Published by Oxford University Press in association with The London School of Hygiene and Tropical Medicine Health Policy

cy and Planning 2009;24:116–128

ß The Author 2008; all rights reserved. Advance Access publication 17 December 2008 doi:10.1093/heapol/czn046

Reducing out-of-pocket expenditures to

reduce poverty: a disaggregated analysis
at rural-urban and state level in India
Charu C Garg1* and Anup K Karan2

Accepted 11 November 2008

Downloaded from http://heapol.oxfordjournals.org/ at University of Alabama at Birmingham on December 11, 2012

Out-of-pocket (OOP) expenditure on health care has significant implications for
poverty in many developing countries. This paper aims to assess the differential
impact of OOP expenditure and its components, such as expenditure on
inpatient care, outpatient care and on drugs, across different income quintiles,
between developed and less developed regions in India. It also attempts to
measure poverty at disaggregated rural-urban and state levels.
Based on Consumer Expenditure Survey (CES) data from the National Sample
Survey (NSS), conducted in 1999–2000, the share of households’ expenditure on
health services and drugs was calculated. The number of individuals below the
state-specific rural and urban poverty line in 17 major states, with and without
netting out OOP expenditure, was determined. This also enabled the calculation
of the poverty gap or poverty deepening in each region.

Estimates show that OOP expenditure is about 5% of total household expen-

diture (ranging from about 2% in Assam to almost 7% in Kerala) with a higher
proportion being recorded in rural areas and affluent states. Purchase of drugs
constitutes 70% of the total OOP expenditure. Approximately 32.5 million per-
sons fell below the poverty line in 1999–2000 through OOP payments, implying
that the overall poverty increase after accounting for OOP expenditure is 3.2%
(as against a rise of 2.2% shown in earlier literature). Also, the poverty head-
count increase and poverty deepening is much higher in poorer states and rural
areas compared with affluent states and urban areas, except in the case of
Maharashtra. High OOP payment share in total health expenditures did not
always imply a high poverty headcount; state-specific economic and social
factors played a role.

The paper argues for better methods of capturing drugs expenditure in house-
hold surveys and recommends that special attention be paid to expenditures
on drugs, in particular for the poor. Targeted policies in just five poor states
to reduce OOP expenditure could help to prevent almost 60% of the poverty
headcount increase through OOP payments.
Keywords Out-of-pocket expenditures, impoverishment, household surveys, equity, India

1
Health economist, World Health Organization (WHO), Geneva.
2
Takemi Fellow, Harvard School of Public Health, USA.
* Corresponding author. World Health Organization, 1211 Geneva 27,
Switzerland. E-mail: gargc@who.int; cgarg@hotmail.com; cgarg@
worldbank.com

116
 REDUCING OUT-OF-POCKET EXPENDITURES TO REDUCE POVERTY 117

KEY MESSAGES

Expenditure on drugs was found to constitute the major part (70%) of out-of-pocket (OOP) health care expenditure
in India.

Both the increase in the number of poor as a result of OOP expenditure, and poverty deepening, were higher in rural
areas and poorer states than in urban areas and wealthier states.

Policymakers need to target specific areas and specific populations in certain states where the poverty impact of OOP
payments is greatest. Targeted policies in just five poor states to reduce OOP expenditure could help to prevent almost
60% of the increase in poverty headcount due to OOP payments.

Downloaded from http://heapol.oxfordjournals.org/ at University of Alabama at Birmingham on December 11, 2012

Introduction finds that three-quarters of total OOP health expenditure is
spent on drugs.
The Millennium Development Goals (MDGs) have put health
While the high share of OOP payments in overall health
at the heart of the development agenda, with three out of
financing caught the attention of policymakers in India dur-
eight Goals directly related to improvement in health status.
ing the early 1990s, the impact of OOP health expenditure on
These goals and targets emphasize the importance of health
households has been studied only recently (Peters et al. 2002;
as a dimension of poverty. This paper examines one of the
Krishna 2004; Garg and Karan 2006; NCMH 2005; Van Doorslaer
financing dimensions of health—out-of-pocket (OOP) expen-
et al. 2006). For example, while analysing health expenditure
diture—and shows how large OOP expenditures exacerbate
and utilization using National Sample Survey (NSS) data for
poverty in India. 1995–96, Peters et al. (2002) showed that the deduction of OOP
OOP spending is an inefficient way of financing health care.
payments from household expenditures lowered the national
It can have a negative impact on equity and can increase the poverty line by 2.2%, i.e. 2.2% of the population fell into poverty
risk of vulnerable groups slipping into poverty. Several studies because of OOP payments. Further, they pointed out that a
have documented the consequences of a high share of OOP quarter of hospital patients were impoverished by the cost of OOP
payments in total health financing in developing countries payments due to hospitalizations, and there were high levels of
(Berki 1986; Peters et al. 2002; Wagstaff and Van Doorslaer borrowing and selling of assets to make these payments. In
2003; Krishna 2004; Russell 2004; van Doorslaer et al. 2006), another study, van Doorslaer et al. (2006) also highlight that
with a higher poverty incidence and a larger proportion of OOP payments alone forced more than 37 million people in
households facing catastrophic expenditures (Xu et al. 2003; India below the $13 poverty line in 1999–2000.
O’Donnell et al. 2007). Overall, payments for health care seem progressive in India,
implying that the rich pay a higher proportion of the total
expenditure on health care than the poor. This is true for
The Indian perspective both taxation-based health finance as well as OOP payments
In India, health expenditure accounts for less than 5% of the (Mahal et al. 2001; Peters et al. 2002; Mahal 2003; O’Donnell
Gross Domestic Product (GDP), with OOP payments constitut- et al. 2007). What needs to be examined here is how the
ing the single largest component of total health expenditure.1 progressivity of OOP payments can be justified with wide
Estimates for OOP health expenditure between 1995–96 and inequalities in living standards across population groups and
2000–01 vary from 80% (Peters et al. 2002) to about 70% of geographical regions. With such large variations in the socio-
total health expenditure (NCMH 2005; WHO 2006). While economic conditions across states and between rural and urban
these proportions are much higher than the 25–50% range settings, it is imperative to assess the magnitude and impact
in several developing and developed countries,2 they vary of OOP expenditure at the disaggregated level, and to observe
widely across Indian states (Garg 2001a, b; Government of its impact on the poor.
India 2005). Further, OOP payments constitute 95% of private
health expenditures, with a weak insurance system and other
Objectives
community-based financing still only emerging (Ahuja 2004;
This paper argues that despite OOP payments being progressive
Garg 2006; Government of India 2005; Mahal et al. 2005). As a
at an all-India level, they have differential impacts across rural
result, OOP health expenditure also forms a substantial pro-
and urban settings, and across Indian states. Based on data
portion of total household spending, constraining expenditure
from the Consumer Expenditure Survey (CES) in 1999–2000
on necessities, and leading to a loss of welfare both at micro
(Government of India 2001), the paper explores the compo-
(household) and at macro (national) levels (Russel 2004).
nents of OOP payments on health care. More specifically, the
OOP health expenditure comprises various types of payments
paper analyses at decentralized levels:
made by households on health care. Recent studies show that
expenditure on medicines accounts for a substantial proportion.
the magnitude and distribution of OOP payments;
The report of the National Commission for Macro-economics
the components that trigger OOP payments; and
and Health (NCMH 2005, p.64) notes that ‘drugs are one of the
the incidence and intensity of poverty that occurs because
three cost drivers of the health care system’. Sakthivel (2005) of OOP payments.
118 HEALTH POLICY AND PLANNING

