Post on 19-Mar-2020
1
Caste, Occupational Status and Wealth Inequality in India*
Vamsi Vakulabharanam
Queen’s College, City University of New York
Ajit Zacharias
Levy Economics Institute of Bard College
*PRELIMINARY DRAFT: PLEASE DO NOT QUOTE OR CITE WITHOUT PERMISSION Paper prepared for the Second ECINEQ Meeting, July 12-14, 2007 Berlin, Germany. Contact Information: Vamsi Vakulabharanam: vamsi.vakul@gmail.com Ajit Zacharias: zacharia@levy.org
2
Abstract
Using two rounds of nationally representative surveys on household wealth
conducted in India during 1991-1992 and 2002-2003, we analyze the relationship
between overall wealth inequality and caste divisions in India. The average minority (SC
or ST) person in India had a substantial disadvantage in wealth relative to the average
non-minority person. In the non-minority group, the forward caste Hindus was the clear
leaders in both the rural and the urban areas. The Other Backward Classes (OBC) and
Non-Hindus occupied positions that placed them noticeably above the minorities, but,
significantly below the forward caste Hindus. Using the ANOGi decomposition
technique, we estimate that between-caste inequality accounts for as much as 13 percent
of overall wealth inequality. The stratification parameters resulting from the procedure
indicate that the forward caste Hindus had a wealth distribution that had a fairly low
degree of overlap with the other caste groups, while the other caste groups had
significantly higher degrees of overlap with each other and the overall population.
3
I. Introduction
It is widely acknowledged among social scientists that caste is a persistent
determinant of power, economic inequality and poverty in contemporary India. Yet,
economics literature on caste relations in India is at best sparse, while non-economists
(mainly anthropologists and sociologists) have made substantial contributions to the
overall literature on caste (e.g. Beteille (2007), Gupta (2000) and Srinivas (2000)). This
gap has been acknowledged recently, and a call for greater attention to this axis of
differentiation has been made (Deshpande, 2000). This has given rise to an acceleration
in the production of quantitative studies on caste (e.g. Barooah (2005), Deshpande
(2001), Kojima (2006), Munshi and Rosenzweig (2006) and Sundaram (2006)).
The quantitative studies on caste can be divided into two broad categories. First,
there are studies that have used either large surveys (mainly National Sample Survey
(NSS) consumption and National Family Health Surveys (NFHS)) or fieldwork based
small sample surveys to show the evidence of caste differentials in consumption, income,
education, occupations, and development indices (e.g. see Deshpande (2001), Hasan and
Mehta (2006), Mehrotra (2006), Mohanty (2006), Srinivasan and Mohanty (2004), and
Sundaram (2006)). The near consensus in these studies is that the less privileged caste
groups tend to be worse off than the others on the measured indicators across the country,
although there are some regional differences. Second, using large survey data, other
studies have employed the Blinder-Oaxaca decomposition (or modifications of this) to
separate the structural differences (e.g. geographical, discrimination-based) among
households from the differences in endowments (physical and human) in the market place
4
(e.g. see Barooah (2005) and Kojima (2006)) that create caste disparities. Barooah, for
instance, using the NCAER survey showed that about a third of the income differentials
in India could be attributed to discrimination in the market place. Using the NSS
consumption surveys, Kojima showed that both the lower endowments of physical and
human capital possessed by disadvantaged groups as well as different structures of
income generation contribute equally to the disparities among caste groups. What is
remarkable across these studies is the persistence of systematic disparities among
households across different caste groups over long periods of time.
Our paper contributes to this literature by analyzing the relationship between
overall wealth inequality and caste divisions in India. There have been no studies on the
wealth disparities (as opposed to consumption or income disparities) within and among
caste groups on indicators, and how these disparities contribute to the overall inequality
in India. Wealth inequality is an integral aspect of economic inequality among persons at
a given point in time as well as across generations. Disparities in wealth can also translate
into disparities in economic security. For a substantial portion of the Indian population
that is dependent on agriculture, land is the major source of livelihood. Inequalities in the
quantity and fertility of land owned are a significant determinant of economic inequality
among households. Quality and quantity of schooling accessible to the children in urban
and semi-urban areas can vary positively with household wealth.
Caste is not the only determinant of social and economic status in India.1 It is
therefore important to compare the schisms generated by different social relations to
1 The 1911 Census of India contained a far more detailed breakdown of caste groups than available in post-Independence data. Tabulations based on the 1911 census for Uttar Pradesh (the largest Indian state, then known as United Provinces) showed that among the 42 castes considered in the census “each caste
5
understand within-group heterogeneity. Therefore, we also analyze the relationship
between overall wealth inequality and occupational status distinctions, a topic on which
no work based on nationally representative microdata exists to the best of our knowledge.
We construct a caste-occupation matrix to study the nature of the contemporary relation
between caste and occupational status. This allows us to examine caste disparities within
status groups and status disparities within castes.
The relationship between overall wealth inequality and social divisions (caste and
occupational status) is analyzed in this study using the Yitzhaki decomposition or
ANOGi (Yitzhaki, 1994; Frick et al. 2004). This allows us to separate the overall
inequality into within-group and intra-group components, rather than obtaining
conditional average effects of social divisions via regression-based decomposition
methods such as the Oaxaca-Blinder method. Furthermore, the overlapping parameters
estimated using our chosen method permits the distinction between caste-stratification
and caste-inequality. This is especially important in the context of ongoing debates in
Indian political economy about the questions of affirmative action and the so-called
“creamy layer.”2
The remainder of the paper is organized as follows. Section II describes the data
used and some of its problems. We also outline the definitions of the caste groups and
status groups used in the study. In the next section, we describe the patterns of wealth
disparities among castes and status groups. Our description is deliberately somewhat
detailed, as the data in question has not been explored from the angle of caste and status
contained landless labourers, cultivators as well as landlords.” (Chaudhury, 2004:1990). Economic differentiation within castes is the rule rather than the exception, then and now. 2 The “creamy layer’ refers to the emergence of an economically well-off group within castes whose average member is worse off relative to the rest of the population.
