Journal of Poverty

ISSN: 1087-5549 (Print) 1540-7608 (Online) Journal homepage: https://www.tandfonline.com/loi/wpov20

How Much Household Healthcare Expenditure

Contributes to Poverty? Evidence from the
Bangladesh Household Income and Expenditure
Survey, 2010

Azaher Molla & Chunhuei Chi

To cite this article: Azaher Molla & Chunhuei Chi (2020): How Much Household Healthcare
Expenditure Contributes to Poverty? Evidence from the Bangladesh Household Income and
Expenditure Survey, 2010, Journal of Poverty, DOI: 10.1080/10875549.2020.1742269

To link to this article: https://doi.org/10.1080/10875549.2020.1742269

Published online: 07 Apr 2020.

Submit your article to this journal

Article views: 3

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=wpov20
JOURNAL OF POVERTY
https://doi.org/10.1080/10875549.2020.1742269

How Much Household Healthcare Expenditure Contributes

to Poverty? Evidence from the Bangladesh Household
Income and Expenditure Survey, 2010
Azaher Mollaa and Chunhuei Chib
a
Department of Applied Health Sciences, Murray State University, Murray, Kentucky, USA; bCollege of
Health and Human Sciences, Oregon State University, Corvallis, Oregon, USA

ABSTRACT KEYWORDS
Bangladesh's health finance characterizes by 63.3% out-of-pocket Healthcare expenditure;
payments and a lack of prepayment mechanism. We assume this Poverty; Contributors of
heavy burden significantly contributes to poverty. The results indi- poverty; Bangladesh;
Household poverty
cate, 3.2% of households are not counted as the poor, but in reality,
they are. This shows an 8.8% underestimation of poverty. The
poverty gap increases from Bangladesh Taka 1,458 to Taka 1,817.
Similarly, the mean positive poverty gap rises from 22.2% to 25.2%,
indicating that the increase in poverty is not only due to more
households being brought into poverty but also deepening of the
poverty among those who were already poor.

Introduction
Ill-health and poverty are inter-related, and the causality runs in both directions
(Ainsworth & Over, 1999; Bloom et al., 1998). Sickness is often associated with
catastrophic health expenditure (CHE) (Narayan et al., 2000). CHE and impover-
ishment are much talked but less acted issue in healthcare finance. Financial risk
protection (FRP) for households and individuals is one of the main objectives of
health systems finance. Four indicators of FRP have been described in the literature:
incidence of CHE, mean positive catastrophic overshoot, the incidence of impover-
ishment, and increase in the depth of poverty due to high out-of-pocket (OOP)
healthcare spending (Saksena et al., 2014). Globally more than 150 million indivi-
duals in 44 million households face financial catastrophe annually because of direct
healthcare payments (WHO, 2005). One of the components of healthcare financing
is fairness in financial contribution (Xu et al., 2003). Bangladesh health systems
finance characterized by 63% OOP payments and a lack of a functional prepayment
system (MoHFW, 2015; Molla et al., 2017). High reliance on OOP payments
increases the financial burden of households (Amaya Lara & Ruiz Gomez, 2011;
O’Donnell et al., 2005; Wagstaff & Van Doorslaer, 2003; Xu et al., 2003). Not only in
Bangladesh, but also in most low and middle-income countries, OOP spending is
the major healthcare payment method.

CONTACT Azaher Molla [email protected] Department of Applied Health Sciences, Murray State
University, Murray, KY 42071
© 2020 Taylor & Francis Group, LLC
2 A. MOLLA AND C. CHI

The consequences of excess OOP expending are enormous with differ-

ent scenarios. Some households may not utilize formal healthcare at all due
to excess OOP expenditure, or they may receive partial care, and thus
aggravate the disease condition that causes the disease to become a chronic
status. Households may sell their movable and immovable properties to
manage the treatment costs, which in turn make them poorer. Due to
health expenditures, households may need to ration their food items, and
thus may become malnourished. Out-of-pocket health expenditure may
affect education; children may drop out of school. Moreover, OOP pay-
ments may mislead planners and policymakers to miscalculate poverty
status. The conventional calculation of poverty assesses the overall house-
hold expenditure. The latter includes all expenditure, including healthcare
payments, which underestimates poverty. It was estimated that 78 million
people in Asia are not currently included as poor despite the fact that their
per capita household expenditure net of spending on healthcare payments
falls below the extreme poverty line (Van Doorslaer et al., 2007).
Many works of literature are available on CHE and poverty (Amaya Lara &
Ruiz Gomez, 2011; Chuma & Maina, 2012; Tomini et al., 2012; Van Doorslaer
et al., 2006, 2007; Wagstaff & Van Doorslaer, 2003; Xu et al., 2003). In a study
among 89 countries, Xu et al. (2007) showed that the incidence of CHE was
estimated to 3.1, 1.8, and 0.6% in low, middle, and high-income countries,
respectively. In India, the CHE was estimated at 30% (Flores et al., 2008). In
Kenya, 1.48 million people were pushed into poverty due to CHE (Chuma &
Maina, 2012). In Uganda, 2.9% of households faced CHE in 2003 (Xu et al.,
2006). In 11 Asian countries, Van Doorslaer et al. (2006) showed that 78 million
people pushed into poverty due to high OOP payments.
A limited number of studies were conducted in Bangladesh in recent
years. Khan et al. (2017) conducted a study on a national representative
data set and showed that 14.2% of households in Bangladesh faced CHE
and 1.9% overshoot. The authors also showed that 16.5% of the poorest
and 9.2% of the richest households faced CHE. In another study in an
urban area, Rahman et al. (2013) observed that nearly 9% of all households
faced financial catastrophe. The findings of this study cannot be general-
ized for the population of Bangladesh. In a separate study, Hamid et al.
(2014) conducted a study on the rural population who were the members
of a micro-insurance project and showed that 3.4% of the households fell
into poverty annually due to high OOP healthcare spending. The findings
of the latter two studies cannot be generalized due to data limitations.
In this study, we investigated the impacts of CHE on poverty in
Bangladesh using the latest available national Household Income and
Expenditure Survey (BBS, 2011). The analysis may be useful for developing
specific policy actions for poverty alleviation related to healthcare financing
mechanisms in Bangladesh.
 JOURNAL OF POVERTY 3

