and three from group (ii). The selection of households is
conditional on being in the unconditional cash transfer
(UCT) 2005 list of poor households. Villages were first
screened on the UCT eligibility of all households and only
villages with at least five UCT-eligible households per ward
were considered for sampling. The UCT program provided
cash transfers to poor households in 2005 and 2008 due
to reduction of gasoline subsidy. All this results in samples
of 6 provinces, 44 districts, 360 subdistricts, and 14,000
households. More details are provided elsewhere (Kusuma
et al., 2016; Alatas et al., 2011; Sparrow et al., 2008) and
in Table A.2 of the Appendix.
In terms of datasets, we use a panel/longitudinal sample
of 4,245 children under two years old with vaccination data
followed for two years (a total of 8,490 observations for
analysis) and repeated cross sectional samples of 4,636 in
2007 and 2,066 children in 2009 (a total of 6,702 observa-
tions for analysis). We limit our analyses to children under
two years because: (1) all basic child vaccinations should be
completed within the age of 9 months; (2) using children
sample of 12–23 month old to see any catch-up effects.
For panel data, compared to 4,636 children under two at
baseline (2007), the attrition rates are 8.7% and 8.6% for
treatment and control group respectively—no significant dif-
ference. Our attrition analysis shows similar characteristics
between attrited and not attrited children and households
(see Table A.3 and Table A.4 of the Appendix). In terms
of data collection, the questionnaire combines two sources
of child immunization status: health card record (KMS/
KIA) and mother’s response on whether a child was immu-
nized. We use the former in our main analysis for more reli-
ability and provide the latter as robustness checks in the
Appendix.
(b) Dependent variables
Dependent variables include the number of doses each child
had received since birth of each of the following: (1) BCG; (2)
OPV; (3) DPT; (4) MCV; and (5) hepatitis B vaccine (HBV). A
binary dependent variable was created to measure coverage
for each vaccine—it took the value one if a child received all
of the recommended doses of that vaccine at the time of each
survey, and zero otherwise. A child was not considered vacci-
nated against DPT or polio unless he or she had received a
complete dose of each vaccine. OPV3 was used in conformity
with the international standard of completion and national
performance indicator (WHO/UNICEF, 2012). The condi-
tionality in PKH is for children to have a complete course
of vaccines.
Table 2 shows the child vaccination schedule published
by the MOH. The schedule stipulates that a child be given
the BCG vaccine at birth; complete the OPV3, DPT3, and
HBV3 vaccines at four months old; and be given the MCV
vaccine at nine months old. For OPV, while the common
practice in the country is four doses, three doses (OPV3)
are used as a performance indicator conforming to the
WHO standard (MOH, 2007a). The schedule shows that
a child should have completed all basic vaccinations by
the end of his/her first year. We examine the effects of
PKH for each type of vaccine and for two age groups:
children aged less than 12 months (<12 months) and 12–
23 months.
4 WORLD DEVELOPMENT
Table 2. Basic vaccination schedule for up-to-date vaccinations in Indonesia
Disease Vaccine
Tuberculosis BCG
Polio OPV
Diphtheria–pertussis–tetanus DPT
Measles MCV
Hepatitis B HBV
Source: MOH (2007a).
(c) Empirical specification
We use a difference-in-differences (DID) estimator to deter-
mine the average effect of treatment since it controls for differ-
ences in vaccination rates at baseline. This estimator also
controls for all characteristics that do not change over time
within treatment and control groups and all characteristics
that do change over time, but in the same way in each of
the groups. We compare the results of panel data DID and
repeated cross-sectional DID in our analyses. The former uses
the same group of children (panel data) and the latter uses the
same age of children (<12 months or 12–23 months) at base-
line and follow-up data. The main regression equation is:
Y ihjtv ¼ a0 þ a1 ðT j Þ þ a2 ðPostt Þ þ bðT j  Postt Þ þ ðX ihj Þc þ eihjtv
ð1Þ
where Y ihjtv ¼ 1 if child i in household h from sub-district j in
time period t is vaccinated with vaccine type v and zero other-
wise; Post = 1 if the year is 2009 (i.e., follow-up period) and
zero otherwise; T = 1 if the household is in the treatment
group and zero otherwise; X is a vector of characteristics for
children, households, and subdistricts; and e is the unobserved
idiosyncratic error (assumed to be uncorrelated with all other
variables).
We estimate the equation using ordinary least squares
(OLS) in Stata 14 (StataCorp, 2014). The characteristics of
children, households, and subdistricts are included as control
variables to improve the precision of estimation. Robust stan-
dard errors are calculated allowing for clustering at the subdis-
trict level. The parameter of interest is the b estimate, which is
the DID estimate of treatment effects for 2009 (relative to
2007). The program effects are identified by randomized
design. And because we do not condition on actual program
participation when using the survey data, but only on whether
a household resides in a treatment locality, the estimates reflect
the ‘‘intent-to-treat” average effect of the program.
4. RESULTS
Table 3 shows the outcome of randomization by examining
the differences in means at baseline between treatment and
control groups for an array of child, household, and subdis-
trict characteristics (panel a) and dependent variables of vac-
cine coverage (panel b). The balance test shows that the
differences are mostly not significant for both characteristics
and vaccine coverage (column 5), which is expected from ran-
dom allocation. In terms of vaccine coverage for children aged
<12 months (panel b), only BCG is provided since it is sched-
uled at birth so the coverage rate should include all children;
other vaccine types are not provided for this age group
because they are scheduled at later months so the coverage
rates would not include all children since many were probably
younger than the scheduled ages. Since we use vaccination
Please cite this article in press as: Kusuma, D. et al. New Evidence on the Impact of Large-scale Conditional Cash Transfers on Child Vaccination Rates:
The Case of a Clustered-Randomized Trial in Indonesia, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.05.007
Dose Ages given
1 0–1 month
3 2, 3, 4 months
3 2, 3, 4 months
1 9 months
3 2, 3, 4 months
status based on health card record in our main analysis, the
vaccination rates (column 1 panel b) are not as high as the
national figures (Section 2). The rates based on mother’s
response are more similar to the national figures and provided
in Table A.5 of the Appendix.
Table 4 shows the effects of PKH on vaccination rates
among children aged <12 months and 12–23 months using
panel data (columns 1–4) and repeated cross-section data anal-
yses (columns 5–8). Results show that PKH increases child
vaccination rates for all vaccine types among children aged
<12 months but also shows some effects among those aged
12–23 months. Among children aged <12 months, panel data
analysis shows that PKH increases BCG, OPV3, DPT3,
HBV3, and MCV by 10.9, 10.3, 8.8, 6.4, and 10.5 percentages
points respectively; relative to the control group means in
2009, these translate into increases of 28.7%, 30.3%, 28.4%,
26.7%, and 30% respectively. PKH also increases complete
vaccination rates by 7.2 percentage points, which translate
into 40% increase compared to control group mean in 2009.
Similarly, repeated cross-section analysis shows that PKH
increases most vaccine types but only significantly increases
two types. Table 4 (column 7) shows that PKH significantly
increases BCG and DPT3 by 6.2 and 5.1 percentage points
respectively, which translate into 11.9% and 24.3% increases
compared to the control group means in 2009. PKH also
increases complete vaccination rates by 2.9 percentage points
or an increase of 40% compared to the control group mean.
Among children aged 12–23 months, panel data analysis
shows that PKH increases OPV3 and MCV by 5.1 and 6.2 per-
centage points, which translate into 20.4% and 23.8% increases
compared to the control group means, respectively. Repeated
cross-section analysis only shows increase for OPV3 by 6.9
percentage points or 15.3% increase compared to the control
group mean. As a robustness check, Table A.6 in the Appen-
dix uses vaccination status based on mother’s response, which
also shows that PKH increases vaccination rates particularly
OPV3, DPT3, and MCV among children aged <12 months
but not much effect among those aged 12–23 months.
Next we provide subgroup analysis to see if PKH is equity
enhancing for instance by reducing the differences in vaccina-
tion coverage between children living in more and less disad-
vantaged areas. Since the results using repeated cross-section
analysis are similar, we only include the results using panel
data analysis. Table 5 provides subgroup analysis on vaccina-
tion rates by mother’s level of education. Results show that
PKH impact tends to be larger among children aged
<12 months living with less educated mothers compared to
those living with more educated ones (columns 3 and 7). For
children living with less educated mothers, results show that
PKH significantly increases vaccination rates for BCG,
OPV3, DPT3, and MCV by 13.6, 13.6, 10.1, and 9.7 percent-
age points respectively, which translate into 44%, 52%, 39%,
and 33% increases compared to the control means in 2009.
For those living with more educated mothers, PKH signifi-
cantly increases vaccination rates for BCG, OPV3, DPT3,
 NEW EVIDENCE ON THE IMPACT OF LARGE-SCALE CONDITIONAL 5

