Introduction

The index of economic status of households, called the wealth index in the third National Family Health Survey (NFHS-3), is based on household asset holdings and housing characteristics.1 This is an indicator of level of wealth widely used in Demographic and Health Surveys (DHS) statistics across the world and is reported to be consistent with the expenditure level of households (Rutstein and Johnson 2004). In the survey, an enquiry was made on the absence or presence of 33 characteristics in all the households in the sample. According to the NFHS-3 report, each household asset was assigned a weight generated through principal component analysis, and the resulting scores were standardised in relation to the normal distribution with a mean of zero and standard deviation of one (IIPS and Macro International 2007; Gwatkin et al 2000). Each household was then assigned a score for each asset owned, and they were summed up to obtain its wealth index factor score (Wifs). Individuals in the sample were assigned the Wifs of the household in which they resided. Based on the wealth score distribution, the sample population in the NFHS-3 was divided into five groups, or quintiles, each with an equal number of individuals.

The DHS wealth index is stated to have several advantages over the standard of living index classification used in the earlier rounds of the NFHS (IIPS and ORC Macro 2000). Such an alternative for comprehending household economic status is used primarily because of the absence of information on household incomes and expenditures in the NFHS. There is also universal agreement on the use of wealth quintile variables available in the NFHS-3 data for describing a household’s relative economic status while analysing inequalities in health outcomes and other health indicators.

However, we do have certain reservations regarding the procedure adopted for classifying the surveyed population into different quintiles, particularly in the Indian context. These do not have to do with the assignment of a wealth quintile score for each household based on a set of 33 variables pertaining to the availability of household assets and characteristics. There can be

o bvious reasons for disagreement among researchers if one is attempting the economic classification of a population based on

NATIONAL FAMILY HEALTH SURVEY-3

Table 1: Variation in Median Value of Household Wealth Index Factor Score, Rural-Urban Divide in Wealth and Wealth Concentration, across Indian States, 2005-06

Source: Computed using the NFHS-3 data set.

the possession of household assets and amenities. Such disagreement is not specific to the NFHS data (Gwatkin et al 2000). These could range from the need for including a particular asset or

amenity to the insignificance of a particular amenity subject to

location (for example, owning a mobile phone in an area where mobile networks are not available, or owning an animal-driven cart or tractor in an urban area). The first NFHS data based

attempt at classifying the Indian population into quintiles was

for studying the effect of wealth on education, where the authors themselves cautioned against rural-urban comparisons using a national quintile distribution (Filmer and Pritchett 2001). An agreement on the selection of variables in computing wealth quintiles is near impossible given the diversity of regional d evelopment in India.

Therefore our primary concern in this paper does not relate to the selection of variables for arriving at a proxy wealth status of the population or households, or on the procedure adopted for arriving at the Wifs in the NFHS-3 data. The issue is the manner in which wealth quintiles have been designated for comparison on outcome indicators, using the derived Wifs. This arises from an inappropriate wealth quintile classification that does not account for (1) the inter-state variations in wealth possession, and

(2) to the intra-state rural-urban divide in wealth possession. The wealth quintile classification adopted in the NFHS-3 data is based

0

–1

| | | | | | | | | | | | | Manipur

–1.5

she resides. Otherwise errors could creep in because of household size differentials across wealth or income status and across different Indian states. It is well known that the average household size is larger among the poor than among the rich. There are also considerable variations in household size across Indian states, depending on fertility transition and the existence of the joint family system.

Interstate Wealth Inequalities

Variations in the possession of household assets and better h ousing characteristics is analysed here, both among states and between rural and urban areas. This is based on the

Mean wealth index factor score

0.5

| | | | | | | | | | | | | | | |

Delhi Mizoram Haryana Meghalaya Arunachal Pradesh Bihar Jharkhand

–0.5

on a simple all-India cut-off point for demarcating the proportion of population in each quintile. This ignores the existence of state or regional patterns, and rural-urban divides within each state or region in relation to most of the health indicators under study (Bhat and Zavier 1999; IIPS and Macro International 2007). Besides, we know there are inter-state variations in the proportion of people below the poverty line (Planning Commission 2007), and differences in consumption expenditure patterns across states (NSSO 2006). By discounting such disparities, the wealth quintiles computed in the NFHS-3 mask real wealthrelated inequalities in the indicators under study. The level of masking depends on inter-state variations in the indicator under study and the rural-urban divide in a state. While analysing wealth or income-related inequalities in any indicator at the national and state level, one has to account for inter-state and rural-urban wealth differences to bring the desired robustness to the assessment.