Data and methods

The data came from Bangladesh Household Income and Expenditure Survey,
2010 (BBS, 2011). This is a national representative cross-sectional survey on
12,240 households consisting of 55,580 populations. A two-stage stratified
random sampling technique was followed in drawing the sample.
This study focuses on the empirical relationship between household health-
care expenditure and ability-to-pay. Under the distributive justice, we follow the
ability-to-pay principle developed by Wagstaff and Van Doorslaer (1992). For
data analysis, we have used Stata 14.0 and the Automated Development
Economics and Poverty Tables (ADePT) software, version 5.0 developed by
World Bank’s experts (Wagstaff et al., 2007).
We present the results both in the quantitative estimation and in graphics.
Both total and nonfood consumption were used to measure the incidence
and intensity of catastrophic payments and their impacts on poverty.
Assuming that the marginal utility of income is different for the rich and
the poor, we assess distribution sensitive measure of catastrophic payments.
A threshold level of 5% to 40% was used for both catastrophic payments and
poverty status. Poverty differences are shown among the quintiles and
between gross and net of health payments.

Impacts of OOP healthcare payments on poverty

A common practice of defining the poverty line involves calculating expen-
diture required to meet subsistence nutrition requirements (2,122 kcal/per-
son/day) and the addition of allowances for nonfood needs (Deaton, 1997).
The conventional method of measuring household poverty is to compare
household total consumption or nonfood consumption with a poverty line.
This measure is not sensitive to household health expenditure. As total
household resources are fixed and healthcare payments are largely nondis-
cretionary, measurement of household total expenditure gross of OOP may
be misleading. The reason is that, during illness, households tend to manage
money by any means, and it adds up to the total expenditure. As a result, the
household that diverts money from other sources spending it on the neces-
sities of healthcare, resulted in falling below the poverty line however, would
not be counted as poor. Similarly, when a household lives below the poverty
line but borrows money to cover healthcare expenses, it is not treated as
poor. The above scenarios underestimate poverty. Therefore, to estimate
poverty in the real term, researchers suggested counting household expendi-
ture net of OOP healthcare spending. Thus, the impoverishing effects of
household OOP expenditure are measured by the differences between pov-
erty estimates derived from household consumption gross and net of OOP
payments (Wagstaff & Van Doorslaer, 2003).
4 A. MOLLA AND C. CHI

Household
expenditure
per capita

Gross of OOP payment

parade
Net of OOP
payment parade
poverty line (PL)
C
A
B

H gross H net Cumulative proportion of population

ranked by hhold. expenditure p.c.

Figure 1. Pen’s parade for households’ gross and net of out-of-pocket health payments. Source:
O’ Donnell et al. (2008)

Measurement of poverty adjusted with health payments

When we consider gross household expenditure, OOP payments included in
expenditure shown in the following figure (Figure 1). The number of house-
hold exceeding the poverty line is the poverty headcount (H), which is the
proportion of household that is poor. It is measured by using Equation (1),
PN gross
i¼1 Si Pi
H gross
¼ PN (1)
i¼1 Si

Where H gross is the headcount household gross expenditure, Si is the size of

the household, and Pigross = 1 if xi < PL and zero otherwise, and N is the
number of households in the sample.
The poverty gap is calculated using Equation (2).
PN gross
i¼1 Si gi
Ggross
¼ PN (2)
i¼1 Si
gross
The net of health payments headcount is given by replacing Pi with Pinet ¼
1 if ðxi  Ti Þ < PL, and zero; otherwise, in Equation (1). In Equation (2), the
gross
poverty gap is given by replacing gi with ginet ¼ Pinet ðPLðxi  Ti ÞÞ.
This measure does not say anything about how much the household falls
below the poverty line. This is the poverty gap (G). When we consider
household expenditure net of OOP payments, then the line (Pen’s parade1)
shifts to the right, adding more households to headcount as well as adding
more poverty gap.
In Figure 1, poverty headcount (Hgross) is the number of household
exceeding the threshold, and the poverty gap is equal to the area A. When
OOP payments are subtracted from the household expenditure before
 JOURNAL OF POVERTY 5

poverty is assessed, then the headcount (H) rises to (Hnet), and the gap is
A + B + C. Therefore, the difference between OOP gross and OOP net is
Hnet – Hgross. The difference between poverty gaps is represented by the area
B + C. The poverty gap increases both because those already counted before
as the poor become poorer once health payments are netted out of household
resources (area B). There is an additional number of households that were
not counted as poor based on gross expenditures but become poor after
netted out OOP expenditures (area C) and considered it.