6
divisions before. The subsequent section (Section IV) presents the decomposition results.
Section VI contains our concluding observations.
II. Data and Concepts
The data used in this paper are from the two rounds of the All India Debt and
Investment Survey (AIDIS) conducted in 1991-92 and 2002-03. Wealth is computed as
the total household assets net of the indebtedness. Household assets are defined as
“physical assets like land, buildings, livestock, agricultural machinery and implements,
non-farm business equipment, all transport equipment, durable household goods and
financial assets like dues receivable on loans advanced in cash or in kind, shares in
companies and cooperative societies, banks, etc., national saving certificates and the like,
deposits in companies, banks, post offices and with individuals” (NSS 2005, p. 5). Debt
is defined as cash loans payable. In the absence of a better deflator, the Consumer Price
Index for agricultural workers is used to make the 1991 and 2002 rural wealth values
comparable across time. Similarly, the Consumer Price Index for industrial workers is
used to make urban wealth values comparable across time.
The unit of analysis for the whole paper is the household adjusted for its size.
That is, the household weight is multiplied by the household size to obtain a distribution
among persons. We use per capita wealth, household wealth divided by household size,
as the measure of wealth. The implicit equivalence scale assumed here is that there are no
“economies of scale” associated with wealth. (For the relative advantages and
disadvantages of this method, see Jayadev, Motiram and Vakulabharanam (2007)).
The definitions of caste groups are completely dictated by the data and do not
adequately reflect the complex and layered reality of caste in India. Both the AIDIS
7
rounds allow for the classification of the entire population into three groups, viz., the
Scheduled Castes or the Dalits (SC), Scheduled Tribes or the Adivasis (ST) and everyone
else whom we call Other Communities (OC). We term this classification Scheme I. The
2002-03 survey introduced the additional category of Other Backward Classes (OBC). 3
In addition, the category of religion was also enumerated. Cross-tabulating caste and
religion allows for the separation of OC into three distinct groups: OBC; Hindus who are
not SC, ST or OBC whom we call Hindu forward castes (FC); and Non-Hindus (NH)
who are not SC, ST or OBC. The 2002-03 survey, therefore, allows for the classification
of the population into five caste groups. We term this classification Scheme II. It should
be noted that the SC and ST individuals might belong to any religion.
A brief note is in order regarding the category of caste. Caste in India is defined
differently along the “Varna” and the “Jati” schemes. Varna scheme has four broad
groups - Brahmins, Kshatriyas, Vis, Sudras - and those people outside the Varna scheme,
ranked in a descending order of ritual status. Brahmins were traditionally associated with
the priestly and scholarly community. Kshatriyas were the ruling groups. Vis were
associated with those groups associated with trading, moneylending and retailing. Sudras
were the peasants and artisans. Among those outside the Varna scheme, the so-called
untouchables (the present day Dalits) were mostly associated with the rural landless
laboring community and the tribal groups (the present day Adivasis) were associated with
those living on the fringes of or outside the settled agricultural society. It is generally
agreed upon that this is a textual scheme defined by the Brahmins.
3 The determination of disadvantaged groups (OBCs) was made according to the 1931 Census. Many changes have occurred since in the socio-economic status of these groups, but this category does not reflect these changes, in part because subsequent censuses have not collected information on caste. The need for a new caste-based census is the subject of a heated debate in contemporary politics and the public sphere.
8
The Jati scheme is very different. There are thousands of Jati groups that vary
spatially and temporally in terms of their ritual rankings, socio-economic status and
occupations. It is also important to recognize that the caste system functions on the
ground along different Jati orderings thus creating a bewildering variety of them, as well
as a system that cannot be neatly captured by structural and closed systems that can be
deployed across space and time (For important renditions of the caste system, for e.g. see
Dumont (1970), Chatterjee (1993) and Gupta (2000)). Similarly, while certain
occupations are traditionally associated with certain caste groups (especially Jati groups),
this relation too is problematic given the significant flux in this relationship over time.
However, it has generally been the case that those outside the Varna scheme have tended
to be concentrated in the menial occupations, and this relationship too needs careful
examination in our times to see if independent India has been able to shake off some of
the entrenched caste based hierarchies. We therefore use the categories (Varna-based)
that are available in the surveys to make broad decomposition analyses of the overall
inequality in India.
We derive the variable Status in the following way. We use the NCO
classification of 1968 (available for both 1991 and 2002), landholdings and the household
type (whether self-employed or working for others) in dividing the households into six
categories viz. Urban elites, Urban middle groups, Urban manual workers, Rural Elites,
Rural middle groups and Rural workers. The NCO category for each household is the
principal occupation of the household based on the occupation of the primary earning
member of the household. All three urban categories are derived from the NCO
classification. Those people who are occupied in professional, managerial, technical and
9
administrative occupations in urban areas are classified as elites (NCO categories, first
digit: 0-2). Those occupied in sales and clerical occupations are classified as middle
(NCO categories, first digit: 3-4). Those working in low level service occupations (e.g.
hotels and households), urban agriculture, factories, construction and other manual
occupations are the urban manual workers (NCO categories, first digit: 5-9). The rural
areas are subdivided into agricultural and non-agricultural occupations. The agricultural
occupation is further classified according to the landholdings they possess. The
conversion factor between wet and dry land is 1.98 based on ICRISAT studies. Those
farmers who possess more than 10 acres of land are classified within rural elites. Those
between 2 and 10 acres of land are classified as rural middle. Those with less than 2 acres
of land are combined with the landless workers and classified as rural working people. Of
the non-agricultural population, those that are self-employed are part of the rural elites.