Second, the Wifs is a household variable, and households should have been split into different quintiles using it. Following which, individuals residing in a particular household should have been assigned to its quintile. Instead of this, in the NFHS-3, the sample population was ranked according to the Wifs of the households in which they resided, and divided into five equal groups for demarcating the wealth score of each quintile. In preparing poverty estimates, the Planning Commission adopts the former procedure where it provides separate estimates of the proportion of households below the poverty line and the proportion of population below the poverty line (Planning Commission 1993). Since the cut-off line is determined on the basis of a household-level value or score, the classification of an individual is on the basis of the relative position of the household in which he or

Chart 1: Mean Weath Index Factor Score across Indian States as in NFHS-3, 2005-06

1.5

1

november 29, 2008

EPW

NATIONAL FAMILY HEALTH SURVEY-3

Table 2: Likelihood Position of a Household as Per National Wealth Quintile
and State-Specific Wealth Quintile Distribution Patterns
National Wealth	State-Specific Wealth Quintile (S)
Quintile (N)	Lowest (1)	Second (2)	Middle (3)	Fourth (4)	Highest (5)	Total
Lowest (1)	IN1S1	IN1S2	IN1S3	IN1S4	IN1S5	N1
Second (2)	IN2S1	IN2S2	IN2S3	IN2S4	IN2S5	N2
Middle (3)	IN3S1	IN3S2	IN3S3	IN3S4	IN3S5	N3
Fourth (4)	IN4S1	IN4S2	IN4S3	IN4S4	IN4S5	N4
Highest (5)	IN5S1	IN5S2	IN5S3	IN5S4	IN5S5	N5
Total	S1	S2	S3	S4	S5	TH

a verage WIFS for each household available in the NFHS-3 data. S tate-wise median values for the WIFS demonstrate the variation in possession of wealth in India. The median value of the WIFS varied from 1.186 in Delhi to -1.243 in Jharkhand. Households in Delhi, Goa, Kerala, Punjab and Mizoram are wealthier than those in other states. A lower median value of the WIFS indicates that households in Uttar Pradesh, Bihar, Madhya Pradesh, Orissa, Chhattisgarh and Jharkhand are lower down the scale, or worse in terms of possession of certain household assets and in terms of housing conditions. In the case of major states, their position as per the median WIFS is close to their position in terms of the proportion of population below the state-specific 2004-05 poverty line (Planning Commission 2007). This close correspondence between the wealth index and the economic index based on National Sample Survey (NSS) data indicates the effectiveness of the WIFS in assessing the economic status of households.

Rural-urban differences are striking at the all-India level as well at the state level, indicating that households in urban areas possess more wealth-determining household assets and have better housing than their rural counterparts. The rural-urban divide exists in all states but the extent of the divide varies from state to state. The rural-urban divide is least in Kerala, Manipur, Tripura, Delhi and Goa. The highest rural-urban differential in possession of wealth is seen in Jharkhand, Rajasthan, Madhya Pradesh, Maharashtra and Uttar Pradesh.

State-Specific, Rural-Urban Adjusted Wealth Quintiles

The variation noted above in the median WIFS demonstrates the need to account for rural-urban and state-wise variations while creating wealth quintile groups. This will give us a state-specific wealth quintile classification, which will be ideal for nationallevel comparisons of the various NFHS-3 indicators on the e conomic background of households.

To arrive at this in each state we have to rank and arrange each household according to the Wifs, separately for rural and urban areas. The sample of households surveyed in each state must then be divided into quintiles, again separately for rural and urban areas. In this way, we identify wealth quintile cut-off points at the household level and create wealth quintile groups for rural and urban areas of each state. Each individual in the sample is then assigned the wealth quintile level of the household in which he or she resides. This gives a quintile classification that take interstate and the intra-state rural-urban wealth differentials into account, hereafter termed as state-specific wealth quintiles.