Apart from providing an in-depth analysis of OOP payments health care payments and that after payments (Wagstaff and
and their impact on the living standards of households, the Doorslaer 2003). It is measured by comparing both the preva-
paper also critically examines and updates earlier results. The lence (headcount ratio) and the intensity of poverty before and
present analysis, however, does not address the impact of OOP after OOP health payments. The pre-OOP (or pre-payment)
payments on the quantity and quality of health care utilized. poverty headcount is calculated by comparing households’
Neither does it capture other potential effects of illness or consumption expenditure gross of payments for health care
disability, which may include direct and/or indirect loss of with a poverty line as defined by the Planning Commission of
income, forgone consumption of needed health care, etc. India (2001). The pre-payment headcount of poverty (or pre
Furthermore, the analysis is based on data for a single year Hp, which has also been the basis for calculating the poverty
which does not allow the investigation and dynamics of headcount by the Planning Commission) can be measured as:
whether health shocks can be absorbed via consumption
pre HP ¼ 1=n  1 ðxi  PLÞ ð1Þ
across several periods, through borrowing or dis-saving
opportunities. where, xi is per capita consumption expenditure (in Rupees),

Downloaded from http://heapol.oxfordjournals.org/ at University of Alabama at Birmingham on December 11, 2012

The paper is organized as follows. The next section presents PL is the poverty line (in Rupees) and n is the number of
the data and methods. This is followed by the findings, starting individuals.
with all-India figures and continuing with state level results. Similarly, the post-OOP payment ‘poverty headcount’ is
These findings are then discussed. We finish with conclusions computed by netting out health care payments (measured by
and recommendations for policy interventions. actual OOP payments for all households) from households’
consumption expenditure and then comparing with the pov-
erty line, i.e.
Methods post HP ¼ 1=n  1 ððxi  OOPÞ  PLÞ ð2Þ
The paper is based on households’ consumption expendi-
Similarly, intensity of poverty, also known as poverty deep-
ture data collected in India by the National Sample Survey
ening, is measured by calculating the average ‘poverty gap’ as
Organisation (NSSO)4 for the year 1999–2000. This is the
defined by:
latest available large CES5 with a sample of more than 120 000
households (71 000 rural and 49 000 urban). Since consump- pre G ¼ 1=n Pi ðPL  xi Þ ð3Þ
tion expenditure data are collected at household level, the
analysis in this paper is based on household expenditure and
rather than individual expenditure. post G ¼ 1=n Pi ðPL  ðxi  OOPÞÞ ð4Þ
In the CES, the NSSO collects data on household expen-
diture on a wide range of items, including expenditure on where Pi ¼ 1 if xi  PL and is zero otherwise.
health services and commodities for institutional and non- OOP being positive, equation (2) results in a higher head-
institutional care. The recall period in the surveys is ‘last one count ratio and greater number of individuals below the
year’ for expenditure on institutional care and ‘last 30 days’ poverty line (PL) compared with that for equation (1).
for expenditure on non-institutional care. Health expendi- The additional number of individuals moving below the
tures for institutional and non-institutional care are recorded poverty line because of OOP expenditures is provided by:
separately under: HP ¼ post HP  pre HP
(a) purchase of drugs and medicines;
Similarly, the ‘average poverty gap’, or poverty deepening
(b) expenditure incurred on clinical tests such as pathological
in terms of the average amount by which people go below
tests, ECG, X-ray etc.;
the poverty line because of OOP expenditures, is measured by:
(c) professional fees of doctors, nurses etc.;
(d) payments made to hospitals and nursing homes for G ¼ post G  pre G
medical treatment;
(e) family planning appliances including IUD (intra-uterine Lastly, to facilitate comparison of poverty gaps computed for
device), oral pills, condoms, diaphragm, spermicide, etc.; different poverty lines (across different states and regions), it
and is useful to express the mean gap as a multiple of the poverty
(f) ‘other health expenditures’ not recorded above. line.6 This is known as the normalized poverty gap, NG ¼ G/PL.
The gap is also standardized with the headcount and this,
The available data at the unit level are added for all these known as the mean positive poverty gap, MPG ¼ G/Hp, depicts
items to obtain total expenditures for institutional and non- the average consumption shortfall because of OOP payments
institutional care. Drugs and medicines expenditures are for the poor.
subtracted from each of the institutional and non-institutional Since the NSSO bases its data on sample surveys, we cal-
totals to provide total expenditures on inpatient and out- culate all our estimates by applying the inbuilt weighting
patient care services, respectively. Adding together the drugs system of the NSSO. However, even after applying the inbuilt
and medicines expenditures under the institutional and non- weights, the total population estimated by the NSSO is usually
institutional categories provides the total expenditure on drugs an underestimate for a particular year. For example, for the
and medicines. year 1999–2000, the NSSO report estimates approximately 7%
The poverty impact of OOP payments is defined as the less population when compared with the interpolated Census
difference between the average level of poverty before data for the same year. The poverty headcounts both pre- and
 REDUCING OUT-OF-POCKET EXPENDITURES TO REDUCE POVERTY 119

post-OOP payments in this paper have been calculated by areas, and the opposite for rural areas, is noteworthy. The gap
adjusting with the interpolated Census population data for between the Lorenz curve and the curve for OOP payment share
1999–2000. indicates the relative inequality of OOP expenditure in comparison
with the inequality in household consumption expenditure, also
denoted as the ability to pay (ATP). Hence, the respective gaps
between the Lorenz curve and the curve for OOP payment share
Key findings in rural and urban areas suggest that OOP expenditure is highly
National level
OOP expenditure estimates
Approximately 30% of all households do not report OOP
expenditure, with a slightly lower proportion for urban areas.
Average monthly per capita OOP payment in 1999–2000 was

Downloaded from http://heapol.oxfordjournals.org/ at University of Alabama at Birmingham on December 11, 2012

Rs. 33 ($0.8), with Rs. 43.3 ($1) in urban areas and Rs. 29.6
($0.7) in rural India. Those in the richest 20% of the popula-
tion spent on average 10 times more than those in the poorest
20% (Rs. 100 in rural and Rs. 133 in urban areas compared
with Rs. 9 in rural and Rs. 14 in urban areas, respectively)
(Table 1).
OOP payments constitute 4.8% of total consumption expen-
diture, which is equal to 10.7% of total non-food expenditure
at an all-India level. The proportions are lower in urban areas
than in rural areas. The proportion of OOP payments in house-
hold total as well as non-food expenditure increases with an
increase in levels of consumption expenditure both in rural
and urban areas. This progressive nature of OOP expenditure
is more pronounced in rural than in urban areas. In urban
regions, this trend becomes weaker and in fact reverses at
the two highest consumption levels when OOP payments are
measured as a proportion of non-food expenditure (Table 1).
The rural–urban differential in the progressiveness of the
OOP payment share of total household consumption expendi-
ture can also be seen with the help of concentration curves for
OOP payment share and household consumption expenditure
(Figure 1). The concentration curves for the OOP payment shares
are lower than those for household consumption expenditure
(Lorenz curve), implying progressivity of OOP payments both
in rural and in urban areas. However, the closeness of the con- Figure 1 Concentration curves of OOP payments in rural and urban
centration curve for OOP share to the Lorenz curve for urban India, 1999–2000

Table 1 Average monthly per capita OOP expenditure (in Rs.) and average OOP share (%) to total and non-food consumption expenditure by
quintile groups for rural, urban and combined areas in India, 1999–2000