Those that are working on public works are part of the rural working people. Of the
residual population, the NCO classification that is used for urban elites is applied and
where a match occurs these are placed under the rural elites. The justification for this is
that some people are working for the government and other established occupations that
are not readily classified in the above mentioned agricultural and non-agricultural
populations. This scheme to devise the status variable in rural areas accounted for all the
households in the sample.
We further separate the rural areas from the urban as we believe that the wealth
accumulation and income generation dynamics vary significantly across this sectoral
division. Thus, all the caste and status groups used in our analysis are differentiated by
their rural/urban status.
10
The problems associated with the wealth data in the surveys are identified in the
literature (e.g. see Subramaniam and Jayaraj (2006) and Jayadev, Motiram and
Vakulabharanam (2007)). They deserve a brief recapitulation. There are basically four
kinds of problems with these data. First, wealth distributions tend to be concentrated at
the very top end. Unless special effort is made to over-sample the very wealthy, the
concentration of wealth tends to be under-represented. This will artificially reduce the
overall inequality. Second, there is a tendency among people of all wealth groups to
under-report their wealth holdings. This tendency to under-report is exacerbated as
wealth holdings rise. This will widen the gap between those with close to no wealth and
those that have some wealth. Third, the reported assets may not be correctly valued. It has
been found in India that the reported values of even recent transactions tend to be lower
than the market values. Given the lack of proper wealth based deflators, the wealth values
that are analyzed can be somewhat off the mark. Fourth, there is a tendency to hide
illegitimate wealth that will lead to under-counting of the assets owned by the wealthy.
Finally, there is a strong tendency to under-report the liability or debt. These problems
add up to a state where in populations belonging to the wealthier groups (more prevalent
among the non-SC/ST population) appear to hold lower wealth than they actually have
and the less wealthy groups (especially the SC/ST groups) report higher wealth than they
have. This will certainly reduce the overall inequality but it will also reduce the between-
caste inequality figures. These problems might be reflected in our findings.
III. Disparities in Wealth and Wealth Distributions
III.1. Disparities among Caste Groups
Most studies of economic inequality in India have used consumption expenditures
as the indicator of economic status. Our choice of wealth as the indicator of economic
11
status would be superfluous if consumption expenditures and wealth are distributed
similarly across individuals. While the two are correlated, the ranks of individuals in the
two distributions can be quite different (Table 1). If all individuals in a given quintile of
one distribution also belonged to the same quintile of the other distribution, then every
number on the principal diagonal of the matrix shown in Table 1 will be equal to 20 and
every off-diagonal numbers will be equal to zero. Inspection of the table shows that the
largest number occurs at the intersection of the top quintiles of the two distributions. This
number indicates that only about half (10.4/20 = 52%) of those in the top quintile of
wealth distribution were also in the top quintile of consumption expenditures. In other
quintiles, at least two-thirds of individuals in a given quintile of wealth distribution were
located in a different consumption quintile, with the third quintile showing the weakest
correlation in rankings.4
Let us now turn to examine disparities in wealth and wealth distributions among
caste groups. Since comparisons between the two years are possible only with the six-
group schema (ST, SC and OC), differentiated by their rural versus urban location), we
begin with a consideration of the estimates shown in the upper panel of Table 2. Between
1991 and 2002, the minority groups (SC and ST) have experienced rates of growth in
mean per capita wealth that were better than the majority group in both the urban and
rural areas. However, the medians tell a different story, especially for the ST. The wealth
of the average person in that group rose only 7 percent in the urban areas (as compared to
42 percent for the urban OC) and 21 percent in the rural areas (versus 25 percent increase
4 Indeed, the picture can be more complicated than suggested by the table since we expect rankings of individuals within quintiles also to be different, depending on whether wealth or consumption is used as the ranking variable.
12
for the rural OC). In contrast, the average SC person experienced a robust increase in
wealth of approximately 40 percent over the same period in both the urban and rural
areas.
In spite of the increases that did occur between the two years, the average
minority person still had a considerable wealth disadvantage in 2002 (see Figure 1).
Compared to the most numerous group, rural OC, the median wealth of rural ST and SC
were, respectively, only 49 and 46 percent; the relative position of the urban ST and SC
were somewhat better at 53 and 58 percent. In contrast, the urban OC had a median
wealth that was 21 percent higher than their rural counterpart. Comparison to the 1991
median values showed that the relative positions of the rural and urban SC were, in fact,
higher then than in 2002 while the relative positions of the rural and urban ST were
somewhat lower. The urban OC group also experienced a strong growth in their relative
position. If we were to compare the relative positions using mean, rather than median,
values then too we would obtain a similar picture of disadvantage for the minority
groups, with the exception of the urban ST whose mean wealth was 86 percent of the
mean wealth of the rural OC (as compared to only 58 percent, in terms of median
wealth).