Economic & Political Weekly

EPW

Now, we have a national wealth quintile (N), which is the one in the NFHS-3 report, and a state-specific wealth quintile (S) obtained using the above procedure. The national wealth q uintile is based on ranking all individuals in the NFHS-3 sample on the basis of their household WIFS and then dividing them into five equal groups. At the national level, 20% of the household population is in each quintile, but the same is not true across states (Table 2). This is due to interstate variations in the possession of wealth among households and differences in household size. In the case of state-specific wealth quintiles, the proportion of households in each quintile is 20% in each state, and so it also turns out to be 20% of households in each quintile at the national level.

When we compare the national wealth quintile and statespecific wealth quintiles, there can be three situations: (1) households which are in same quintile as per both, (2) households which are in a lower quintile as per the national wealth quintile but in a higher quintile as per the state-specific wealth quintile, and (3) households which are in a higher quintile as per the national wealth quintile but in a lower quintile as per the statespecific wealth quintile. Table 2 demonstrates these three situations. Situations (2) and (3) may affect the national and statelevel estimates of wealth inequalities using indicators from the NFHS survey. The households in the same national wealth quintile and state-specific wealth quintile are on the right diagonal in Table 2. Cases above this diagonal represent households which are in a lower quintile as per the national wealth quintile but in a higher quintile as per the state-specific wealth quintile. This situation is largely seen in states that are relatively poor in terms of wealth possession of its households and also in states where there are severe household-level inequalities in possession of wealth. Conversely, cases below the diagonal represent households which are in a higher quintile as per the national wealth quintile but in a lower quintile as per the state-specific wealth quintile classification. They occur in states that are comparatively better off in terms of the wealth possession of households.

Table 3 (p 80) presents the proportion of population in the same wealth quintile as per the national wealth quintile in the NFHS-3 report and the state-specific wealth quintile. The remaining proportion comprises individuals in a national wealth quintile either lower or higher than the state-specific wealth quintile. On the whole, only 35% of the sample is in the same quintile as per the two classifications. This indicates the extent of bias likely to creep in while computing quintile-wise wealth inequalities using the national wealth quintile. The estimates of urban areas are more

Unbound Back Volumes of Economic and Political Weekly from 1976 to 2007 are available.

Write to: Circulation Department,

Economic and Political Weekly

320, 321, A to Z Industrial Estate, Ganpatrao Kadam Marg, Lower Parel, Mumbai 400 013. Circulation@epw.in

NATIONAL FAMILY HEALTH SURVEY-3

state. Hindus are evenly distributed across wealth

Lowest Second Middle Fourth Highest Total Lowest Second Middle Fourth Highest Total

But it is not so when we look at the distribution

across state-specific wealth quintiles. In the case of

Muslims, the wealth quintile-based distribution

across quintiles is more uneven. According to the

respectively, while state-specific wealth quintiles

show it to be 17% and 23% respectively. In the case

of Sikhs, who are mostly concentrated in Punjab and

three or four other states, the national wealth quin-

tile figure suffers from a Punjab bias and the com-

The same is true of Buddhists and Neo-Buddhists.

North-East

quintiles. In the case of scheduled castes, the

d istribution of population across the two types of

wealth quintiles do not vary. However, one should

note that the population under the same quintile as

per the two types of classification may not neces-

wealth quintile distribution. Among all the caste/

tribe groups, the other backward castes are the

q uintile. This evenness increases if we adopt state-

specific wealth quintiles. The other castes, which

are essentially the forward castes, are better

equipped in terms of wealth possessions, and

Source: 1 IIPS and Macro International (2007); 2 Computed using the NFHS-3 data set.

likely to suffer than that of rural areas. In urban areas, the match between the national and state-specific wealth quintiles occurs in 24% of the population. This is least in the lowest quintile and universal in the highest quintile. Such a pattern is because urban households are better equipped in terms of wealth status indicators than rural households.