Rural Urban Combined

Average of OOP shares as % of Average of OOP shares as % of Average of OOP shares as % of
Average per Average per Average per
Consumption capita OOP Total Non-food capita OOP Total Non-food capita OOP Total Non-food
expenditure spending consumption consumption spending consumption consumption spending consumption consumption
quintile* (in Rs.) expenditure expenditure (in Rs.) expenditure expenditure (in Rs.) expenditure expenditure
Poorest 20% 9.02 3.1 8.3 13.84 3.5 8.4 10.25 3.2 8.3

2nd poorest 20% 16.33 4.1 10.3 25.28 4.2 9.1 18.53 4.1 10
Middle 24.86 4.9 11.6 39.74 4.6 9.3 28.49 4.8 11

2nd richest 20% 40.98 6.1 13.4 56.06 4.6 8.6 44.74 5.7 12
Richest 20% 100.38 8.3 15.7 132.99 5.3 8.3 109.18 7.4 13.4
All households 29.62 5 11.5 43.53 4.4 8.8 33.12 4.8 10.7

Standard error 0.700 0.029 0.025 0.338 0.032 0.026 0.318 0.022 0.019
N 71 385 48 924 120 309

Note: In 1999–2000, Rupees (Rs). 43.3 ¼ US$1.

*Quintile groups are made separately for rural and urban areas.
120 HEALTH POLICY AND PLANNING

concentrated among the rich in rural areas, while it is largely expenses for every household in the case of non-institutional
distributed in accordance with the ATP of population groups in care. Although the magnitude of such contamination is not
urban areas. The concentration index7 of OOP payment share discernable from the present database, the NSSO instruction
to total household consumption expenditure is estimated to be manual recognizes this problem only in the case of non-
(þ)0.194 in rural areas compared with (þ)0.079 in urban areas. institutional care in rural areas.
The corresponding standard errors are 0.004 in rural and 0.007 The average monthly per capita expenditure on drugs in
in urban areas.8 The very low value of the concentration index India is Rs. 25 (Rs. 23 and Rs. 30 in rural and urban areas,
of OOP payments in urban areas signifies a very small difference respectively). This amounts to an estimated Rs. 250 billion
between inequity in OOP payments and inequity in ATP. (US$5.7 billion) in total annual expenditure on drugs by
In order to identify the relative contribution of the main households in India for the year 1999–2000 [approximately
components of OOP expenditure, we attempt here to separate Rs. 180 billion (US$4.1 billion) for rural and Rs. 70 billion
household expenditures on drugs from the overall expenditure (US$1.6 billion) for urban India]. This excludes the total gov-
on institutional and non-institutional health care. Expenditure ernment procurement of drugs worth approximately Rs. 20

Downloaded from http://heapol.oxfordjournals.org/ at University of Alabama at Birmingham on December 11, 2012

on drugs is the single largest component of OOP payments billion (US$0.5 billion) in the year 1999–2000 (Sakthivel 2005).
across all consumption quintiles in both rural and urban areas, The estimated value of the total drugs supply in the retail
constituting up to 60% of total expenditure on institutional markets in India for the same year is nearly Rs. 200 billion
(or inpatient) care and 85% of non-institutional (outpatient) (US$4.6 billion).9 The difference between the estimate of
expenditure. Altogether expenditure on drugs is approximately expenditure on drugs by households and the value of total
75% of OOP expenditure; 77% in rural areas and a little less supply of drugs in retail markets is therefore about Rs. 50
than 70% in urban areas. Overall, richer quintiles spend a lower billion (US$1.1 billion) for the year 1999–2000.
proportion of OOP expenditure on drugs compared with poorer
quintiles, both in rural and in urban areas. However, even in Poverty estimates
the top quintile, who are also the top beneficiaries of insti- The increase in number of poor after accounting for OOP
tutional health care services (Mahal et al. 2001), the share of payments is 3.2% (3.5% in rural areas and 2.5% in urban).
OOP payments earmarked for drug purchases is as high as The pre-payment headcount (pre Hp) in 1999–2000 is 25.9%10
70% in rural and 60% in urban areas. For the poorest quintile and post-payment headcount (post Hp) after deducting the
the share is 86% in rural and 83% in urban areas (Table 2). OOP payment from total consumption expenditure is 29.2%.
These estimates (particularly estimates of drugs share in rural The additional number of persons falling into poverty is
areas) should, however, be treated with caution. Given the 32.5 million, with 25.5 million in rural and 7 million in
survey design of the NSSO, expenditure on drugs by house- urban areas. Seventy-nine per cent of the incremental poor
holds in rural areas cannot be separated from total OOP are from rural areas (more than the rural share of the

Table 2 Percentage share of OOP expenditure on institutional and non-institutional health care and drugs in rural and urban India, 1999–2000

Consumption expenditure quintile

Poorest 20% 2nd poorest 20% Middle 2nd richest 20% Richest 20% All households Standard error
Rural
Institutional care 3.28 4.72 6.76 9.15 14.84 9.86 0.064
Non-institutional care 10.59 10.66 10.39 11.59 15.97 12.87 0.233
Institutional drugs 9.47 9.49 12.41 12.74 14.65 12.71 0.123
Non-institutional drugs 76.66 75.13 70.44 66.52 54.54 64.56 0.270
Total drugs 86.13 84.62 82.85 79.26 69.19 77.28 0.270

Urban
Institutional care 5.90 8.66 11.20 14.03 22.08 15.16 0.094
Non-institutional care 10.84 12.48 13.76 15.44 17.81 15.23 0.110
Institutional drugs 10.87 12.04 13.68 13.29 14.31 13.36 0.143
Non-institutional drugs 72.39 66.82 61.36 57.24 45.80 56.25 0.194
Total drugs 83.26 78.86 75.04 70.53 60.11 69.61 0.194

Combined
Institutional care 4.20 6.03 8.27 10.68 17.23 11.61 0.053
Non-institutional care 10.68 11.27 11.54 12.79 16.58 13.65 0.158
Institutional drugs 9.96 10.34 12.84 12.91 14.53 12.93 0.094
Non-institutional drugs 75.17 72.36 67.35 63.62 51.66 61.82 0.189
Total drugs 85.13 82.70 80.19 76.53 66.19 74.75 0.189
 REDUCING OUT-OF-POCKET EXPENDITURES TO REDUCE POVERTY 121

Table 3 Poverty increase after accounting for OOP payments: poverty headcounts and poverty gaps, India, 1999–2000

Poverty measures Rural Urban Combined

Poverty headcounts (in %)
Pre-payment headcount (pre-Hp) 26.84 (0.165) 23.53 (0.193) 25.93 (0.126)
Post-payment headcount (post-Hp) 30.35 (0.173) 26.06 (0.201) 29.17 (0.132)
Poverty impact – headcount (post-Hp  pre-Hp) 3.51 (0.076) 2.53 (0.079) 3.24 (0.056)

Poverty gaps (in Rs.)

Pre-payment gap (pre-G) 17.11 (0.138) 23.35 (0.255) 18.69 (0.124)
Post-payment gap (post-G) 19.97 (0.150) 26.56 (0.275) 21.63 (0.134)
Poverty impact – gap (post-G  pre-G) 2.85 (0.039) 3.21 (0.066) 2.94 (0.033)

Downloaded from http://heapol.oxfordjournals.org/ at University of Alabama at Birmingham on December 11, 2012

Normalized poverty gaps (in %)
Pre-payment normalized gap (pre-NG) 5.22 (0.042) 5.14 (0.054) 5.17 (0.033)
Post-payment normalized gap (post-NG) 6.10 (0.045) 5.85 (0.058) 5.99 (0.035)
Normalized poverty impact (post-NG  pre-NG) 0.87 (0.0112) 0.71 (0.0137) 0.82 (0.009)

Notes: Figures in parentheses are Standard Errors.