Table 3. Differences in means of characteristics and dependent variables by treatment status at baseline
Treatment (T) Control (C) Difference (T-C)
Mean SD Mean SD Mean SE
[1] [2] [3] [4] [5] [6]
(a) Characteristics
Child N = 4245
Female (=1) 0.49 0.50 0.50 0.50 0.00 (0.02)
Age in months 11.7 7.25 11.85 7.24 0.16 (0.24)
Household N = 4122
Block wall (=1) 0.42 0.49 0.44 0.50 0.02 (0.03)
Dirt floor (=1) 0.36 0.48 0.33 0.47 0.03 (0.03)
Zinc roof (=1) 0.19 0.39 0.17 0.38 0.01 (0.03)
Tile roof (=1) 0.70 0.46 0.71 0.45 0.01 (0.05)
Latrine in house (=1) 0.42 0.49 0.45 0.50 0.03 (0.03)
House has electricity (=1) 0.84 0.36 0.83 0.37 0.01 (0.03)
Piped water into house (=1) 0.12 0.33 0.12 0.33 0.00 (0.02)
Wood & coal cooking fuel (=1) 0.81 0.39 0.77 0.42 0.04 (0.02)
Land owned (square meters) 1615.85 4673.41 1960.15 5309.07 344.29 (354.26)
At least one animal (=1) 0.54 0.50 0.54 0.50 0.00 (0.03)
Per capita expenditure (Ln) 5.30 0.44 5.33 0.45 0.03 (0.02)
Mother’s education >6 years (=1) 0.68 0.47 0.69 0.46 0.01 (0.02)
Mother’s age in years 36.46 9.93 35.66 9.44 0.80** (0.39)
Household size 5.87 1.82 5.81 1.79 0.06 (0.10)
Number of under-five 1.40 0.58 1.41 0.57 0.01 (0.03)
Subdistrict N = 360
Doctor (number of, #) 6.13 6.44 6.04 6.32 0.09 (0.67)
Nurse (#) 10.87 7.90 9.89 6.37 0.98 (0.76)
Midwife (#) 11.38 4.72 11.37 4.69 0.02 (0.50)
Pharmacy (#) 2.43 2.34 2.55 2.28 0.12 (0.24)
Health Facility (#) 50.32 25.07 49.87 26.87 0.44 (2.74)
Stock-out at clinic last 2 months:
BCG 0.16 0.37 0.11 0.31 0.05 (0.04)
OPV 0.06 0.24 0.08 0.28 0.02 (0.03)
DPT 0.04 0.20 0.06 0.24 0.02 (0.02)
HBV 0.06 0.23 0.06 0.24 0.00 (0.03)
MCV 0.06 0.23 0.07 0.26 0.02 (0.03)