As noted earlier, we are forced to treat the OC as a single category for comparing
the two years because the 1991 data does not allow further breakdowns of this group
along caste/religion lines. However, such a breakdown is possible in 2002 and the
structure of disparities among caste groups can be better seen in terms of what was
referred to earlier as Scheme II (Panel B, Table 1 and Figure 2). Irrespective of their
urban or rural location, the average OBC person had an amount of wealth that was a little
13
less than 90 percent of the average rural OC person. The average person in the group
labeled “Non-Hindu Others” and living in an urban area had as much wealth as the
average OBC; but those in the rural areas had significantly less, though more than that of
the average SC or ST person. The most advantaged subgroup in the OC group was the
Hindu forward castes (FC): the median wealth in the urban segment of this group was
twice as much as rural OC while its rural segment had a median that was 54 percent
higher than rural OC.
The ranking of the ten groups (in Scheme II) in terms of median wealth follow a
pattern that one might expect a priori: the Hindu forward castes are at the top (urban,
followed by rural). Immediately below them are the OBC groups and urban non-Hindu
others who have quite similar levels of median wealth. At the bottom, we have the
minorities (urban, followed by rural). The rural non-Hindu others occupied a place
immediately above the minorities and below everyone else.
If we were to use the mean values to rank the groups, the pattern shifts somewhat
(Figure 2). The top group—urban, Hindu FC—still maintained their lead and the rural
minorities held their status as the worst-off. Rural Hindu FC slipped to the third place,
with the second place taken by the urban, non-Hindu others. Rural non-Hindu others
occupied the fourth place, followed by the urban OBC, urban ST, rural OBC and then
urban SC. The re-ranking of the groups is an indication of the extent to which within-
group inequalities differ, a subject to which we shall return later.
Comparison of within-group distributions reveals that caste divisions and the
urban-rural divide act as distinct, yet interrelated influences on the overall wealth
distribution (see Table 3). The differences between the distributions of the individual
14
groups is plotted on the vertical axis in Figure 3 as ( ij ip p− ), which expresses the
deviation between the percentile cutoff of the jth group ( ijp ) from the overall percentile
cutoff ( ip ) at the ith percentile. Strikingly, only the Hindu FC stayed in the positive
territory throughout the distribution while the SC and ST groups stayed in the negative
territory throughout the distribution. The cutoff values for the former became
increasingly higher than the overall values (most markedly for the urban, forward caste
Hindus), while for the latter they became increasingly lower, as we move to higher
echelons of the wealth distribution. The other two groups, OBC and non-Hindu other,
displayed more complex patterns. Lower portions of the urban OBC and non-Hindu other
distributions had cutoff values that were below the cutoff values for the overall
distribution, but the higher portions had values that were higher, especially for the non-
Hindu others. The rural segments of these communities diverge from one another
markedly. While the bottom 60 percent of rural OBC enjoy higher than overall cutoff
values, the top of 40 percent in their distribution have cutoff values that are increasingly
lower. The opposite pattern can be observed for the rural non-Hindu others.
The direction and amount of the urban-rural disparity within caste groups varies
across the distribution. This can be illustrated by defining the following statistic for group
j at percentile i :
,u rij ij
ij rij
p pg
p−
= (1)
where the urban-rural gap in wealth is expressed as a percentage of the percentile
cutoffs ( p ) in the rural area for each caste group (The superscripts and u r , represent,
respectively, the urban and rural areas).
15
Estimates of the urban-rural gaps are shown in Figure 4 for selected percentiles,
with the bold horizontal reference line representing a situation of zero urban-rural
disparity. The wealth gap was in favor of rural individuals at the bottom of the
distributions of all castes. This is a reflection of the incidence of land ownership
(however meager the farm size might be) in the rural areas among the poor, in contrast to
the greater presence of property-less individuals among the urban poor, irrespective of
their caste identity. Notable differences existed among the castes in the percentile point at
which their respective curves crossed above the zero line. At one extreme was the non-
Hindu others for whom the switch favoring the urban areas occurred at the 20th
percentile; and, at the other extreme, the switch occurred only at the 50th percentile for
the OBC. The variation in the amount of urban-rural disparity among the castes appeared
to be much smaller at any given percentile point below the zero-line, i.e. when the
disparity was in favor of the rural individuals. Above the zero-line, when the disparity
turns in favor of the urban persons, the amount of disparity (at any given percentile point)
among the castes appeared to vary much more. Clearly, the evidence suggests that the
wealth advantage enjoyed by the urban individuals within every caste becomes higher at
the higher percentiles, with the non-Hindu others standing out as a clear exception to this
rule because the disparity in favor of the urban individuals in this group declined after the
70th percentile. The urban advantage skyrockets within the ST group in the top portions
of the distributions, a result consistent with the well-known fact that the rural tribal areas
fall among the most economically backward areas in India.
We now revert to Scheme I in order to examine whether any significant
differences could be found among the groups in terms of the changes in wealth that
16
occurred between 1991 and 2002 across the entire distributions (Figure 5). Urban ST was
the only group in which some of the percentile cutoffs in 2002 were roughly the same as,
or lower than their 1991 levels. In contrast, the bottom half of the urban SC group
generally saw a much higher boost in their wealth levels than their counterparts in the
other groups. For the upper-middle portion (roughly from the 50th to 80th percentile), the
urban OC group experienced much faster growth than their counterparts in other groups.
The sharpest increases in wealth between 1991 and 2002 among the top 20 percent in all
groups occurred for the urban minority groups. A negative correlation between the initial
amount of wealth and the subsequent gain could be found in the bottom half of the urban
SC and OC groups as well as the rural ST and SC groups. In fact, the schedule for the
rural SC group slopes downward to the right almost throughout the distribution. Finally,
the rural OC group displayed the most stable pattern: their schedule remained largely flat
for most of the distribution. The overall picture of changes across the distributions
suggests a pattern of wealth accumulation that is not heavily biased in favor of those at
the top within each caste group, with the exception of the urban ST.