The estimates of poverty headcount are slightly lower than the official estimates of poverty (26.1%) by Planning Commission of India (2002);
mainly because the estimate of poverty ratio based on unit level data in north eastern states is much lower than the official figures. The official
poverty ratio takes a common figure for all north eastern states which is that of Assam.

total population) and 21% are from urban areas (Table 3).
These figures do not include persons already below the
poverty line and who are pushed further down to acute
poverty due to OOP payments.
At the national level, the consumption level of the poor dips
by an average of Rs. 3 per capita per month because of OOP
payments, which also serves to measure poverty deepening or
the increase in the poverty gap. The intensity of the poverty gap
is higher in urban areas (Rs. 3.21) than in rural areas (Rs. 2.85)
but the normalized poverty gap, which is standardized by
respective poverty lines, is higher in rural areas (0.87%) than Figure 2 Pre-payment and post-payment consumption expenditure,
in urban areas (0.71%). This shows that the relative burden 1999–2000
Notes: Expre – Shows monthly per capita expenditure before OOP.
in rural areas is much greater. The impact of OOP payments Expost – Monthly per capita expenditure of households after OOP.
on poverty can be clearly observed with the help of a Pen
parade graph (Figure 2).
The Pen parade graph plots households’ pre-payment as well
in the middle of the gross expenditure distribution and some
as post-payment per capita consumption against the cumula-
at the high end of pre-payment consumption who are pulled
tive percentage of individuals ranked by pre-payment con-
below the poverty line by OOP payments. It is evident from
sumption (upper boundary of the Pen parade). The x-axis, at
the figure that for households just above the poverty line
the intersection of the pre-payment curve with the national
even a small expenditure on OOP payments will drop them
poverty line (horizontal line set at Rs. 361 for 1999–2000),
below the poverty line.
measures the poverty headcount. As indicated by this point
in the graph, the pre-payment poverty ratio (which is also the
official estimate of poverty in India) is 26%. The ‘paint drops’ OOP expenditures and poverty differentials
from the pre-payment curve depict the consumption expen- across states
diture of individuals who are pulled below the pre-payment OOP expenditure differentials
curve because of OOP payments. The lower boundary of the In general, OOP payment share in total expenditure is
‘paint drops’ plots the post-payment curve. The proportion found to be higher in high-income states than in poor states
below the poverty line is the post-payment headcount of (Pearson correlation coefficient is (þ)0.5 between per capita
poverty, which is 29.5%. The difference in the two headcounts state domestic product and OOP share in total consumption
measured again on the x-axis is the poverty headcount impact expenditures). Developed states such as Punjab, Haryana,
of the OOP payment and is 3.2%. The area below the poverty Maharashtra and Kerala have a higher share of OOP expen-
line and that above the pre-payment curve shows the extent diture in consumption expenditures (5% or above) compared
of poverty gap. It is clear that many individuals below the with the 2–4% range in poorer states such as Bihar, Jammu and
poverty line are dragged further down by medical expenses Kashmir, Orissa, Rajasthan and Assam. An exception is Uttar
when these are netted out. There are also many individuals Pradesh, a poorer state, with an OOP share at 6.5%, just below
122 HEALTH POLICY AND PLANNING

Downloaded from http://heapol.oxfordjournals.org/ at University of Alabama at Birmingham on December 11, 2012

Figure 3 Concentration curves of OOP payments in four states with high OOP share, 1999–2000

the highest level in Kerala. Kerala, which has the highest rank In these states, the concentration curves of OOP payments
in terms of human development indicators in India, has the lie below the concentration curves for total consumption
highest OOP expenditure share, at over 7%. Two higher income expenditure, implying that OOP expenditure increases with an
states, namely Gujarat and Tamil Nadu, show a contrasting increase in consumption expenditure in all these states. The
picture from the general trend. difference across these four states lies in terms of the intensity
While all the poor states are characterized by lower per of the OOP payments share between the rich and the poor,
capita annual OOP expenditures (between Rs. 400–800), all as indicated by the difference between the two concentration
high-income states except Maharashtra are characterized by curves (the OOP concentration curve and the Lorenz curve).
annual per capita OOP expenditures of more than Rs. 1000. This gap is lowest in Kerala and highest in Uttar Pradesh. In
Most middle-income states, such as West Bengal, Andhra Haryana and Uttar Pradesh, the gap widens only among the
Pradesh, Karnataka and Himachal Pradesh, have shares of upper consumption quintiles, implying that OOP payments
OOP expenditures between 4–5% of total expenditure, and per are largely concentrated among higher consumption quintiles.
capita annual OOP expenditures ranging from Rs. 500–850. In Punjab and Kerala, OOP payments are largely distributed in
The states of Orissa, Assam and Rajasthan, characterized accordance with the ATP.
by mass poverty, have the lowest annual per capita OOP Our findings suggest that share of expenditure on drugs
expenditure (less than Rs. 600), and Assam and Bihar have in total OOP expenditure is lower in developed states. For
the lowest OOP payment shares in total household consump- example, in poorer states like Bihar, Orissa, Uttar Pradesh,
tion expenditure at 2 and 3.8%, respectively. Jammu and Kashmir, and Rajasthan, share of expenditure
The pattern of OOP payment across quintile groups was on drugs is as high as 90% or more. In more developed states
examined in the four states of Haryana, Punjab, Uttar Pradesh like Maharashtra, Tamil Nadu, Gujarat and Karnataka, it is
and Kerala, which have very high levels of OOP expenditure, 70% or less, implying comparatively higher expenditure on
over 6%, and per capita private expenditures of over Rs. 1000, institutional and/or non-institutional care. In Kerala and
but which vary extensively in levels of development.11 To illus- Gujarat, the share of OOP expenditure on institutional care is
trate the distribution of OOP payments in these four states, higher than in all other states, and in Tamil Nadu, Karnataka,
the concentration curves of OOP payments are presented in Maharashtra and West Bengal, OOP share on non-institutional
Figure 3. care is among the highest.
 REDUCING OUT-OF-POCKET EXPENDITURES TO REDUCE POVERTY 123

Headcount in % Poverty gap in Rs. Normalised poverty gap in % Table 4 Poverty headcount (number of persons in 1000s) after
7.00 accounting for OOP payments in major Indian states and by rural
and urban areas, 1999–2000
6.00
Rural Urban Combined
5.00 (1000 persons) (1000 persons) (1000 persons)
Uttar Pradesh 8687.4 1380.4 10 067.8
4.00
Bihar 4237.6 394.3 4631.9

3.00 Maharashtra 2099.7 959.1 3058.8

Madhya Pradesh 2087.2 634.0 2721.2
2.00
West Bengal 2040.8 347.7 2388.6
1.00 Andhra Pradesh 1137.2 492.4 1629.5
Orissa 1349.6 137.0 1486.7

Downloaded from http://heapol.oxfordjournals.org/ at University of Alabama at Birmingham on December 11, 2012

0.00
Tamil Nadu 1059.6 230.0 1370.8
28)

)
2)

)
)

)
)

)
5)

)
12)

)
38)