(b) Dependent variables

Cohort A: <12 months N = 2195
BCG 0.44 0.50 0.46 0.50 0.02 (0.03)
Cohort B: 12–23 months N = 2050
BCG 0.45 0.50 0.43 0.50 0.02 (0.03)
OPV3 0.37 0.48 0.37 0.48 0.01 (0.03)
DPT3 0.38 0.48 0.35 0.48 0.02 (0.03)
HBV3 0.33 0.47 0.32 0.47 0.01 (0.03)
MCV 0.36 0.48 0.36 0.48 0.00 (0.03)
Complete 0.23 0.42 0.23 0.42 0.00 (0.03)
Note: SD = standard deviation; SE = standard errors; LN = natural log. Data include children aged <12 months and 12–23 months old and their
households and subdistricts. Analyses use OLS regression comparing the means of treatment and control groups at baseline. For dependent variables,
vaccination status is based on health card record; only BCG is provided among children aged <12 months since the national schedule is at birth; other
vaccine types are not provided for this age group because the rates are not comprehensive since many children might be younger than the scheduled age.
SEs are robust and clustered at the subdistrict level. *p < 0.1, **p < 0.05.
The bold and italics show statistically significant estimates.

and MCV by 12.2, 7.9, 7.4, and 10.9 percentage points respec-
tively, which translate into 29%, 20%, 22%, and 29% increases
compared to the control group means. However, those differ-
ences are mostly not statistically significant (column 9). As for
children aged 12–23 months, our results show limited impact.
Table 6 provides subgroup analysis on vaccination rates by
urbanicity. PKH impact tend to be larger among children aged
<12 months in urban areas compared to those in rural areas.
For children aged <12 months in urban areas, PKH signifi-
cantly increases vaccination rates for BCG, OPV3, DPT3,
HBV3, and MCV by 9.5, 9.7, 9.2, 7.8, and 10.2 percentage
points respectively, which translate into 26%, 29%, 30%,
34%, and 30% increases compared to the control group means
in 2009. For those in rural, PKH significantly increases vacci-
nation rates for BCG and OPV3 by 15.2 and 11.6 percentage
points respectively, which translate into 39% and 33%
increases compared to the control group means. However,
those differences are not statistically significant (column 9).
As for children aged 12–23 months, our results show limited
impact.