III.2. Caste and Occupation
Heterogeneity within caste groups is well known to social scientists of modern
India. A perennial question has been whether class disparities, understood in terms of
occupation and ownership, “trump” caste disparities. As shown in Table 4, people of
every caste can be found in all status groups. However, the distribution of members over
status groups differed markedly among castes. Within rural areas, the Hindu FC seem to
have roughly the same proportion of their members in all status groups; the same pattern
also holds, albeit in a weaker form, in urban areas. For the other communities, the
17
distribution of persons is much more polarized with the Urban Manual (low-grade service
occupations, nonskilled factory workers etc) and Rural Working (consisting mainly of
landless agricultural workers and, small and marginal farmers) making up the majority in,
respectively, the urban and rural areas. However, the combined share of these two groups
was considerably higher for the SC at 76 percent in 2002, while the other three castes—
ST, OBC and Non-Hindu Others—had around 60 percent and the Hindu FC had only 39
percent.
Urban middle class membership (mainly sales and clerical workers etc.) was
higher, in percentage terms, among the Hindu FC and non-Hindu Others (10 to 12
percent in 2002). Among the other groups, it was under 6 percent. The OBC and Hindu
FC had about the same proportion of their population in the Rural Middle group (20
percent in 2002), which was lower than that of the ST (26 percent), but higher than that
of the SC and Non-Hindu Others (10 and 13 percent respectively). The SC, ST and OBC
communities had only less than 4 percent of their members in the urban elite (consisting
mainly of those living in households where the principal earner is a professional or
managerial employee) while the forward caste Hindus had 10 percent and non-Hindu
Others had 7.1 percent in 2002. A notable change in the status structure in India between
1991 and 2002 was the shrinking of the Rural Middle and Rural Elite groups and the
growth of the Rural Working in all three groups between 1991 and 2001, suggesting
greater of proletarianization in the countryside and greater migration to the urban areas.
We next turn to our estimates of mean values of wealth for the status groups
cross-tabulated with caste (Table 5, panel A). Looking first at within-caste disparities in
status, it is clear that the Rural Working group ranks at the bottom and the Urban Elite at
18
the top of the status ladder in all castes, with the exception of Non-Hindu Others in 2002.
The Urban Elite also experienced the fastest growth in wealth in the ST and SC between
1991 and 2002, but, surprisingly, the lowest growth in the OC. The wealth gap between
status groups can be expressed in terms of the ratio of the average wealth of each group
to the wealth of Rural Working (Table 5, panel B).
Expressed in these terms, the rankings of the different status groups based on
mean values followed a clear pattern of wealth hierarchy in all castes. In the urban areas,
the professional and managerial employees are at the top of the wealth hierarchy,
followed by semi-skilled white-collar workers in the middle and unskilled workers at the
bottom. Big landlords along with the professional class have the highest average wealth
in the rural areas, followed by farmers with medium-sized landholdings and semi-skilled
workers in the middle, and landless agricultural workers and other unskilled workers at
the bottom.
If we consider the urban and rural groups together within each caste, rankings
based on mean values do not follow the “neat” pattern that can be observed for the OC
group as a whole: Elites at the top, middle-status groups in the middle, and the mass of
unskilled workers and poor peasants at the bottom, with the urban residents within each
group possessing greater wealth than their rural counterparts. In the SC and OBC groups,
the Urban Middle has less wealth than the Rural Middle; and, in the ST group, the Rural
Middle had only about the same wealth as the Urban Manual; and, for Non-Hindu Others
the Rural Elite had about 11 percent more wealth than the Urban Elite. We speculate that
in the first and third cases the presence of a relatively prosperous middle- or upper class
19
peasantry could explain the apparent anomaly, while the “ST anomaly” requires further
investigation.
Our discussion above has focused on status disparities that exist within castes.
But, there are also considerable caste disparities within status groups. These disparities
can be expressed in terms of the ratio of mean wealth of each caste to the OC group and
the resulting estimates are presented in Table 5, panel C. Inspection of the numbers
shown in the ST and SC columns, shows that the OC had substantially more average
wealth than the SC and ST in every status group. For the ST group, the extent of wealth
disparity was much lower for the Urban Elite in 2002 as compared to 1991 (0.62 versus
0.45), while it was much higher for the Urban Middle (0.60 versus 0.74). The standing of
other status groups in the ST improved relative to their counterparts in the OC. The
Urban and the Rural Elites in the SC group saw a divergence in their fortunes relative to
their OC counterparts as the ratio of mean wealth values rose from 0.36 to 0.45 for the
first group while it declined from 0.46 to 0.41 for the second. Unlike the Urban Middle in
the ST, the Urban Middle in the SC improved relative to the OC Urban Middle between
1991 and 2002. Apart from the top two urban status groups, every other status group in
the SC community either stagnated or declined relative to their OC counterparts.
Disaggregating the OC in 2002 revealed that the OBC in every status group also
suffered from a wealth disadvantage relative to the OC group as a whole although to a
much lesser extent than the SC and ST. Hindu FC emerge as the leader in terms of
average wealth in every urban status group, while in the Rural Elite and Middle groups,
the Non-Hindu Others dominated. Closer examination showed that the Sikhs (primarily
in Punjab), Jains (primarily in Karnataka and Maharashtra), and Christians (primarily in
20
Kerala and the northeastern states) are substantially ahead of Hindu FC in terms of wealth
in both the groups.