)
907

141
533

398
715

012

343

040
625

11 4
26 2
569
916

976
(63

(96

123
(10

(1 9
(12

(23
(14

(16
(15

(23
(18

(21
(18
(15
(

Rajasthan 1030.4 340.3 1289.6

BI

K(
OR

UP
AS

MP

TN
RA

MA
AP

KA
HP

PU
GU

HA
KE
WB
J&

Karnataka 937.9 342.7 1280.6

Figure 4 Poverty headcount and poverty gap as a result of OOP Gujarat 647.3 420.6 1067.8
expenditures in major Indian states, 1999–2000.
Notes: Analysis has been done for 17 major states. The acronyms for Kerala 564.1 244.3 808.3
states are: AP – Andhra Pradesh; AS – Assam; BI – Bihar; GU – Gujarat; Punjab 318.2 155.7 473.9
HA – Haryana; HP – Himachal Pradesh; J&K – Jammu and Kashmir;
KA – Karnataka; KE – Kerala; MA – Maharashtra; MP – Madhya Assam 454.4 9.8 464.2
Pradesh; OR – Orissa; PU – Punjab; RA – Rajasthan; TN – Tamil Nadu; Haryana 307.5 121.8 429.3
UP – Uttar Pradesh; WB – West Bengal.
Figures in parentheses against the state names are per capita state Himachal Pradesh 66.3 4.8 71.0
domestic product in Rupees (Rs.). Data for Bihar, Madhya Pradesh Jammu & Kashmir 49.9 7.5 57.4
and Uttar Pradesh relate to erstwhile states before their bifurcation.
All India 25 528.7 6992.6 32 519.9

Notes: States are arranged in descending order of total number of persons

Poverty differentials going below poverty line.
The poverty impact of OOP expenditure shows that states with Data for Bihar, Madhya Pradesh and Uttar Pradesh relate to erstwhile states
a relatively low per capita state domestic product have a higher before their bifurcation.
increase in poverty incidence (headcount) and intensity (gap)
compared with richer states (Pearson correlation coefficient the four richest states—Gujarat, Haryana, Maharashtra and
of (-)0.392 and (-)0.395, respectively, significant at the 10% Punjab—the proportion is only 67%. This clearly shows that
level). Exceptions are Assam among the poor and Maharashtra OOP payments have a greater impact on poverty levels in rural
among the rich states (Figure 4). areas of poorer states. In contrast, the urban poor are affected
Among all the major states, Uttar Pradesh shows the highest more by OOP payments, in relative terms, in richer states than
increase in poverty, followed by three other poor states: Bihar, in poorer states. In absolute terms, Maharashtra ranks second
Orissa and Madhya Pradesh. These four states taken together in terms of the increase in number of urban poor because of
constitute 58% of the total increase in poverty headcount OOP expenditures, following closely behind Uttar Pradesh.
because of OOP payments. In Uttar Pradesh alone, a 6% Poverty differentials among states in terms of the OOP
increase in the poverty ratio implies more than 10 million payment share and increase in poverty headcount show three
persons or over one-fifth of the state’s total number of poor distinct categories: (1) high OOP payment share and high
have fallen into poverty through OOP payments. Similarly, in poverty increase (Uttar Pradesh, Maharashtra and Madhya
Bihar, 4.6 million, and in Maharashtra and Madhya Pradesh, Pradesh), (2) low OOP payment share but high poverty increase
3.1 million and 2.7 million people, respectively, have joined (Bihar, Orissa and Rajasthan), and (3) high OOP payment
the ranks of the poor as a result of OOP expenditure on share but low poverty increase (Kerala, Punjab and Haryana).
health care (Table 4). These are also the states with the largest Underlying economic and human development factors in
poverty deepening, with average consumption shortfalls of these states can explain the variations in poverty increase
over Rs. 3 per capita per month, the highest being in Uttar due to OOP payments.
Pradesh at almost Rs. 6 per capita per month. Among the
low- and middle-income states, Assam, Jammu and Kashmir,
and Himachal Pradesh have the lowest numbers of people
moving below the poverty line and have the smallest impact Discussion
in terms of poverty deepening. Before discussing the main findings, we will explain the choice
The rural–urban breakdown of the poverty increase in dif- of CES data instead of data from the NSSO Health Survey (HS)
ferent states provides an interesting insight. In most of the for estimates related to OOP expenditures and impoverishment
poorer states, such as Assam, Orissa, Uttar Pradesh, Bihar, in India. Reporting of health expenditures in the Health Survey
Jammu and Kashmir and West Bengal, rural areas account is based on self-reporting of illness of family members, while
for 87% of the total increase in poverty. The proportion is as in the CES it is based on recall of expenditure on treatment.
high as 90% in Assam, Bihar and Orissa. In comparison, in Recall periods in both the surveys are ‘last one year’ for
124 HEALTH POLICY AND PLANNING

Table 5 Comparison of 1995–96 NSS Health Survey and 1999–2000 Consumer Expenditure Survey for OOP payments

NSSO survey Rural Urban Combined

Households reporting OOP (%)
1995–96 HS 24.87 (0.162) 24.12 (0.192) 24.61 (0.124)
1999–2000 CES 70.21 (0.172) 69.87 (0.209) 70.07 (0.133)
Total OOP as a share of total household expenditure (%)
1995–96 HS 6.83 (0.032) 5.78 (0.039) 6.47 (0.025)
1999–2000 CES 6.09 (0.037) 5.09 (0.042) 5.72 (0.028)

Note: Figures in parentheses are Standard Errors.

Downloaded from http://heapol.oxfordjournals.org/ at University of Alabama at Birmingham on December 11, 2012