Results
Incidence and intensity of catastrophic payments
A threshold of 5 to 40% was used to measure the incidence and intensity of
catastrophic payments. When the threshold is raised from five to 10% of total
expenditure, the incidence of catastrophic payments in the lowest quintile falls
by half (30.2 vs. 15.0%), and the mean overshoot drops from 3.6% to 2.5%.
When we consider a 40.0% threshold, the incidence of catastrophic payments is
two percent. Incidence and intensity of catastrophic payments decrease by the
threshold. Unlike the incidence/headcount, the mean positive overshoot among
those exceeding the thresholds increases as the threshold raised. Those in the
lowest quintile at 10.0% threshold level spent, on average, 26.9% (threshold level
10.0% plus MPO 16.9%) of their total consumption in healthcare. While those
spending at 25.0% threshold level spent on average, 53.3% (threshold level
25.0% plus MPO 28.3%) of their total consumption in healthcare. The main
equity concern here is that the incidence, intensity, and mean positive over-
shoot are prevalent in all threshold (Table 1).
When catastrophic payments are defined with respect to nonfood expen-
diture, both incidence and intensity (overshoot) of catastrophic payments are
much higher (Table 1 vs. Table 2) than total consumption. An important
finding from this analysis is that when we assess the incidence of catastrophic
payments in healthcare against household total nonfood consumption, it is
higher than when it was assessed against household’s total consumption (as
a proxy measure of adjusted household income). In other words, poorer
households cut proportionately more nonfood expenditure to cope with the
health outlays (Table 2).

Measures of distribution-sensitive catastrophic payments

This part of the analysis focuses on catastrophic payment headcount and
overshoot to household income distribution. The concentration index of
the incidence of catastrophic payments (CE) is negative for all thresholds
(Table 3), which indicates that the poor exceed the thresholds. Similarly,
6 A. MOLLA AND C. CHI

Table 1. Incidence and intensity of catastrophic health payments using household consumption,
Bangladesh, 2010.
Threshold budget share
5% 10% 15% 25% 30% 40%
Headcount (H)
Lowest quintile 30.2 15.0 8.9 4.6 3.6 2.0
Standard error 1.17 0.91 0.72 0.51 0.46 0.34
2nd quintile 25.0 11.7 6.8 4.0 2.9 2.1
Standard error 1.06 0.75 0.59 0.47 0.40 0.33
3rd quintile 24.4 10.6 6.3 3.0 2.5 1.4
Standard error 1.04 0.71 0.55 0.37 0.34 0.23
4th quintile 20.5 8.2 4.8 2.0 1.2 0.6
Standard error 0.92 0.60 0.47 0.30 0.23 0.17
Highest quintile 12.3 5.4 2.9 1.0 0.7 0.4
Standard error 0.71 0.47 0.33 0.18 0.15 0.12
Total 22.5 10.2 5.9 2.9 2.2 1.3
Standard error 0.45 0.32 0.25 0.17 0.15 0.11
Overshoot (O)
Lowest quintile 3.6 2.5 1.9 1.3 1.1 0.8
Standard error 0.33 0.31 0.29 0.26 0.24 0.22
2nd quintile 3.3 2.5 2.0 1.5 1.4 1.1
Standard error 0.48 0.46 0.45 0.43 0.42 0.40
3rd quintile 3.3 2.5 2.1 1.6 1.5 1.3
Standard error 0.76 0.75 0.74 0.73 0.73 0.72
4th quintile 1.9 1.3 1.0 0.6 0.6 0.5
Standard error 0.27 0.26 0.25 0.23 0.22 0.21
Highest quintile 1.0 0.6 0.4 0.2 0.2 0.2
Standard error 0.11 0.10 0.09 0.08 0.07 0.07
Total 2.6 1.9 1.5 1.1 0.9 0.8
Standard error 0.20 0.19 0.19 0.18 0.18 0.18
Mean positive overshoot (MPO)
Lowest quintile 11.9 16.9 21.8 28.3 30.9 41.2
Standard error 1.00 1.82 2.77 4.49 5.45 8.03
2nd quintile 13.3 21.2 29.9 38.3 46.5 53.1
Standard error 1.83 3.70 6.05 9.69 12.68 16.61
3rd quintile 13.4 23.3 32.4 54.7 60.9 93.9
Standard error 3.06 6.92 11.48 23.94 28.89 49.85
4th quintile 9.2 15.4 19.9 32.1 45.8 76.9
Standard error 1.25 2.91 4.70 10.18 15.29 25.35
Highest quintile 8.5 11.8 15.1 24.5 29.2 33.6
Standard error 0.79 1.57 2.73 6.76 9.18 14.20
Total 11.7 18.4 24.9 36.7 43.4 59.0
Standard error 0.86 1.83 3.05 5.92 7.79 12.41

the concentration index of payment overshoot (CO) for all threshold are
also negative, indicating that the average payments exceed the thresholds
is greater among the worse off. Rank weighted headcount (HW) and
overshoot (OW) are commonly used to measure sensitivity to the distribu-
tion of income. In all cases, HW is greater than H, indicating that
catastrophic payments are more frequent among the poor. According to
social welfare interpretation, such catastrophic payments are worse than if
they were not at all related to income (Wagstaff & Van Doorslaer, 2003).
 JOURNAL OF POVERTY 7