Since the salience of occupational status as an axis of wealth disparities emerges
clearly from the above discussion, it is useful to examine the disparities among the
occupational status groups for the nation as a whole, leaving aside the caste divisions
(Table 6). It is at once evident that the “neat” pattern of wealth stratification that was
noted earlier for the OC group as a whole earlier holds nationally too, if we use the mean
values to rank the groups. The same pattern did not hold strictly in terms of the median
values, because the Rural Middle has a somewhat higher (13 percent) median wealth than
the Urban Middle. Status disparities in wealth encompass a much larger range than caste
disparities in wealth (in terms of mean or median values). This should not be surprising
given the wealth disparities within castes along the status dimension that was discussed
earlier.
IV. Decomposition of Wealth Inequality
IV.1. The Yitzhaki Decomposition
The picture of caste and status disparities in India sketched out so far can be made
richer by relating them to an analysis of overall wealth inequality. Using the tools of
decomposition analysis allows us to examine, for a given level of overall inequality,
whether the groups formed on the basis of caste are more equal in terms of their average
wealth than the groups formed on the basis of occupation. It would also allow us to
develop summary measures that would express how demarcated in terms of its wealth
holdings a particular caste or status group is from another group or from the total
population. Further, wealth inequality exists within every group irrespective of whether
the “grouping” is done in terms of caste or occupational status and it would be useful to
21
have comparisons of the degree of inequality among groups. The method of Gini
decomposition developed originally by Shlomo Yitzhaki offers a unified framework for
addressing these issues.
Let G be the Gini coefficient of wealth. The Yitzhaki decomposition allows us to
separate G into inter-group inequality ( bI ) and a remainder ( rI ) that can be interpreted
as intra-group inequality (Yitzhaki, 1994):
b rG I I= + (2)
The amount of inter-group inequality is:
2cov( , ( )) ,i oib
F yI μμ
= (3)
where y is wealth, μ is mean wealth for all persons, iμ is mean wealth for group
i , and ( )oiF y is the mean rank of group i , i.e., the average position of the members of a
group in the overall wealth distribution.5 Thus, the amount of inter-group inequality is
twice the covariance between the mean amounts of wealth and mean ranks of groups
divided by the mean wealth for all individuals.6
The remainder term is calculated as:
,r i i ii
I s G O=∑ (4)
5 For example, if the mean rank is 0.25 for SC then the average SC person’s position in the wealth distribution for all persons will be at the 25th percentile. 6 In contrast, in the standard decomposition the between-group component would be equal to twice the covariance between the wealth of each group and the rank of each group’s mean wealth divided by overall mean wealth. The Yitzhaki decomposition takes into account the ranking of each individual within each group in the overall distribution.
22
where is is the share of group i in aggregate wealth, iG is the Gini coefficient of the
wealth distribution within group i , and iO is the overlapping index for group i . The
Yitzhaki decomposition provides group-specific measures of overlapping unlike the
standard decomposition of the Gini where only a summary measure of overlapping by all
groups can be obtained. The index of overlapping proposed by Yitzhaki is a measure of
the degree to which the range of wealth in each group overlaps with the range of wealth
for all persons. Overlapping can thus be seen as the opposite of stratification: the higher
the amount of overlap between a group and the population, the less stratified they are as a
group in terms of wealth (Yitzhaki, 1994, pp.148–149). This feature of the decomposition
is crucial for us since our objective is to ascertain the extent to which castes occupy or do
not occupy different segments of the wealth distribution.
The amount to which group i overlaps with the overall distribution is defined as:
cov ( , ( )) ,cov ( , ( ))
i oii
i i
y F yOy F y
= (5)
where ( )oiF y is the function that assigns to the members of group i their ranks in the
overall distribution, iF is the function that assigns to the members of group i their ranks
in the wealth distribution within that group, and covi indicates that the covariance is
according to the distribution within group i .7 The minimum value of iO is given by the
share of group i in the population and its maximum value is equal to 2. When the index
equals the minimum possible value, it suggests that the group in question is a perfect
stratum, i.e., it occupies an exclusive segment of the overall wealth distribution. If a
7 In theory, the functions are actually cumulative distribution functions. However, when working with actual samples, the cumulative distribution function is estimated by the rank of the observation and hence our description of the functions as rank-assigning functions (Yitzhaki 1994, p.149, n.1).
23
particular group has a range of wealth that coincides with the range of wealth of all
persons then the index will be equal to 1. Finally, if the index is greater than 1, the
distribution of wealth within the group is much more polarized than in the overall
distribution. This can happen if the members of the group constitute two strata, one that
has higher and the other that has lower wealth than μ , the average wealth of all
individuals in all groups (Yitzhaki and Milanovic, 2002, pp.162–163).
The index of overlapping defined in equation (4) is constructed from indexes that
indicate the amount by which a group overlaps with each of the other groups:
i i j jij i
O p p O≠
= +∑ (6)
where ip is the share of group i in the total population and jiO is the index of
overlapping of group j by group i . Since the overlapping of a group by itself is equal to
1 by definition, its contribution to iO is equal to its relative size. The index of
overlapping of the overall distribution by a group is the weighted sum of overlapping of
each of the other groups by that group, with the relative size of each group serving as the
weights.