hospitalizations (i.e. inpatient), but they differ for outpa- expenditure can be estimated only by comparing two or more
tient care, being ‘last 15 days’ in the Health Survey and similar rounds of CES surveys.
‘last 30 days’ in the CES. With the longer recall period for Further, the Health Survey does not separately record drug
outpatients in the CES, a larger proportion of households expenditure, particularly in respect of institutional or hospital
reported OOP expenditure in the 1999–2000 CES compared care. This is one of the most important reasons why most of
with the 1995–96 Health Survey (Table 5). Although the sam- the earlier literature concludes that hospitalization accounts
pling error of both the proportion of households reporting for the largest outlay of both public and private resources
OOP and the proportion of OOP to total household expendi- in India and is the largest cause of catastrophic payments
ture is marginally higher in the CES than in the Health Survey, (Peters et al. 2002; Roy and Hill 2007). Based on the CES
the substantially higher proportion of households reporting data, where expenditure on drugs and medicines can be
OOP in the CES led us to use this for poverty impact separated for institutional as well as non-institutional health
calculations. care, our findings show that expenditure on drugs is the
The CES captures OOP payment as a part of total household largest component of OOP payments, both for institutional
consumption expenditure, whereas the Health Survey concen- and for non-institutional health care.
trates more on measuring health expenditures exclusively for In the CES, the proportion of OOP payments spent on drugs
those households who report any ailment and/or hospitaliza- for non-institutional care may be slightly overestimated as the
tion of family members. Since the Health Survey reports only NSSO instructions to enumerators mention that: ‘In the rural
about 25% of households experiencing illness and/or hospital- areas, doctors charge a consolidated amount for consultation as
izations, only that proportion reports expenditure on health well as providing medicines. In such a case, the total amount
care. Further, all households that report OOP expenditures will be recorded against item 420 (medicine)’ (Government of
only when they have illness/hospitalisation, report a signifi- India 2001, p. D-26). However, this problem is associated only
cantly higher fraction of their total household expenditure as with outpatient care in rural areas, where it is common practice
OOP expenditure. This is further accentuated by the shorter for doctors to provide drugs and to charge for them with their
recall period and lower total household consumption expen- consultation fee. In rural areas, the share of drug expenditure
diture reported in the Health Survey.12 In the CES, in contrast, in total OOP payments is 70% in the top quintile compared
approximately 70% of households report expenditure on health with 86% in the lowest quintile. Since the level of amalgama-
care. tion (of doctors’ fees and drug charges) may be considered to
The large difference in households reporting health expen- be lowest in the top population quintile, the lowest estimate
ditures in these two surveys could be due to exclusion of of the share of drugs in OOP payments may be taken as 70%,
households incurring expenditures on self-medication, or those rather than 77% as shown in the CES data. In urban areas
incurring expenditures on payments to informal providers, also the combined share of drugs expenditure in total OOP
because of the manner in which questions on health expen- payments is 70%.
ditures are posed in the Health Survey. It could also be due The estimate that 70% of OOP expenditure is spent on
to differences in recall for expenditures vs. ailments. In any drugs implies a total annual expenditure on drugs by house-
case, this large difference cannot be ignored, as using the holds of Rs. 230 billion (US$5.3 billion). This figure is
Health Survey leads to relatively smaller estimates of the approximately 14% higher than the total retail market value
proportion of households making OOP payments and of of drugs from all pharmaceutical companies in India (IDMA
the increase in poverty headcount because of them. For 2004). One reason for this difference could be the purchase of
example, the use of Health Survey data in Peters et al. (2002, traditional (Ayurvedic, homeopathic, Unani, Siddha) and other
p. 216) led them to estimate a 2.2% increase in poverty head- drugs by households, which are not accounted for in the IDMA
count because of OOP expenditure. Our estimates are approxi- data. Hence, if we propose an average of 65–70% as the share
mately 1% higher at 3.2%. Part of this difference in poverty of medicines in total OOP payments, it may not be far from
headcount may also be explained by a real increase in poverty the truth. Further research is needed to separate drugs expen-
between the two referred periods, but methodological dif- ditures from consultation fees and correctly estimate the actual
ferences highlighted above cannot be discounted. The actual expenditures on drugs from households’ OOP expenditures,
magnitude of increase in poverty after discounting for OOP including those on traditional Indian systems of medicine.
 REDUCING OUT-OF-POCKET EXPENDITURES TO REDUCE POVERTY 125

There is enough evidence from NSS data to infer that poor medical expenses (Ranson 2002). Gujarat, Karnataka and
people spend the largest proportion of OOP expenditure on Tamil Nadu are known for the good performance of NGOs,
the purchase of drugs and least on inpatient treatment. One particularly regarding midday meal schemes, immunization
reason for this may be the irrational use of drugs, by both of children and other health awareness programmes.
prescribers and users. Further, due to poor access to formal The poverty estimates based on 1999–2000 CES data show
health care services, most of the poor use informal health that 3.2% of the total population in India plunged into
services in large quantities and thereby spend still higher poverty because of OOP expenditure, compared with 2.2%
quantities on purchase of drugs and self-medication compared estimated by Peters et al. (2002) based on the 1995–96 Health
with those in wealthier quintiles. Hence, it is argued that Survey. While the official poverty line, used as a yardstick to
expenditure on drugs is one of the major causes of impover- measure poverty headcount, has been widely observed as a
ishment in India. narrow interpretation of poverty, we use this mainly to show
In India, OOP health expenditure is relatively high as a the comparative level of poverty headcount before and after
share of total household resources in general and total health OOP payments. As mentioned above, the lower estimates in

Downloaded from http://heapol.oxfordjournals.org/ at University of Alabama at Birmingham on December 11, 2012

expenditure in particular, compared not only with developed Peters et al. (2002) may be due to a real increase in poverty
countries but also with other low-income countries.13 The between two time periods, or to a large number of households
increasing share of OOP health care payments in households’ purchasing drugs and services without reporting illness and
total as well as non-food expenditure is in accordance with the not being counted in the estimates by Peters et al. As indicated
literature on health being a ‘normal’ good (McGuire et al. 1988; earlier, it is noteworthy that sampling errors of the two surveys
Gertler and Gaag 1990). In the top consumption expenditure are only marginally different (see Table 5).
quintile, purchase of a greater quantity coupled with better Rural areas and poor states experience a higher increase
access to health services pushes OOP expenditure up. People in in the poverty headcount through OOP expenditure mainly
middle and higher expenditure quintiles in rural areas may because a large proportion of their population is concentrated
be better off because of being able to purchase better care around the poverty line, and hence even a small amount of
than their poorer counterparts, but they are worse off than OOP expenditure will push many households below the poverty
their urban counterparts as they have to spend a higher pro- line. In contrast, in urban areas and richer states, where more
portion of their consumption expenditure on OOP payments people have monthly per capita expenditure well above the
(Table 1). This could also be due to households incurring more poverty line, the same level of OOP payment will not cause the
same level of impoverishment.
expenses on travel and other related expenditures to access
A high share of OOP payments would normally imply a higher
facilities in urban areas.
increase in poverty headcount, as in the case of Uttar Pradesh
The results for state-level differentials computed in this
and Maharashtra, but the states of Bihar and Orissa had a
paper are very similar to those found in other studies (Mahal
high increase in poverty despite a low OOP payment share. The
et al. 2001; Peters et al. 2002, 2003; NCMH 2005; Government of
states of Bihar, Madhya Pradesh, Orissa and Uttar Pradesh are
India 2005). Public spending on health is generally very low.
characterized by mass poverty. Taken together they constitute
Per capita private spending on health in Kerala, Haryana and
more than 58% of the total increase in poverty because of
Punjab is four times higher than that in Rajasthan and three
OOP expenditure. The level of development in Punjab, Haryana
times that in Bihar. One reason for this is the level of health
and Kerala explains their high OOP share but low poverty
transition in different states. According to Peters et al. (2003),
increase. In these three states, which have some of the lowest
Kerala is in the ‘late transition’ stage and Uttar Pradesh the
percentages below or just above the poverty line, even high
‘early transition’ stage. OOP expenditure in Kerala is high
OOP payments do not generally push people below the
because of higher expenditure on institutional care and on
poverty line. Further probing is required, however, to discover
lifestyle diseases such as heart conditions. Poorer states such
what proportion of the hidden poor uses borrowing, remit-
as Assam, Bihar, Jammu and Kashmir, West Bengal, Rajasthan,
tances, sale of assets and past savings to finance their
etc., have a relatively low share of OOP expenditure because
compulsory health care needs, which, in turn, might answer
they are in the ‘early’ to ‘mid’ transition stage, when people
the question of the cumulative poverty impact of OOP expen-
spend less on health care services on account of low incomes, diture in subsequent years. Mahal (2006) also argues that
limited access to health care, lack of awareness (poor literacy convincing evidence on long-term or chronic poverty will
rates) and poor infrastructure (in terms of number of facili- come only from longitudinal analysis and not from cross-
ties, availability of medicines and number of health workers). sectional analysis as has been done in this study.
This coupled with low public spending makes the situation
particularly severe.
The lower share of OOP expenditure in Karnataka, Gujarat
and Tamil Nadu needs a special mention. These are middle- Conclusions and recommendations
income states in the ‘mid’ to ‘late’ transition stages. One In India, an average of 4.8% of total household consump-
reason for their low OOP expenditure may be to do with other tion expenditure is spent on OOP health care payments. Poor
sources of financing, such as higher government expenditures quintiles spend a relatively lower proportion of their con-
and better risk-pooling systems through insurance, or high sumption expenditure on OOP payments than rich quintiles,
expenditure by non-governmental agencies. It has been shown in both rural and urban areas. Middle and wealthier expen-
in Gujarat that community-based health insurance schemes diture quintiles in rural areas bear a greater burden in
help to protect poor households against uncertain risks of comparison with their urban counterparts. The OOP payment
126 HEALTH POLICY AND PLANNING