Table 2. Incidence and intensity of catastrophic health payments, using nonfood consumption,
Bangladesh, 2010.
Threshold budget share
5% 10% 15% 25% 30% 40%
Headcount (H)
Lowest quintile 66.8 45.6 33.8 20.6 17.0 11.8
2nd quintile 58.1 37.9 25.9 15.7 12.3 8.0
3rd quintile 55.3 34.0 23.9 13.9 10.6 7.9
4th quintile 46.9 27.3 17.8 8.9 7.2 4.7
Highest quintile 29.1 13.8 8.7 5.0 3.6 2.1
Total 51.2 31.7 22.0 12.8 10.2 6.9
Overshoot (O)
Lowest quintile 17.8 15.0 13.1 10.4 9.5 8.1
2nd quintile 14.1 11.7 10.2 8.1 7.4 6.5
3rd quintile 12.9 10.8 9.3 7.5 6.9 6.0
4th quintile 8.6 6.8 5.7 4.5 4.1 3.5
Highest quintile 4.0 3.0 2.5 1.8 1.6 1.3
Total 11.5 9.5 8.1 6.5 5.9 5.1
Mean positive overshoot (MPO)
Lowest quintile 26.7 33.0 38.6 50.6 55.9 68.3
2nd quintile 24.2 31.0 39.2 51.7 60.4 80.6
3rd quintile 23.4 31.7 39.1 54.1 65.3 76.6
4th quintile 18.3 25.0 32.1 50.4 56.5 73.7
Highest quintile 13.7 21.7 28.3 36.3 44.0 63.2
Total 22.4 29.9 37.0 50.5 58.2 73.5

Table 3. Measures of distribution-sensitive catastrophic payments, Bangladesh, 2010.

Threshold budget share
5% 10% 15% 25% 30% 40%
Total expenditure
Concentration index (CE) −0.150 −0.190 −0.203 −0.269 −0.302 −0.306
Standard error 0.01 0.02 0.02 0.03 0.03 0.04
Rank-weighted headcount (HW) 25.848 12.107 7.151 3.717 2.841 1.717
Standard error 0.58 0.43 0.34 0.25 0.22 0.17
Concentration index (CO) −0.224 −0.244 −0.256 −0.268 −0.265 −0.257
Standard error 0.03 0.04 0.04 0.05 0.06 0.07
Rank-weighted overshoot (OW) 3.209 2.331 1.863 1.361 1.199 0.974
Standard error 0.26 0.25 0.24 0.23 0.23 0.22
Nonfood expenditure
Concentration index (CE) −0.143 −0.196 −0.224 −0.253 −0.267 −0.281
Rank-weighted headcount (HW) 58.554 37.918 26.966 16.059 12.872 8.853
Concentration index (CO) −0.251 −0.268 −0.277 −0.286 −0.289 −0.291
Rank-weighted overshoot (OW) 14.353 12.006 10.399 8.326 7.618 6.557

In general, the distribution of catastrophic payments depends on whether

health payments are expressed as a share of total expenditure or a share of
nonfood expenditure. In the former case, catastrophic payments fall, in all
six thresholds, with total expenditure. As a result, the rank weighted head-
count (HW) and overshoot (OW) are larger than the unweighted indexes in
Tables 2 and 3. When health payments are assessed relative to nonfood
expenditure, the concentration indexes are negative for all thresholds. It
8 A. MOLLA AND C. CHI

Table 4. Weighted vs. unweighted incidence and intensity of catastrophic payments, Bangladesh,
2010.
Total Expenditure
Incidence of catastrophic payments by thresholds
Thresholds 5% 10% 15% 25% 30% 40%
Unweighted (H) 22.5 10.2 5.9 2.9 2.2 1.3
Weighted (HW) 25.8 12.1 7.2 3.7 2.8 1.7
The overshoot of catastrophic payments by thresholds
Unweighted (H) 2.6 1.9 1.5 1.1 0.9 0.8
Weighted (HW) 3.2 2.3 1.9 1.4 1.2 1.0
Nonfood Expenditure
Incidence of catastrophic payments by thresholds
Unweighted (H) 51.2 31.7 22.0 12.8 10.2 6.9
Weighted (HW) 58.6 38.0 27.0 16.0 12.9 8.9
The overshoot of catastrophic payments by thresholds
Unweighted (H) 11.5 9.5 8.1 6.5 5.9 5.1
Weighted (HW) 14.4 12.0 10.4 8.3 7.6 6.6

indicates that households with low nonfood expenditures are more likely to
incur catastrophic payments. As a result, the weighted indexes are larger
than the unweighted indexes.
Table 4 shows that considering the economic status of the household, both
incidence and intensity/overshoot changes dramatically (22.5 vs. 25.8% and
2.6 vs. 3.2%). When we consider nonfood consumption, both incidence and
intensity vary drastically (51.2 vs. 58.6 and 11.5 vs. 14.4%).