Please cite this article in press as: Kusuma, D. et al. New Evidence on the Impact of Large-scale Conditional Cash Transfers on Child Vaccination Rates:
The Case of a Clustered-Randomized Trial in Indonesia, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.05.007
6 WORLD DEVELOPMENT

Table 4. Mean vaccination rates by treatment status and treatment effects

Year Panel DID Repeated cross-sectional DID
Mean Treatment effects Mean Treatment effects
T C b SE T C b SE
[1] [2] [3] [4] [5] [6] [7] [8]
<12 months N = 4390 N = 3437
BCG 2007 0.44 0.46 0.109** (0.032) 0.44 0.46 0.062* (0.037)
2009 0.47 0.38 0.56 0.52
OPV3 2007 0.23 0.26 0.103** (0.031) 0.23 0.26 0.046 (0.030)
2009 0.42 0.34 0.30 0.29
DPT3 2007 0.18 0.21 0.088** (0.028) 0.18 0.21 0.051* (0.028)
2009 0.38 0.31 0.23 0.21
HBV3 2007 0.17 0.18 0.064** (0.026) 0.17 0.18 0.030 (0.026)
2009 0.29 0.24 0.15 0.14
MCV 2007 0.08 0.09 0.105** (0.032) 0.08 0.10 0.000 (0.021)
2009 0.43 0.35 0.09 0.11
Complete 2007 0.04 0.06 0.072** (0.021) 0.04 0.06 0.029* (0.015)
2009 0.22 0.18 0.05 0.04
12–23 months N = 4100 N = 3265
BCG 2007 0.45 0.43 0.04 (0.033) 0.45 0.43 0.018 (0.040)
2009 0.35 0.28 0.56 0.52
OPV3 2007 0.37 0.37 0.051* (0.029) 0.37 0.37 0.069* (0.039)
2009 0.31 0.25 0.52 0.45
DPT3 2007 0.38 0.35 0.02 (0.029) 0.37 0.35 0.015 (0.040)
2009 0.30 0.24 0.46 0.42
HBV3 2007 0.33 0.32 0.04 (0.030) 0.33 0.32 0.021 (0.043)
2009 0.26 0.20 0.33 0.30
MCV 2007 0.36 0.36 0.062** (0.031) 0.37 0.36 0.024 (0.039)
2009 0.32 0.26 0.48 0.44
Complete 2007 0.23 0.23 0.04 (0.026) 0.23 0.23 0.003 (0.039)
2009 0.19 0.15 0.25 0.24
Note: T = treatment, C = control, SE = standard errors. Vaccination status is based on health card record. Panel DID data include panel children aged
0–23 months old followed for two years; sample size is panel sample both 2007 and 2009 of each children cohort (Table 3). Repeated cross-section DID
data include children aged 0–23 months in 2007 and 2009; sample size shown is from both 2007 and 2009. Analyses use OLS regression and the treatment
effects are b coefficient in equation (1). For children aged <12 months, OPV3 rates were low (23% and 26%) for treatment and control groups in 2007
because some children were too young for full doses; the rates went up as this cohort got older in 2009 (columns 1 and 2). Control variables include
characteristics of children, households, and subdistricts (Table 3 panel a). Robust standard errors (SE) are in parentheses and clustered at the subdistrict
level. *p < 0.1, **p < 0.05.
The bold and italics show statistically significant estimates.

5. DISCUSSION AND CONCLUDING REMARKS
Our analyses provide new evidence on the impact of large-
scale CCTs on child vaccination rates using a clustered-
randomized trial in Indonesia. After two years of implementa-
tion, results show that PKH increases child vaccination rates
for all vaccine types including BCG, OPV3, DPT3, HBV3,
and MCV among children aged <12 months by up to 11 per-
centage points or up to 30% compared to the control group
means. PKH also increases complete vaccination rates by 7.2
percentage points or 40% increase compared to the control
group mean. However, results show modest effects among chil-
dren aged 12–23 months. Our analyses also show that PKH is
equity enhancing by increasing child vaccination rates for
most vaccine types by up to 52% among children aged
<12 months living with less educated mothers (below six
years).
All this contributes to the currently inconclusive evidence in
the literature on the impact of CCTs on child vaccination rates
particularly at scale (Johri et al., 2015; Owusu-Addo & Cross,
2014; Ranganathan & Lagarde, 2012). This also contributes to
the previous evaluation that showed PKH did not increase
complete vaccination rates among children under three years
old (Alatas et al., 2011).
The more prominent PKH effects among children aged
<12 months might be because of the timing since all the vac-
cine types should be completed by the first year and also
because mothers might be more attentive to younger children
since they are more prone to illness compared to older ones.
Further, the PKH effects on OPV3 and MCV both among
children aged <12 months and 12–23 months might be
explained by the popularity of and public awareness toward
those vaccines among PKH field facilitators and beneficiaries
from massive national television advertisements since 1990 s.
Also, local policymakers are likely to align their efforts with
national/global agenda like MCV as a tracking indicator for
MDG 4.
Results also show that vaccination rates among children
aged <12 months raised substantially over the period of
2007 to 2009 for both treatment and control groups. For
instance, OPV3 rose from 23% and 26% in 2007 to 42% and
34% in 2009 in treatment and control groups, respectively.
While general health systems effects through routine vaccina-
tion programs (see Section 2) are likely to contribute to that
increase in both treatment and control groups, survey and
administrative data do not indicate any big push vaccine pro-
gram during the period. The national health survey (Riskes-
das) reported that the national MCV rates were 82% in 2007