In turn, the group-by-group overlapping indexes are calculated as:
cov ( , ( ))
,cov ( , ( ))
i jiji
i i
y F yO
y F y= (7)
where jiF is the function that assigns members of group i their ranks in the
wealth distribution of group j . The index jiO indicates the extent to which the wealth of
individuals in group j falls in the range of wealth of individuals in group i ; the higher
the fraction of group j that falls in the range of group i , the higher will be the value of
24
jiO ; and, for a given fraction of group j that falls in the range of group i , the closer the
wealth of the individuals in that fraction are to the mean wealth of group i , the higher
will be the value of jiO . The index can take values between 0 (no overlap) and 2. Perfect
overlap occurs when the index equals 1, indicating that the rankings of members of group
i produced by iF and jiF are identical (Yitzhaki, 1994, pp.150–152).
IV.2. Within-Group vs. Between-Group Inequality
We now turn to the results of the Yitzhaki decomposition for our data.8 It is useful
to begin with the estimates of within-group and between-group inequality under the
alternative “grouping” variables—caste and occupational status (Table 7). Overall wealth
inequality showed very little change between 1991 and 2002. The share of within-group
and between-group inequality in overall inequality also remained roughly the same
between the two years. With either caste or status as the grouping variable, within-group
inequality (the rI term in equation (2)) accounted for the bulk of overall inequality in
both years. Yet, the share of between-group inequality in overall inequality was at least
twice as high (around 25 percent) when individuals were differentiated according to
occupational status rather than caste. Interpreted in light of equation (3), it appears that
taken as groups, caste groups are more equal than status groups, i.e., the covariance
between the average amounts of wealth and average ranks are twice as high for status
than caste.
The dominating effect of within-group inequality on overall inequality is
consistent with our earlier discussion of the extent of status disparities within each caste
8 Decomposition of the Gini by groups was performed using the ANOGI module for STATA (Version 9).
25
and the caste disparities within each status group. Of course, in addition to status and
caste, there are other wide variations in the characteristics of household members that are
also expected to contribute to wealth differentials within castes—age, education, industry
of employment, and number of earners in the household, to mention a few. Additionally,
we would also expect product mix and fertility, among other things, also to have effects
on the wealth of farmer households. In 2002, we found the share of within-group
inequality to be somewhat lower (87 percent) under the more elaborate Scheme II (10
groups as compared to 6 in Scheme I). Since the subgroups included in the OC group are
themselves quite different from each another in terms of their average wealth and
distributions, the modest increase in the share of between-group inequality under Scheme
II was not out of line with our expectations.
IV.3. Within-Caste Inequality and Overlapping
The results from decomposing the remainder term along caste lines are shown in
Table 8. Looking first at the column of overlapping indexes for caste groups under
Scheme I reveals that the Urban ST and SC groups are hardly homogenous groups. Both
have values exceeding 1 for their overlapping indexes, indicating that there might be two
distinct strata, one quite rich and the other extremely poor, within each of these groups.
This is most striking in the case of the Urban ST in 2002. The overlapping index for the
Urban OC in 1991 is almost 1 in 1991 and slightly lower in 2002, indicating the close
similarity between their distribution function and the distribution function for the entire
population. However, when these values are reckoned against their share in population
(the minimum value that can be taken by the overlapping index), they appear far more
modest than the two urban minority groups.
26
The rural groups in Scheme I had lower values for their overlapping indexes than
the urban groups, a result that is not surprising in light of the considerable rural-urban
wealth gaps that were discussed earlier (Figure 4). Once again, when compared relative
to their shares in population, the rural OC group had substantially lower degree of
overlapping than the rural, minority groups. Estimates for the subgroups included in OC
in 2002 (Scheme II) showed that the Hindu FC was the group with the lowest amount of
overlapping among all groups, while the Non-Hindu Other rural and urban groups took,
respectively, the second and third places in terms of overlapping (the Urban ST was first,
as noted above). The higher degree of overlapping by the rural Non-Hindu Others as
compared to their urban counterparts was an exception to the pattern observed for the
other groups, but consistent with our earlier finding of relatively rich Non-Hindu Others
dominating the Rural Elite and Rural Middle status groups.
Within-caste inequality is the highest (above 0.670) among the urban ST, urban
Non-Hindu Others, and rural Non-Hindu Others, which as we noted above were also
characterized by overlapping indexes above 1. Excluding the latter, the other rural groups
all had roughly similar amount of within-caste inequality (0.560 to 0.580). The urban SC,
OBC and FC groups occupied an intermediate position (0.610 to 0.660) in within-caste
inequality. Comparisons against the 1991 values show that the only groups that saw
substantial change in wealth inequality were the ST groups, for whom there was a big
increase in inequality. This is was especially true for the urban ST and is consistent with
our earlier finding about the big increases in the percentile cutoffs in the upper tail of the
ST wealth distribution (Figure 5). Considered in conjunction with the jump in the
overlapping indexes, it appears that there was the emergence of a “nouveau rich” and
27
growing income polarization within the ST groups. We will return to the discussion of
within-group inequality in conjunction with the estimates for the status groups later.
Apart from the index of overlapping for each group with the overall population,
the Yitzhaki decomposition also allows us to estimate pair-wise indexes of overlapping
among the groups (equation (7)). The estimates of the resulting overlapping matrix using
Scheme II for 2002 are shown in Table 9 (Panel A). The reference group (the caste
represented by the subscript i in the overlapping index jiO ) is shown in the rows of the
table; other groups are shown in the columns (the castes represented by the subscript j ).
Urban and Rural FC have the highest degree of overlap with one another and much lower
degree of overlap with all others. Thus, their status as the groups with the lowest degree
of overlapping with the population did not hold for the pair-wise comparison.
Overlapping of each of the other groups by the urban ST, SC, OBC, and the Non-Hindu
Others groups are generally high. In contrast, the overlapping of each of them by the
Hindu FC was much lower.