Impacts of household healthcare expenditure on poverty

Poverty lines (PLs) are usually based on the consumption of basic needs
(CBN). Using the CBN method, calculation of PLs entails estimation of the
average level of per capita expenditure at which individuals can meet basic
food and nonfood needs. The upper poverty line consists of gross consump-
tion plus amount spent on nonfood consumption and uses the limit of
Bangladesh Tk.19, 813 per household per year (Titumir & Rahman, 2011).
The lower PL is assessed, adding gross consumption with the median amount
spent on nonfood consumption equivalent to Tk. 15,888. When the assess-
ment is based on total household consumption, 33.2% of the population is
estimated to be in poverty. If OOP payments for healthcare are netted out of
household consumption, this percentage rises to 36.4%. Therefore, 3.2% of
the population is not counted as living in poverty but would be considered
poor if spending on healthcare was discounted from household resources.
This represents a substantial increase of 10.0% in the estimation of poverty.
The estimated poverty gap (average deficit to reach the poverty level) also
increases by about 25.0% from Tk.1, 458 to Tk.1, 817. The normalized
poverty gap increases from 7.4% to 9.2% when health payments are netted
 JOURNAL OF POVERTY 9

Table 5. Measures of poverty based on consumption gross and net of spending on healthcare,
Bangladesh, 2010.
Difference
Gross of health Net of health
payments payments Absolute Relative
Poverty line = Upper poverty line
(Titumir & Rahman, 2011) (Tk. 19,813)
Poverty headcount 33.2% 36.4% 3.2% 9.6%
Poverty gap (Tk.) Tk.1,458 Tk.1,817 Tk.359 24.6%
Normalized poverty gap 7.4% 9.2% 1.8% 24.3%
Normalized mean positive poverty gap 22.2% 25.2% 3.0% 13.5%
Poverty line = Lower poverty line
(Tk. 15,888)
Poverty headcount 17.0% 20.0% 3.0% 17.6%
Poverty gap Tk.483.5 Tk.720.3 Tk.237 49.0%
Normalized poverty gap 3.0% 4.5% 1.5% 50.0%
Normalized mean positive poverty gap 17.9% 22.7% 1.3% 26.8%

out of household consumption. The normalized mean positive poverty gap

increases from 22.2% to 25.2%. This suggests that the rise in the poverty gap
is due to more households being brought in to poverty as well as the
deepening of poverty among those who were already poor (Table 5).
Similarly, for the lower poverty line, 3.0% of the population is not counted as
poor as the OOP includes household consumption. Here poverty gap increases
from Tk.484 to Tk.720, indicating a 49.0% increase. The normalized poverty gap
increases from 3.0 to 4.5%, predicting a 50.0% increase. The normalized mean
positive poverty gap increases from 17.9 to 22.7%. These findings conclude that,
even in the lower poverty line, the rise in the poverty gap is due to both an
increase in headcounts and pushing households into deeper poverty.
The relationship between health-payment budget share and a cumulative frac-
tion of households ranked by decreasing the value of household prepayment
budget share shows that the catastrophic payment headcount depends on the
budget share threshold. The sharper the curve, the less sensitive the headcounts
are to the choice of threshold. Both the curves are nearly right at just a 1-point
threshold. The total expenditure curve shows that a threshold of 5.0% leads to
a catastrophic payment headcount of 22.5%. When the threshold raised to 10.0%,
the proportion of households with catastrophic payments falls to 10.2%. Again,
when the threshold raised further 15.0 and 25.0%, the headcounts falls to 2.9 and
1.0%, respectively (Figure 2). There is also a sharp difference when we consider
total expenditure and nonfood expenditure.
This stylized version (Figure 3) of the Pen’s parade charts household total
consumption as a fraction of the cumulative proportion of households ranked
in ascending order of total consumption. When health payments produce re-
ranking in the income distribution, it is still possible to visualize the effect of
healthcare payments on the parade using the “paint drips” chart developed by
Wagstaff and Van Doorslaer (2003). It shows the Pen’s parade for household
10 A. MOLLA AND C. CHI

Health payments as % of household expenditure

1.4

1.2

0.8 OOP/total exp.

0.6
OOP/nonfood
0.4 exp.

0.2

0
0.0 0.1 1.0

Cummulative proportions of households ranked by decreasing health

payments budget share

Figure 2. Health payments share by total expenditure and nonfood expenditure, Bangladesh,
2010.

pre-OOP consumption
5
post-OOP consumption

0
Consumption as multiple of PL

-5

-10

-15

0 .2 .4 .6 .8 1
Cumulative proportion of population, ranked from poorest to richest

Figure 3. Pen’s parade of the household consumption gross and net of health payments,
Bangladesh, 2010.

consumption gross of health payments. For each household, the vertical bar or
“paint drips” shows the extent to which health payments reduce consumption. In
other words, households become impoverished by health payments. The graph
shows that health payments are larger at higher values of consumption, but in the
 JOURNAL OF POVERTY 11

lower and middle part, 80.0% of households brought below the poverty line by
health payments.