Please cite this article in press as: Kusuma, D. et al. New Evidence on the Impact of Large-scale Conditional Cash Transfers on Child Vaccination Rates:
The Case of a Clustered-Randomized Trial in Indonesia, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.05.007
 NEW EVIDENCE ON THE IMPACT OF LARGE-SCALE CONDITIONAL 7

Table 5. Mean vaccination rates by treatment status and treatment effects by mother’s level of education
Year Mother education <6 years Mother education >6 years Difference
Mean Treatment effects Mean Treatment effects [3]–[7]
T C b SE T C b SE b SE
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10]
<12 months N = 1449 N = 2941 N = 4390
BCG 2007 0.39 0.41 0.136** (0.058) 0.46 0.49 0.122** (0.042) 0.005 (0.069)
2009 0.42 0.31 0.49 0.42
OPV3 2007 0.17 0.24 0.136** (0.054) 0.26 0.28 0.079* (0.041) 0.060 (0.064)
2009 0.36 0.26 0.44 0.39
DPT3 2007 0.14 0.19 0.101** (0.050) 0.21 0.23 0.074** (0.037) 0.030 (0.059)
2009 0.33 0.26 0.40 0.34
HBV3 2007 0.12 0.17 0.03 (0.049) 0.19 0.19 0.051 (0.034) 0.002 (0.056)
2009 0.24 0.20 0.31 0.26
MCV 2007 0.05 0.09 0.097* (0.055) 0.09 0.10 0.109** (0.041) 0.014 (0.060)
2009 0.37 0.29 0.47 0.38
Complete 2007 0.02 0.07 0.004 (0.037) 0.06 0.06 0.092** (0.029) 0.082* (0.044)
2009 0.16 0.16 0.26 0.19
12–23 months N = 1342 N = 2758 N = 4100
BCG 2007 0.43 0.35 0.03 (0.062) 0.46 0.47 0.027 (0.039) 0.020 (0.067)
2009 0.30 0.22 0.37 0.32
OPV3 2007 0.31 0.29 0.08 (0.052) 0.39 0.41 0.029 (0.037) 0.055 (0.059)
2009 0.26 0.18 0.33 0.29
DPT3 2007 0.33 0.28 0.002 (0.052) 0.40 0.39 0.017 (0.036) 0.002 (0.058)
2009 0.25 0.17 0.33 0.28
HBV3 2007 0.26 0.26 0.042 (0.055) 0.36 0.35 0.026 (0.037) 0.030 (0.062)
2009 0.23 0.16 0.27 0.21
MCV 2007 0.32 0.28 0.022 (0.055) 0.38 0.39 0.064* (0.037) 0.030 (0.062)
2009 0.28 0.20 0.35 0.28
Complete 2007 0.16 0.16 0.043 (0.040) 0.26 0.26 0.027 (0.034) 0.024 (0.049)
2009 0.16 0.11 0.21 0.17
Note: T = treatment, C = control, SE = standard errors. Vaccination status is based on health card record. Data include panel data of children aged
0–23 months divided into two cohort groups and followed for two years. Analyses use OLS regression; the treatment effects are b coefficient in equation
(1) for each subgroup. Difference uses triple-difference estimates by interacting the treatment effects with a dummy for whether the mother had less than
6 years of education (=1). No control variables included. Robust standard errors (SE) are in parentheses and clustered at the subdistrict level. *p < 0.1,
**
p < 0.05.
The bold and italics show statistically significant estimates.