Inter-state variations in the match between the national and state-specific wealth quintiles vary between 20% in Kerala and 60% in Uttaranchal. A higher or lower mismatch is found in states representing the extremes in terms of wealth p ossession of its households. All these caution on the potential biases likely to surface with the use of the national wealth q uintiles in the NFHS-3 data set.

Socio-Religious Differentials in Wealth

An illustration of this bias can be seen if we examine the distribution of religious or caste groups on the basis of the two types of wealth quintiles. It is known that the social and religious

more than one third of this population is in the r ichest quintile as per the national wealth quintile d istribution. The relative advantage marginally declines when we adopt the state-specific wealth quintile c lassification.

Wealth Inequalities in Outcome Indicators

The analysis so far looked at the extent of interstate differentials in wealth inequalities, dissimilarities between the national and the state-specific wealth quintiles, and biases that creep in when studying wealth distribution across rural and urban areas, social groups, and religions using the national wealth quintile classification. We now demonstrate the flaws that may occur while analysing outcome indicators in accordance with the two types of wealth quintiles. Towards this, we use five selected indicators from the NFHS-3 data. Of the five, the following two indicators are positively related to the wealth index: the proportion of currently married women aged 15 to 49 years using contraception, and the proportion of children under five whose births were registered. The other three indicators are negatively related to the

november 29, 2008

wealth status: the proportion of women aged 15 to 49 years with quintiles in state-level analyses. If not, the real wealth-wise anemia, the proportion of men aged 15 to 49 years using any kind i nequalities in the NFHS-3 indicators may be misrepresented. It is of tobacco, and the percentage of men aged 15 to 49 years with a to be noted that the NFHS-3 state-level reports for Rajasthan and low body mass index. Orissa, which were released recently, have adopted the national

The difference between the lowest and highest quintile is very wealth quintile for examining state-level economic inequalities high when we use the national wealth quintile but much less in health indicators. Such an approach masks the real inequaliwhen we use state-specific wealth quintiles. In other words, ties in relatively wealthier states and exaggerates inequalities in

Table 4: Distribution of the de jure Population by Different Wealth Quintiles According to Religion and Caste of Head of the Household taken note of before the remaining state-wise NFHS-3

State-Specific Wealth Quintile2 summary reports are released. Since information

Lowest Second Middle Fourth Highest Total

Religion Hindu 20.7 20.6 20.2 19.2 19.2 100.0 18.5 18.8 19.8 20.7 22.1 100.0 one has to rethink the procedure of directly classi-

Muslim 18.9 19.7 20.6 23.6 17.2 100.0 19.9 23.5 21.3 20.2 15.1 100.0 fying the sample population into quintiles. For more

Christian 12.1 13.0 20.8 22.9 31.1 100.0 17.4 17.7 19.4 22.2 23.4 100.0 accurate results, the households in the sample sur-Sikh 1.5 6.3 12.6 26.9 52.7 100.0 12.1 13.4 18.1 23.4 32.9 100.0

Buddhist-Neo-Buddhist 18.2 21.0 18.0 20.5 22.3 100.0 28.7 22.9 22.5 15.6 10.3 100.0

Jain 1.6 1.8 1.6 8.3 86.8 100.0 1.0 3.0 8.5 24.2 63.3 100.0

Caste or Tribe Scheduled caste 27.9 24.6 20.8 16.6 10.2 100.0 28.7 23.0 21.9 16.3 10.2 100.0 belongs to.

Discussion

Having understood that there are inter-state differentials in household-level wealth inequalities and that they make a difference to the measurement of wealth-wise inequalities in health outcome indicators, one needs to be clear regarding roles of the national wealth quintile classification and the state-specific wealth quintile classification. The above analysis provides sufficient ground to argue the need for use of state-specific wealth

different states are at different stages of attainment when good health/status indicators are considered. Further, health is a state subject, and the efficiency of health interventions depends on the responsiveness of the intervention machinery in each state. So a national-level wealth quintile classification that divides the population into five equal groups will not be sensitive to analysing wealth-based inequities in health outcome indicators using the NFHS-3 data. In this context, we emphasise that using state- specific wealth quintiles in a national-level analysis using the NFHS-3 data will yield more accurate results. One could take a leaf out of the Planning Commission’s book where the poverty lines are first determined at the state level and then used for estimating the number of poor in each state. In this case, the rural and urban poverty lines in different states are determined using consumer price indices. These estimates provide the number of poor in the country as a whole. We suggest that the same procedure be followed when identifying cut-off points for including households within particular wealth quintiles. We do agree that the Planning Commission does this to adjust for interstate variations in the prices of commodities, while in the NFHS-3 it is inter-state variations in living standards due to disproportionate wealth possession that have to be adjusted for.