Discussions
Although strong, our findings should be interpreted with caution. As usual,
for all survey data sources, this study may suffer from recall bias. Household
income and expenditure were mainly collected for the last 12 months. Food
consumption data were collected on a daily and weekly basis, and non-food
consumption data were collected on monthly and yearly consumption.
Using the household as a unit of analysis does not account for any
complexity or diversity within families. In addition, the data contain
a significant number of zeros OOP payments. This might be caused by
poor households not being able to use health services and could not make
any payments at all, which needs further research.
Longitudinal data are ideally used for the analysis of the effects of
catastrophic illness shocks. This would allow researchers to assess how
spending on non-medical goods and services changes due to health shocks
(Gertler & Gruber, 2002; Wagstaff, 2004). As our data is cross-sectional,
we made an approximation of material disruption following catastrophic
payments. This approach is well accepted in the literature (O’ Donnell
et al., 2008). Households’ catastrophic payments may have two types of
impacts; short term – when financed by cutting back on current consump-
tion and long term- when the expenditure is financed through savings, sale
of assets, or credits. To identify long-run impacts is beyond the scope of
this study. With cross-sectional data, we could not distinguish between
short-term and long-term impacts of catastrophic payments and subse-
quent poverty status.
The choice of the threshold of catastrophic payments is obviously
a normative value judgment, and we do not impose our judgment. Rather,
in our analysis, we have used a threshold from 5% to 40%. It is noticeable
that incidence, overshoot, and mean positive overshoot are prevalent in all
levels of the threshold. This may be a concern for policymakers who are
interested in improving equity in healthcare financial burden. Both incidence
and intensity decrease with income/consumption, because the richer house-
holds spent a lesser budget share on healthcare.
We identified only those households that incurred medical expenses. What
was left out in this study are those who did not use healthcare due to the
unavailability of funds for the minimum payments or lack of access? These
groups of people face subsequent deterioration of health and may have
a greater welfare loss than those incurring catastrophic payments. To avoid
these difficulties, Pradhan and Prescott (2002) suggested measuring exposure
to, rather than incurrence of, catastrophic payments.
12 A. MOLLA AND C. CHI

Earning losses is another issue of concern for equity in healthcare financial

burden, besides catastrophic payments. Lost earnings have long-term eco-
nomic consequences and loss of welfare. In Indonesia, earning losses were
more important in disrupting household living standards than medical
spending (Gertler & Gruber, 2002).
Researchers debated on the issue of which denominator – income, expenditure,
or consumption would be the best for measuring catastrophic payments. Income
is not directly corresponding to medical spending. On the other hand, the health
payments-to-income ratio is also not responsive to the means of healthcare
financing. When household expenditure is used as the denominator, the cata-
strophic payments are defined in relation to the health payments budget share.
There is a problem with this measurement. That is, the budget share may be low
for the poor in low-income households, and high for the high-income households.
For the poor, most resources are absorbed by essential items like foods, leaving
little to spend on healthcare. Thus, researchers proposed to define catastrophic
payments with respect to health payments as a share of nondiscretionary expen-
diture or expenditure net of spending on necessities. The definition of nondiscre-
tionary also generated some debates; what is discretionary for the rich may be
nondiscretionary for the poor. Therefore, nonfood expenditure may be more
capable of distinguishing between the rich and the poor than total expenditure.
When we consider diminishing the marginal utility of income, the opportu-
nity costs of health spending by the poor is always greater than the rich are.
Therefore, it is essential that opportunity cost should be weighted differently for
the poor and the rich. This can be measured by using concentration indexes of
headcount (Hw) and overshoot (OW). A positive value indicates a greater
tendency for the better off to exceed the payment threshold. On the other
hand, a negative value indicates a greater tendency of worse-off to exceed the
threshold. Wagstaff and Van Doorslaer (2003) recommended adjusting both the
headcount and the overshoot by multiplying each measure by its compliment.

Conclusions
This study provides a comprehensive estimation and understanding of the distribu-
tion of healthcare financial burden in Bangladesh. The incidence and intensity of
catastrophic payments delineate how healthcare expenditure contributes to poverty.
Our findings suggest that the incidence and intensity of catastrophic payments are
highly prevalent in Bangladesh health systems finance, which is concordant with
Van Doorslaer (2007). The poorest quintile bears the most. It is recommended to
devise alternative healthcare financing, including new pooling mechanisms like
social, community, and private insurance through taxation be devised.
The distribution of catastrophic payments clearly depends on whether
payments are expressed as a share of total expenditure or of nonfood
expenditure. This explains the fact that OOP payments budget share rise
 JOURNAL OF POVERTY 13