shares were generally found to be positively correlated to per done is for policymakers to target specific areas and specific
capita state domestic product. In general, poor states have populations in certain states where the poverty impact of
low OOP shares of consumption expenditure because of low OOP payments is greatest.
incomes, limited access to health care, lack of awareness and It should be mentioned here that the Government of India,
poor infrastructure therein. Conversely, middle and higher Ministry of Health and Family Welfare, started a new scheme
income states generally have high OOP expenditure, mainly called the National Rural Health Mission (NRHM) in April
on account of better health care seeking behaviour and pro- 2005 to provide quality health care to every household through
vider choice as shown by larger expenditures on institutional its upgraded health infrastructure and provision of round-
care. Sometimes, governmental and/or non-governmental inter- the-clock health services (Ministry of Health and Family
ventions may reverse the situation, as in the case of Gujarat Welfare, undated). When fully operational, it is hoped that
and Tamil Nadu. More often, very poor government facilities this will reduce OOP expenditures to a bare minimum and
and greater dependence on private facilities in nearby towns will stop people sliding down the poverty line.
leads to a high OOP payment share, as is probably the case Further research needs be undertaken to identify character-

Downloaded from http://heapol.oxfordjournals.org/ at University of Alabama at Birmingham on December 11, 2012

for Uttar Pradesh. Epidemiological changes such as lifestyle istics of households that are moving below the poverty line
diseases and greater utilization of institutional care may also and to analyse the impact of OOP payments on poverty as a
lead to high OOP payments, as in case of Kerala. dynamic process, that is, whether households move out of
Of the 4.8% OOP share of total consumption expenditure, poverty by seeking medical care and who constitutes the
about 65–70% is spent on drugs and only 30–35% (urban/rural new poor.
variations) is for inpatient and outpatient care. In rural areas
and poorer states, the average share of expenditure on inpa-
tient and outpatient care is lower, but expenditure on Acknowledgements
medicines is higher. There is a need not only to improve the
The authors were lead persons from India for the research
availability of services in rural areas and in poor states, but
project ‘Equity in Health Financing in Asia Pacific Countries’
also for a policy to rationalize high drug expenditure in
(referred to as the EQUITAP project), undertaken at the
these areas. Our research also suggests improving survey Institute for Human Development, New Delhi. The EQUITAP
designs for estimating OOP expenditure, and expenditure project was funded by the European Commission under the
on its components, for poverty estimations. INCO-DEV programme (ICA4-CT-2001–10015).
OOP expenditures have a striking impact on increasing the The authors are extremely thankful to Dr P Berman,
poverty ratios in the country. Over 32 million persons fall Dr O O’Donnel, Dr A Mahal, Dr M K Premi, Dr Sakthivel
below the poverty line in a single reference year. The increase and Dr A Singh, and to two anonymous referees who pro-
in the number of poor as a result of OOP expenditure was vided insightful comments that have greatly improved this
higher in rural areas (3.5%) than in urban areas (2.5%). Also paper. The authors would also like to thank Ms S Prasad for
the intensity of poverty in terms of poverty deepening when editorial help. However, the authors alone are responsible for
standardized by poverty lines was higher in rural areas. OOP all errors and omissions.
payments, in addition to pushing a large number of people
below the poverty line, also severely affect the living status
of many households already below the poverty line. In terms
of the poverty gap, those in urban areas below the poverty Endnotes
1
line face a larger average consumption shortfall than those Of total GDP, approximately 5% is spent on health care; government
spends less than 1% of GDP on financing health care, with the rest
in rural areas.
coming from the private sector, including individual and household
The poverty impact of health payments is greater in the expenditure.
poorer states, especially in rural areas. Targeted policies in 2
In OECD countries, the average OOP share is even lower at 15 to 20%
just five poor states to reduce OOP expenditure could help (based on WHO National Health Accounts data, WHO 2006).
3
to prevent almost 60% of the poverty headcount from OOP Two lines of absolute poverty that have been developed and used by
the World Bank are $1.08 and $2.15 per capita per day at 1993
payments. In the five richest states, targeting is required among purchasing power parities (Ravallion 1998; Chen and Ravallion
the urban poor to enable households to avoid slipping below 2001). The lower of these is the median of the 10 lowest poverty
the poverty line and also to prevent those already below the lines operational in a sample of low-income countries (Ravallion
poverty line from falling deeper into poverty. This would be et al. 1991). It represents a very low living standard that is often
referred to as ‘extreme poverty’ (Chen and Ravallion 2004).
an important contribution to achieving the first Millennium 4
The NSSO is a premier institution of the Government of India under
Development Goal (of halving the population below the poverty the Ministry of Statistics and Programme Implementation, which
line) to which India is a signatory. has been collecting household data on a regular basis since 1950.
5
We conclude that there is a need for: improved survey The large CES (also known as the full round) is conducted by the
NSSO after an interval of approximately 5 years. The previous large
designs for impoverishment work; rationalized drug policies samples are the 50th round (1993–94), 47th round (1987–88), 43rd
(including free supplies); pro-poor health financing policy round (1983), 38th round (1977–78) and 32nd round (1972–73).
focusing on financial protection not only for those close to In-between these quinquennial rounds, annual rounds (also
the poverty line, but also those who are already below it in known as the thin sample) are conducted with a smaller sample.
The CES 55th round for 1999–2000 was the latest available large
both rural and urban areas; and innovative financing mech-
sample at the time the study was completed.
anisms on the collection, pooling and purchasing side to 6
There are many reference groups, rural and urban, in different
reduce the intensity of poverty. One of the ways this could be states. Each of these has a different official poverty line,
 REDUCING OUT-OF-POCKET EXPENDITURES TO REDUCE POVERTY 127

commonly known as state-specific rural and urban poverty lines, Kakwani NC, Wagstaff A, van Doorslaer E. 1997. Socioeconomic
which have been considered for this study. inequalities in health: Measurement, computation and statistical
7
The concentration index measures the underlying inequity in OOP inference. Journal of Econometrics 77: 87–104.
payments in relation to the ATP. The formula used for calculating
the concentration index (C) is: C ¼ [2/a] cov (yi,Ri), where y is the Krishna A. 2004. Escaping poverty and becoming poor: who gains,
OOP variable, a is mean OOP, and Ri is the ith household’s who loses, and why? World Development 32: 121–36.
fractional rank in ATP distribution (i.e. the household’s rank in Mahal A. 2003. The distribution of public health subsidy in India.
the consumption expenditure distribution). In: Yazbeck AS, Peters DH (eds). Health policy research in South Asia:
8
In order to correct for likely autocorrelation and heteroscedasticity building capacity for reform. Human Development Network, Health,
in the regressor (fractional rank of the ATP variable), the stan- Nutrition and Population series. Washington DC: The World Bank,
dard error has been calculated using the Newey-West variance-
pp. 33–63.
covariance matrix (Kakwani et al. 1997).
9
The size of the Indian pharmaceutical industry, both bulk drugs Mahal A. 2006. Health spending and poverty. The Lancet 368: 1308–9.
and formulations, was estimated at Rs. 354.7 billion in 2003–04 Mahal A, Singh J, Afridi F et al. 2001. Who benefits from public health
(US$7.7 billion) (IDMA 2004). spending in India? New Delhi: National Council of Applied Economic
10
The pre-payment headcount is the same as the official estimate of the