Given that there is a mix of community variables and h ousehold-specific variables in the construction of the Wifs, the

Table 5: Variations in Percentage of the de jure Population with Selected Characteristics across Different Wealth Quintile Classifications

EPW

NATIONAL FAMILY HEALTH SURVEY-3

use of a state-specific, rural-urban adjusted wealth quintile classification is further justified. In addition, the state-specific wealth quintile classification allows differential weights for the possession of a particular household asset in rural and urban areas and for the irrelevance of a particular household asset in a particular state.

This exposition is intended to caution NFHS-3 data users to make use of the WIFS to compute quintiles for the population they wish to stratify according to economic status. Further, a quintile classification is always subject to change depending on the purpose and context. While the WIFS is a household

Note References

attribute, its quintile distribution should be in accordance with the population, and the relevant outcome that one wishes to stratify in terms of this wealth score. An analysis based on the existing national wealth quintile classification not only makes a false pronouncement on wealth inequality, but also leads to misleading data on wealth-related inequalities in demographic and health outcomes. To conclude, the dis-aggregation according to wealth quintiles presented in the NFHS 3 does not providet a true picture of wealth disparity because it ignores both the prevailing differences among states and between rural and urban areas.

International Institute for Population Sciences (IIPS)

TT

hh

e Ne N

FF

HH

SS

-3 w-3 w

ee

aa

ll

tt

hh

ii

nn

dd

ee

x ix i

ss

bb

aa

ss

ee

d od o

nn

tt

hh

e fe f

oo

ll

ll

oo

ww

--

ing 33 assets and housing characteristics: household electrification, type of windows, drinking water source, type of toilet facility, type of flooring, material of exterior walls, type of roofing, cooking fuel, house ownership, number of members per sleeping room, ownership of bank or post office account, ownership of a mattress, pressure cooker, a chair, a cot/bed, a table, an electric fan, a radio/transistor, a black and white television, a colour television, a sewing machine, a mobile phone, any other telephone, a computer, a refrigerator, a watch or clock, a bicycle, a motorcycle or scooter, an animal-drawn cart, a car, a water pump, a thresher, and a tractor.

Bhat, P N M and F Zavier (1999): “Finding of the National Family Health Survey: Regional Analysis”, Economic & Political Weekly, 34 (2), pp 3008-32.

Filmer, D and L H Pritchett (2001): “Estimating Wealth Effects Without Expenditure Data – or Tears: An Application to the Educational Enrolments in States of India”, Demography, 38 (1), pp 115-132.

Gwatkin, D R, S Rutstein, K Johnson, R P Pande and A Wagstaff (2000): Socio-economic Differences in Health Nutrition and Poverty, HNP/Poverty Thematic Group of World Bank (Washington DC: The World Bank).

International Institute for Population Sciences (IIPS) and ORC Macro (2000): National Family Health Survey (NFHS-2), 1998-99 (Mumbai: IIPS).

The Review of Women’s Studies appears twice yearly as a supplement to the last issues of April and October. Earlier issues have focused on: Gender in Medical Education (April 2005); Gender Budgeting (October 2004); Women, Work and Family (April 2004); New Challenges (October 2003); Women, Work, Markets (October 2002); Women and Education (April 2002); Women: Security and Well-Being (October 2001); Women and Philosophy (April 2001); Reservations and Women’s Movement (October 2000); Women, Censorship and Silence (April 2000); Women and Ageing (October 1999); Gender Inequities: Focus on Tamil Nadu (April 1999).

For copies write to Circulation Manager

Economic and Political Weekly

320-321, A to Z Industrial Estate, Ganpatrao Kadam Marg, Lower Parel, Mumbai 400 013. email: circulation@epw.in