with total household resources (Van Doorslaer et al., 2007). Consequently,

the rank weighted headcount and overshoot are smaller than unweighted.
Therefore, nonfood expenditure is treated as the gold standard for assessing
household ability-to-pay, and subsequently to measure incidence and inten-
sity of catastrophic payments.
In 2010 purchasing power parity (PPP), using the cost of basic needs (CBN)
method, the lower and upper poverty line was fixed. In Bangladesh, 17.0% of the
population is estimated to be in extreme poverty. If OOP payments are netted
out of household consumption, this percentage rises to 20.0%. Therefore, using
the conventional method, 3.0% of the population is not counted as living in
extreme poverty. All the poverty measures, including the poverty gap, normal-
ized the poverty gap, and normalized mean positive poverty gap, rise from total
expenditure to nonfood expenditure. This rise in the poverty gap is due to more
households being brought into poverty as well as deepening the poverty of
households that are already poor.
We hope these findings may benefit the country in designing new health
policies aiming for the financial protection of households. The results of the
study may be applicable for the low and middle countries with similar socio-
economic conditions in formulating the health policy to lessen the burden of
OOP payments. Further, the findings of this study may stimulate future
researches on how to minimize the burden of healthcare expenditure.
Research on the financial burden of healthcare costs has to gain more attention
among researchers and policymakers around the world who are concerned about
equity in health systems financing. One indication is that on 15 November 2016, the
Interagency Expert Group on Sustainable Development Goals (IAEG-SDG) in
Geneva has revised the measuring indicator 3.8.2: Coverage by Health Insurance
or a Public Health System per 1,000 populations, to Proportion of Population with
Large Household Expenditure on Health, as a Total Share of Household Expenditure
or Consumption (McIntyre et al., 2016). Future research along this topic may
include short term and long-term impacts of catastrophic payments. Further, to
assess the healthcare financial burden more comprehensively, we also need more
research on the healthcare financial burden of those who cannot afford to use any
healthcare services.

Note
1. Jan Pen, a Dutch economist described pen’s parade or the income parade in 1971. The
parade is defined as a succession of people in the economy, with their height propor-
tional to their income, and ordered from the lowest to the greatest.

Disclosure statement
No potential conflict of interest was reported by the authors.
14 A. MOLLA AND C. CHI

References
Ainsworth, M., & Over, A. M. (1999). Confronting AIDS: Public priorities in a global epidemic.
World Bank Publications.
Amaya Lara, J. L., & Ruiz Gomez, F. (2011). Determining factors of catastrophic health
spending in Bogota, Colombia. International Journal of Health Care Finance and
Economics, 11(2), 83–100. https://doi.org/10.1007/s10754-011-9089-3
BBS. (2011). Household income and expenditure survey 2010. Bangladesh Bureau of Statistics.
Bloom, D. E., Sachs, J. D., Collier, P., & Udry, C. (1998). Geography, demography, and
economic growth in Africa. Brookings Papers on Economic Activity, 1998(2), 207–295.
https://doi.org/10.2307/2534695
Chuma, J., & Maina, T. (2012). Catastrophic health care spending and impoverishment in Kenya.
BMC Health Services Research, 12(1), 413. https://doi.org/10.1186/1472-6963-12-413
Deaton, A. (1997). The analysis of household surveys: A microeconometric approach to devel-
opment policy. World Bank Publications.
Flores, G., Krishnakumar, J., O’Donnell, O., & Van Doorslaer, E. (2008). Coping with
health-care costs: Implications for the measurement of catastrophic expenditures and
poverty. Health Economics, 17(12), 1393–1412. https://doi.org/10.1002/hec.v17:12
Gertler, P., & Gruber, J. (2002, March). Insuring consumption against illness. The American
Economic Review, 92(1), 51–70. https://doi.org/10.1257/000282802760015603
Hamid, S. A., Ahsan, S. M., & Begum, A. (2014, August). Disease-specific impoverishment impact
of out-of-pocket payments for health care: Evidence from rural Bangladesh. Applied Health
Economics and Health Policy, 12(4), 421–433. https://doi.org/10.1007/s40258-014-0100-2
Khan, J. A., Ahmed, S., & Evans, T. G. (2017). Catastrophic healthcare expenditure and
poverty related to out-of-pocket payments for healthcare in Bangladesh—an estimation of
financial risk protection of universal health coverage. Health Policy and Planning, 32(8),
1102–1110. https://doi.org/10.1093/heapol/czx048
McIntyre, D., McKee, M., Balabanova, D., Atim, C., Reddy, K. S., & Patcharanarumol, W.
(2016). Open letter on the SDGs: A robust measure for universal health coverage is
essential. The Lancet, 388(10062), 2871–2872. https://doi.org/10.1016/S0140-6736(16)
32189-4
MoHFW. (2015). Bangladesh national health accounts 1997–2012. Ministry of Health and
Family Welfare, Government of Bangladesh.
Molla, A. A., Chi, C., & Mondaca, A. L. (2017, January). Predictors of high out-of-pocket healthcare
expenditure: An analysis using Bangladesh household income and expenditure survey, 2010.
BMC Health Services Research, 17(1), 94. https://doi.org/10.1186/s12913-017-2047-0
Narayan, D., Patel, R., Schafft, K., Rademacher, A., & Koch-Schulte, S. (2000). Can anyone
hear us? Voices of the poor. The World Bank.
O’ Donnell, O., Van Doorslaer, E., Rannan-Eliya, R. F., Somanathan, A., Adhikari, S. R.,
Akkazieva, B., Harbianto, D., Garg, C. C., Hanvoravongchai, P., Herrin, A. N., Huq, M. N.,
Ibragimova, S., Karan, A., Kwon, S.-M., Leung, G. M., Lu, J. F. R., Ohkusa, Y., Pande, B. R.,
Racelis, R., Tin, K., & Zhao, Y. (2008). Who pays for health care in Asia? Journal of Health
Economics, 27(2008), 460–475. https://doi.org/10.1016/j.jhealeco.2007.08.005
O’Donnell, O., Van Doorslaer, E., Rannan-Eliya, R. F., Somanathan, A., Adhikari, S. R.,
Akkazieva, B., Harbianto, D., Garg, C. C., Hanvoravongchai, P., Herrin, A. N., Huq, M. N.,
Ibragimova, S., Karan, A., Kwon, S.-M., Leung, G. M., Lu, J. F. R., Ohkusa, Y., Pande, B. R.,
Racelis, R., Tin, K., & Zhao, Y. (2005). Explaining the incidence of catastrophic expendi-
tures on health care: Comparative evidence from Asia. Working paper no 5.. Mimeo.
Pradhan, M., & Prescott, N. (2002, July). Social risk management options for medical care in
Indonesia. Health Economics, 11(5), 431–446. https://doi.org/10.1002/()1099-1050
 JOURNAL OF POVERTY 15