Downloaded from http://heapol.oxfordjournals.org/ at University of Alabama at Birmingham on December 11, 2012

Research.
poverty headcount for India. For the state-wise and all-India
poverty estimates see Planning Commission (2002). Mahal A, Sakthivel S, Nagpal S. 2005. National health accounts for
11
Private expenditures include expenditures of firms, non-gov- India. In: National Commission on Macroeconomics and Health
ernment organizations and households (Government of India (NCMH). Financing and delivery of health care services in India. NCMH
2005: Table 1.3). OOP expenditures form over 90% of private Background Papers. New Delhi: Ministry of Health and Family
expenditures. Welfare, Government of India, pp. 256–63.
12
In all NSSO surveys other than the CES (but including the Health McGuire A, Henderson J, Mooney G. 1988. The economics of health care:
Survey), an abridged version of the consumption expenditure
An introductory text. London: Routledge.
schedule is canvassed. This usually provides lower estimates of
total household consumption expenditure as compared with Ministry of Health and Family Welfare. Undated. National Rural
that found in the quinquennial CES rounds (see also NSSO Health Mission: A partnership for meeting people’s health
1998, p. 12; NSSO 2006, p. 21). needs. Framework for implementation 2005–2012. New Delhi:
13
OOP payments constitute 80% of total health expenditure in India, Ministry of Health and Family Welfare, Government of India.
compared with an average of 65% in low-income countries in 2000 NCMH. 2005. Report of the national commission on macro economics
(WHO 2006, and the WHO National Health Accounts website:
and health. New Delhi: Ministry of Health and Family Welfare,
http://www.who.int/nha).
Government of India.
NSSO. 1998. Morbidity and treatment of ailments (NSS Fifty-Second
Round, July 1995–June 1996). New Delhi: National Sample Survey
Organisation, Department of Statistics, Government of India.
References NSSO. 2006. Morbidity, health care and the condition of the aged (NSSO
Ahuja R. 2004. Health insurance for the poor. Economic and Political 60th Round, January – June 2006). New Delhi: National Sample
Weekly 39: 3171–8. Survey Organisation, Ministry of Statistics and Programme
Berki SE. 1986. A look at catastrophic medical expenses and the poor. Implementation, Government of India.
Health Affairs 5: 138–45.
O’Donnell O, van Doorslaer E, Rannan-Eliya RP et al. 2005. Explaining
Chen S, Ravallion M. 2001. How did the world’s poorest fare in the the incidence of catastrophic expenditures on health care: comparative
1990s? Review of Income and Wealth 47: 283–300. evidence from Asia. EQUITAP Working Paper #5. Colombo: Erasmus
Chen S, Ravallion M. 2004. How have the world’s poorest fared since University, Rotterdam and IPS.
the early 1980s? World Bank Research Observer 19: 141–69. O’Donnell O, van Doorslaer E, Rannan-Eliya RP et al. 2007. The
Garg CC. 2001a. A conceptual framework for national health accounts in India: Incidence of public spending on healthcare: comparative evidence
A pilot study on Punjab. New Delhi: Institute of Economic Growth, from Asia. World Bank Economic Review 21: 93–123.
September. Peters DH, Yazbeck AS, Sharma RR et al. 2002. Better health system for
Garg CC. 2001b. Financing of health care in India: results of NHA from India’s poor: findings, analysis and options. Human Development
Karnataka. Journal of Health Management 3: 199–237. Network, Health, Nutrition and Population series. Washington DC:
Garg CC. 2006. Is health insurance feasible in India? Issues in private The World Bank.
and social health insurance. In: Prasad S, Satyamala C (eds). Peters DH, Rao SK, Fryatt R. 2003. Lumping and splitting: the health
Securing health for all: dimensions and challenges. New Delhi: Institute policy agenda in India. Health Policy and Planning 18: 249–60.
for Human Development, pp. 430–50. Planning Commission. 2001. Poverty in India, press release. New Delhi:
Garg CC, Karan AK. 2006. Level and pattern of catastrophic health Planning Commission, Government of India.
expenditure in India. In: Prasad S, Satyamala C (eds). Securing Planning Commission. 2002. National Human Development Report.
health for all: dimensions and challenges. New Delhi: Institute for New Delhi: Planning Commission, Government of India.
Human Development, pp. 46–69. Ranson MK. 2002. Reduction of catastrophic health expenditures by
Gertler P, van de Gaag J. 1990. The willingness to pay for medical care. a community based health insurance scheme in Gujarat, India:
Baltimore, MD: Johns Hopkins University Press. current experiences and challenges. Bulletin of World Health
Government of India. 2001. Consumer expenditure survey, 55th round, Organization 80: 613–21.
1999–2000. New Delhi: National Sample Survey Organisation, Ravallion M. 1998. Poverty lines in theory and practice. LSMS working
Ministry of Statistics and Programme Implementation. paper no. 133. Washington DC: World Bank.
Government of India. 2005. National Health Accounts India (2001–2002). Ravallion M, Datt G, van de Walle D. 1991. Quantifying absolute poverty
New Delhi: Ministry of Health and Family Welfare. in the developing world. Review of Income and Wealth 37: 345–61.
IDMA. 2004. Annual report. New Delhi: Indian Drug Manufacturers Roy K, Hill HD. 2007. Equity in out of pocket payments for hospital
Association (IDMA). care: evidence from India. Health Policy 80: 297–307.
128 HEALTH POLICY AND PLANNING

Russell S. 2004. The economic burden of illness for households in Wagstaff A, Van Doorslaer E. 2003. Catastrophe and impoverishment
developing countries: a review of studies focusing on malaria, in paying for health care: With applications to Vietnam 1993–98.
tuberculosis, and human immunodeficiency virus/acquired immu- Health Economics 12: 921–34.
nodeficiency syndrome. American Journal of Tropical Medicine and WHO. 2006. Working together for health. World Health report 2006.
Hygiene 71(Supp. 2): 147–55. Geneva: World Health Organization. Annex 2. Updates also
Sakthivel S. 2005. Access to essential drugs and medicines in financ- available at: http://www.who.int/nha/country/ind/en/, accessed
ing and delivery of health care services in India. In: National 31 March 2007.
Commission on Macroeconomics and Health (NCMH). Financing Xu K, Zeramdini R, Klavas J et al. 2003. Household catastrophic
and delivery of health care services in India. NCMH Background Papers. health expenditure: a multicountry analysis. The Lancet 362:
New Delhi: Ministry of Health and Family Welfare, Government of 111–7.
India, pp. 185–212. Yazbeck AS, Peters DH. 2003. Health policy research in South Asia:
Van Doorslaer E, O’Donnell O, Rannan-Eliya RP et al. 2006. Effects of building capacity for reform. Human Development Network,
payments for health care on poverty estimates in 11 countries in Health, Nutrition and Population series. Washington DC:
Asia: an analysis of household survey data. The Lancet 368: 1357–64. The World Bank.

Downloaded from http://heapol.oxfordjournals.org/ at University of Alabama at Birmingham on December 11, 2012