and 74% in 2010 and the MOH data show the proportion of
village with universal coverage of immunization was 72% in
2007 and 70% in 2009 (MOH, 2007b, 2010b, 2009). The causal
evidence from our analyses shows that PKH significantly con-
tribute to that increase in vaccination rates particularly in the
treatment group by up to 30%. 3 And since PKH specifically
targeted poor households, our results provide policy options
to increase child vaccination among poor children who might
be difficult to reach by the routine health systems.
The relatively limited impact of PKH (e.g., among children
aged 12–23 months) might be due to implementation issues
such as payment delays and untimeliness, lack of condition-
ality enforcement, and inadequate supply-side, which might
limit its impact on child vaccination. Payment delays were
due to the lack of post offices to collect the cash or transfer
delays from central offices; payment was often not at the time
of need such as during childbirth (Febriany, Toyamah, Sodo,
& Budiyati, 2011). Lack of enforcement of conditionality and
cash penalty were due to slow establishment of MIS, which
made PKH relied on ‘‘perceived conditionality” from field
facilitators informing beneficiaries about the potential cash
penalty. The follow-up survey shows that 50% of sampled
households in the treatment areas said ever received PKH.
While this relatively low coverage rate could partly explain
the limited impact, such coverage rates might be due to
households not really being aware of the various poverty
programs they received namely BLT (unconditional cash
transfers) vs. PKH (Conditional Cash Transfers) vs. BOS
(school operational assistance). However, it is also possible
that the payments were delayed or not received at all, which
implies that our findings are potentially an underestimate of
potential effectiveness of PKH. In terms of inadequate
supply-side, our baseline data show that 19% of health cen-
ters in treatment areas experienced stock-outs of at least
one vaccine type the previous two months. In addition to
stock-outs, the PKH spot check also found inadequate num-
ber of midwives and lack of well-baby card (KMS) in some
areas (CHR-UI., 2010).
There are at least two policy implications. First, efforts
should be done to improve PKH effectiveness (e.g., solving
some of the implementation issues) particularly among chil-
dren aged 12–23 months since the effects are modest and their
vaccination rates are still low. The proportion of children aged
12–23 months with HBV3 was 63% and 58% in treatment and
control groups in 2009, respectively. Secondly, efforts should
be done to improve the recording of vaccination status into
health card (KMS/KIA) since there is still considerable cover-
age gap between vaccination rates based on health card record
and that based on mother’s response. Improved recording
should be useful for mother’s record (reminder), provider’s
record (tracking), and better evaluation by reducing the risk
of recall bias.

Please cite this article in press as: Kusuma, D. et al. New Evidence on the Impact of Large-scale Conditional Cash Transfers on Child Vaccination Rates:
The Case of a Clustered-Randomized Trial in Indonesia, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.05.007
8 WORLD DEVELOPMENT

Table 6. Mean vaccination rates by treatment status and treatment effects by urbanicity
Year Urban Rural Difference
Mean Treatment effects Mean Treatment effects [3]–[7]
T C b SE T C b SE b SE
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10]
<12 months N = 3288 N = 1102 N = 4390
BCG 2007 0.45 0.45 0.095** (0.037) 0.42 0.49 0.152** (0.072) 0.057 (0.080)
2009 0.46 0.37 0.50 0.39
OPV3 2007 0.24 0.27 0.097** (0.036) 0.21 0.24 0.116* (0.062) 0.018 (0.071)
2009 0.41 0.34 0.45 0.35
DPT3 2007 0.18 0.22 0.092** (0.034) 0.18 0.19 0.069 (0.053) 0.024 (0.063)
2009 0.37 0.31 0.41 0.33
HBV3 2007 0.17 0.19 0.078** (0.031) 0.16 0.15 0.016 (0.046) 0.062 (0.056)
2009 0.29 0.23 0.29 0.25
MCV 2007 0.08 0.10 0.102** (0.037) 0.08 0.09 0.107 (0.070) 0.005 (0.079)
2009 0.42 0.34 0.47 0.36
Complete 2007 0.04 0.07 0.075** (0.025) 0.05 0.05 0.05 (0.044) 0.025 (0.050)
2009 0.22 0.17 0.23 0.18

12–23 months N = 3224 N = 876 N = 4100

BCG 2007 0.44 0.42 0.043 (0.036) 0.47 0.48 0.055 (0.075) 0.012 (0.083)
2009 0.33 0.26 0.41 0.35
OPV3 2007 0.36 0.38 0.054 (0.033) 0.38 0.37 0.048 (0.065) 0.006 (0.072)
2009 0.29 0.24 0.37 0.29
DPT3 2007 0.37 0.34 0.031 (0.033) 0.39 0.38 0.008 (0.062) 0.024 (0.070)
2009 0.29 0.22 0.34 0.30
HBV3 2007 0.32 0.31 0.035 (0.033) 0.33 0.35 0.08 (0.063) 0.045 (0.071)
2009 0.24 0.19 0.30 0.22
MCV 2007 0.36 0.35 0.071** (0.034) 0.37 0.38 0.033 (0.068) 0.038 (0.075)
2009 0.31 0.24 0.36 0.32
Complete 2007 0.22 0.23 0.047 (0.030) 0.24 0.24 0.026 (0.057) 0.021 (0.064)
2009 0.19 0.14 0.22 0.18
Note: T = treatment, C = control, SE = standard errors. Vaccination status is based on health card record. Data include panel data of children aged 0–
23 months divided into two cohort groups and followed for two years. Analyses use OLS regression; the treatment effects are b coefficient in equation (1)
for each subgroup. Difference uses triple-difference estimates by interacting the treatment effects with a dummy for whether a child lived in urban (=1). No
control variables included. Robust standard errors (SE) are in parentheses and clustered at the subdistrict level. *p < 0.1, **p < 0.05.
The bold and italics show statistically significant estimates.