Rahman, M. M., Gilmour, S., Saito, E., Sultana, P., & Shibuya, K. (2013). Health-related
financial catastrophe, inequality and chronic illness in Bangladesh. PLoS One, 8(2), e56873.
https://doi.org/10.1371/journal.pone.0056873
Saksena, P., Hsu, J., & Evans, D. (2014). Financial risk protection and universal health
coverage: Evidence and measurement challenges. PLoS Medicine, 11(9), e1001701.
https://doi.org/10.1371/journal.pmed.1001701
Titumir, R. A., & Rahman, K. M. (2011). Poverty and inequality in Bangladesh. Unnayan
Onneshan.
Tomini, S. M., Packard, T. G., & Tomini, F. (2012). Catastrophic and impoverishing effects of
out-of-pocket payments for health care in Albania: Evidence from Albania living standards
measurement surveys 2002, 2005 and 2008. Health Policy and Planning, 28(4), 419–428.
https://doi.org/10.1093/heapol/czs073
Van Doorslaer, E., O’Donnell, O., Rannan-Eliya, R. P., Somanathan, A., Adhikari, S. R.,
Garg, C. C., Harbianto, D., Herrin, A. N., Huq, M. N., Ibragimova, S., & Karan, A. (2006).
Effect of payments for health care on poverty estimates in 11 countries in Asia: An analysis
of household survey data. Lancet, 368(9544), 1357–1364. https://doi.org/10.1016/S0140-
6736(06)69560-3
Van Doorslaer, E., O’Donnell, O., Rannan-Eliya, R. P., Somanathan, A., Adhikari, S. R.,
Garg, C. C., Harbianto, D., Herrin, A. N., Huq, M. N., Ibragimova, S., & Karan, A. (2007,
November). Catastrophic payments for health care in Asia. Health Economics, 16(11),
1159–1184. https://doi.org/10.1002/hec.1209
Wagstaff, A. (2004). The Millennium development goals for health: Rising to the challenges.
World Bank Publications.
Wagstaff, A., O’Donnell, O., Van Doorslaer, E., & Lindelow, M. (2007, November 2).
Analyzing health equity using household survey data: A guide to techniques and their
implementation. World Bank Publications.
Wagstaff, A., & Van Doorslaer, E. (1992). Equity in the finance of health care: Some
international comparisons. Journal of health economics, 11(4), 361–387.
Wagstaff, A., & Van Doorslaer, E. (2003). Catastrophe and impoverishment in paying for
health care: With applications to Vietnam 1993–1998. Health Economics, 12(11), 921–934.
https://doi.org/10.1002/()1099-1050
WHO. (2005). Sustainable health financing, universal coverage and social health insurance.
Xu, K., Evans, D. B., Carrin, G., Aguilar-Rivera, A. M., Musgrove, P., & Evans, T. (2007).
Protecting households from catastrophic health spending. Health Affairs (Project Hope), 26
(4), 972–983. https://doi.org/10.1377/hlthaff.26.4.972
Xu, K., Evans, D. B., Kadama, P., Nabyonga, J., Ogwal, P. O., Nabukhonzo, P., &
Aguilar, A. M. (2006). Understanding the impact of eliminating user fees: Utilization
and catastrophic health expenditures in Uganda. Social Science & Medicine, 62(4),
866–876. https://doi.org/10.1016/j.socscimed.2005.07.004
Xu, K., Evans, D. B., Kawabata, K., Zeramdini, R., Klavus, J., & Murray, C. J. (2003).
Household catastrophic health expenditure: A multicountry analysis. Lancet, 362(9378),
111–117. https://doi.org/10.1016/S0140-6736(03)13861-5