Our paper is limited in terms of subgroup analysis using
distance to health facility due to data limitation. Further
research should explore this to see if there is variation of
PKH effects by distance or supply readiness. Also, qualita-
tive studies could be proposed as a follow-up to understand
the barriers and facilitators for lack of impact among
children aged 12–23 months and low proportion of vaccina-
tion coverage based on health card record. Lastly, due to
implementation issues, our study is limited to estimate only
the effects from perceived conditionality (lack of enforce-
ment of compliance) rather than full-blown conditionality.
Future research should analyze the effects of different
conditionality types namely perceived conditionality vs.
partial (random sample check) conditionality vs. full-blown
conditionality. This is to provide policy options based on
a country’s capacity to implement.

NOTES

1. For quasi-experimental evidence, Attanasio, Gomez, Heredia, and
Vera-Hernandez (2005) used Propensity Score Matching (PSM) and found
that the Familias en Accion in Colombia increased DPT vaccination
coverage among children under two years old but not among those aged
2–4 years. Shei, Costa, Reis, and Ko (2014) also used PSM and found that
the Bolsa Familia in Brazil was associated with increased odds for child
vaccinations. Carvalho, Thacker, Gupta, and Salomon (2014) also
employed PSM and found that Janani Suraksha Yojana in India
increased the proportion of fully vaccinated children by 9.1 percentage
points.
2. As a robustness check, we use a dummy variable that accounts for
these other regular programs (PNPM Urban or PNPM Rural) in 2007
and 2009 when estimating PKH program effects and found similar
results.
3. Another reason is positive spillover effects. However, we were
not able to estimate the latter due to lack of administrative data
but we note that potential spillovers might limit the observed
effects.

Please cite this article in press as: Kusuma, D. et al. New Evidence on the Impact of Large-scale Conditional Cash Transfers on Child Vaccination Rates:
The Case of a Clustered-Randomized Trial in Indonesia, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.05.007
 NEW EVIDENCE ON THE IMPACT OF LARGE-SCALE CONDITIONAL
REFERENCES
Adato, M., Roopnaraine, T., & Becker, E. (2011). Understanding use of
health services in conditional cash transfer programs: Insights from
qualitative research in Latin America and Turkey. Social Science and
Medicine, 72(12), 1921–1929.
Alatas, V. et al. (2011). Program Keluarga Harapan: Main Findings from
the Impact Evaluation of Indonesia’s Pilot Household Conditional Cash
Transfer Program. World Bank.
Attanasio, O., Gomez, L. C., Heredia, P., & Vera-Hernandez, M. (2005).
The short-term impact of a conditional cash subsidy on child health and
nutrition in Colombia. London: The Institute of Fiscal Studies.
Ayala, F. (2010). Program Keluarga Harapan (PKH) Operational Assess-
ment Report. Jakarta: Prepared for the Government of Indonesia.
Barham, T. (2005). The impact of the Mexican conditional cash transfer on
immunization rates Unpublished manuscript. CA: Department of
Agriculture and Resource Economics, University of California at
Berkeley.
Barham, T., & Maluccio, J. A. (2009). Eradicating diseases: The effect of
conditional cash transfers on vaccination coverage in rural Nicaragua.
Journal of Health Economics, 28(3), 611–621.
Carvalho, N., Thacker, N., Gupta, S. S., & Salomon, J. A. (2014). More
evidence on the impact of India’s conditional cash transfer program,
Janani Suraksha Yojana: Quasi-experimental evaluation of the effects
on childhood immunization and other reproductive and child health
outcomes. PLoS One, 9(10), e109311.
CHR-UI (2010). Spot check Program Keluarga Harapan Conditional Cash
Transfers Program 2007–2009. Depok: Center for Health Research
University of Indonesia.
Febriany, V., Toyamah, N., Sodo, J., & Budiyati, S. (2011). Qualitative
impact study for PNPM-generasi and PKH on the provision and the
utilization of the maternal and child health services and basic education
services in the provinces of West Java and East Nusa Tenggara. Jakarta,
Indonesia: The SMERU Research Institute.
Fiszbein, A., Schady, N. R., & Ferreira, F. H. (2009). Conditional cash
transfers: Reducing present and future poverty. World Bank Publica-
tions.
Grossman, M. (2000). The human capital model. Handbook of Health
Economics, 1, 347–408.
Hicklin, G. (2008). Program Keluarga Harapan–PKH. Working Paper for
Pro-Poor Planning and Budgeting Project. Jakarta: National Planning
Agency (Bappenas).
Johri, M., Pérez, M. C., Arsenault, C., Sharma, J. K., Pai, N. P., Pahwa,
S., & Sylvestre, M. P. (2015). Strategies to increase the demand for
childhood vaccination in low-and middle-income countries: A system-
atic review and meta-analysis. Bulletin of the World Health Organiza-
tion, 93(5), 339–346.
Judarwanto, W. (2012). 20 Mitos Kampanye Hitam Anti Imunisasi.
Retrieved from <http://health.kompas.com/read/2012/05/17/
14501446/20.Mitos.Kampanye.Hitam.Anti.Imunisasi>.
Kremer, M., & Glennerster, R. (2012). Improving Health in Developing
Countries: Evidence from Randomized Evaluations. Handbook of Health
Economics. Elsevier.
Kusuma, D., Cohen, J., McConnell, M., & Berman, P. (2016). Can cash
transfers improve determinants of maternal mortality? Evidence from
the household and community programs in Indonesia. Social Science &
Medicine, 163, 10–20.
Available online at www.sciencedirect.com
ScienceDirect
Please cite this article in press as: Kusuma, D. et al. New Evidence on the Impact of Large-scale Conditional Cash Transfers on Child Vaccination Rates:
The Case of a Clustered-Randomized Trial in Indonesia, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.05.007
9
Madon, T., Hofman, K. J., & Glass, R. I. (2007). Implementation science.
Science Magazine.
MOH (2007a). Indonesia Health Profile. Jakarta: Ministry of Health.
MOH (2007b). Report on Riskesdas (National Basic Health Survey).
Jakarta: Ministry of Health.
MOH (2009). Indonesia Health Profile. Jakarta: Ministry of Health.
MOH (2010a). Gerakan Akselerasi Imunisasi Nasional Universal Child
Immunization 2010–2014. Jakarta: Ministry of Health.
MOH (2010b). Report on Riskesdas (National Basic Health Survey).
Jakarta: Ministry of Health.
Morris, S. S., Flores, R., Olinto, P., & Medina, J. M. (2004). Monetary
incentives in primary health care and effects on use and coverage of
preventive health care interventions in rural Honduras: Cluster
randomised trial. The Lancet, 364(9450), 2030–2037.
MOSA (2007). Pedoman Umum Program Keluarga Harapan (PKH
Guideline). Jakarta: Ministry of Social Affairs of Indonesia.
Olken, B. A., Onishi, J., & Wong, S. (2011). Indonesia’s PNPM generasi
program: Final impact evaluation report. World Bank.
Owusu-Addo, E., & Cross, R. (2014). The impact of conditional cash
transfers on child health in low-and middle-income countries: A
systematic review. International journal of public health, 59(4), 609–618.
Ranganathan, M., & Lagarde, M. (2012). Promoting healthy behaviours
and improving health outcomes in low and middle income countries: A
review of the impact of conditional cash transfer programmes.
Preventive Medicine, 55, S95–S105.
Robertson, L., Mushati, P., Eaton, J. W., Dumba, L., Mavise, G.,
Makoni, J., ... Garnett, G. P. (2013). Effects of unconditional and
conditional cash transfers on child health and development in
Zimbabwe: A cluster-randomised trial. The Lancet, 381(9874),
1283–1292.
Shei, A., Costa, F., Reis, M. G., & Ko, A. I. (2014). The impact of Brazil’s
Bolsa Famı́lia conditional cash transfer program on children’s health
care utilization and health outcomes. BMC International Health and
Human Rights, 14(1), 1.
Sparrow, R., Moeis, J. P., Damayanti, A., & Herawati, Y. (2008).
Conditional cash transfers in Indonesia: Baseline survey report. World
Bank: Program Keluarga Harapan and PNPM-Generasi.
StataCorp (2014). Stata Statistical Software: Release 14. College Station,
TX: StataCorp LP.
Utomo, B., Sucahya, P. K., & Utami, F. R. (2011). Priorities and realities:
Addressing the rich-poor gaps in health status and service access in
Indonesia. International Journal of Equity in Health, 10(47), 1–14.
WHO, UNICEF. (2016). Progress and Challenges with Achieving Universal
Immunization Coverage: 2015 Estimates of Immunization Coverage.
WHO/UNICEF Estimates of National Immunization Coverage (Data
as of July 2016).
APPENDIX A. SUPPLEMENTARY DATA
Supplementary data associated with this article can be
found, in the online version, at http://dx.doi.org/10.1016/
j.worlddev.2017.05.007.