The reason behind this apparent discrepancy can be understood by considering the
overlapping between the urban ST and urban FC. The overlapping of urban ST by urban
FC is only 0.716. This reflects the fact there were relatively few urban ST individuals in
the urban FC wealth range. Consequently, the ranks of urban FC individuals, when each
of them are placed in the wealth distribution of urban ST, did not differ much from each
other for a large number of them, thus reducing the size of the covariance in the
numerator of equation (7). On the other hand, the overlapping of urban FC by urban ST
was much larger at 1.05, reflecting the fact that there are relatively more urban FC
individuals in the urban ST wealth range
28
The overlapping of rural ST and SC by each of these groups was higher than the
overlapping of their urban counterparts by the same groups. For example, the overlapping
of rural ST by rural SC was 1.02 while the overlapping of urban SC by rural SC was
lower at 0.92. Further, the overlapping of rural OBC, FC, and NH groups by,
respectively, the rural SC and ST were higher than the overlapping of urban OBC, FC,
and NH groups (e.g. the overlapping of rural FC by rural SC was 0.934 as against only
0.739 for urban FC). This suggests that the distributions of rural ST and SC are more
similar to each another than to the members of their own community in the urban areas
and that they have at least some members with amounts of wealth that match the wealth
of wealthier individuals from the rural residents of other communities.
However, the rural-urban patterns of overlapping are quite different for the rural
OBC and FC groups. Their wealth distribution is more similar to the urban residents of
their own communities than to the SC or ST in the rural areas. For example, the
overlapping of rural SC by rural OBC was only 0.831 while the overlapping of urban
OBC by rural OBC was higher at 0.908. Similarly, the overlapping of rural ST by rural
FC was quite low at 0.625 compared to the overlapping of urban FC by rural FC that
stood at 0.903. The overlapping relation between the rural OBC and rural FC as well as
that between the rural NH and rural FC mirror the relationship between urban ST and
urban FC that was discussed above.
The index of overlapping is sensitive to extreme values because it depends on the
ranks and amounts of wealth of individuals in each caste. Hence, an examination of the
ranking of one caste in terms of another is instructive. Such an exercise can answer the
following type of question: at what percentile of the forward caste wealth distribution is
29
an average SC person located? The average rank of each caste in the distribution of other
castes can be displayed in a matrix of ranks. Along the row labeled “Urban ST”, for
example, we can read off the average rank of an individual in that group in the wealth
distribution of each of the other groups. Since the ranks are normalized to lie between 0
and 1, the average rank of a group in its own distribution will be 0.5 (i.e. the 50th
percentile).
The matrix of ranks for caste groups under Scheme II is shown in Table 9 (Panel
B). Forward castes clearly dominate other groups in terms of this indicator too. If we look
at the entries under the column labeled “Urban FC”, it is evident that the average rank of
all groups except their rural counterparts are placed below the 40th percentile of the urban
FC wealth distribution; the rural FC’s average rank was at the 45th percentile. Similarly,
the entries in the “Rural FC” column were also below the 40th percentile for all groups,
except, obviously, their urban counterparts.9 Viewed from another angle, this means that
the average rank of all the other groups were at their lowest levels when they were placed
in the distribution of forward castes. The most numerous of the groups, the rural OBC,
had a mean rank above the 50th percentile in the distributions of all SC and ST groups
and close to the 50th percentile for the Non-Hindu Others and urban OBC distributions.
The average rural ST and SC ranks were below the 40th percentile in the
distributions of all other non-ST/SC groups, except for the Non-Hindu others where their
ranks were at the 41-42nd percentile, and slightly below the middle in the distributions of
their urban counterparts. Even though they had high values for their overlapping index,
the average urban ST and SC ranks were in the bottom half of the distribution of all other
9 The sum of the average rank of group j’s rank in group i's and the average rank of group i's rank in group j’s distribution will be equal to 1.
30
groups, except that of their rural counterparts where they were slightly above the middle.
Their ranking was the lowest (roughly at the 30th percentile) in the FC distributions,
somewhat higher (roughly at the 40th percentile) in the OBC distributions, and the
highest (roughly at the 45th percentile) in the NH distributions.
IV.4. Within-group inequality and status stratification
As we pointed out before, the amount of intra-group inequality was lower when
the decomposition was done using status rather than status groups. Further breakdown of
the portion of overall inequality accounted for by inequality within status groups is
shown in Table 10. It may be recalled that when we used the same number of caste
groups (i.e. six, under Scheme I), the values of the overlapping indexes were all above
0.9 and three of them were equal to, or above 1. In contrast, the overlapping indexes for
status groups exceeded 0.9 in only three cases (namely, for the Urban Manual, Urban
Middle, and Rural Working groups). The first of these groups, with an overlapping index
exceeding 1, appears to have a polarized distribution, similar to what we found for the
Urban SC and ST groups. Relative to their population shares, the Urban Elite and the
Urban Middle displayed comparatively high amounts of overlap, while the rural status
groups, especially the Rural Middle, lower amounts of overlap. This result is similar to
the relatively lower degree of overlapping that we found for the rural caste groups and
serves as a further evidence of the rural-urban divergence in wealth distributions.
Inequality within the Urban Elite and Middle groups were substantially lower in
2002 than in 1991. About 90 percent of the members of these two groups belonged to the
OC group, for whom, as we saw earlier, inequality declined by 1.7 Gini points. While the
share of ST and SC in these groups increased slightly over the period (9.1 to 9.6 percent),
31
and the Gini coefficients for these groups rose, these changes ended up reducing the Gini
coefficients for the Urban Elite and Middle Groups. A partial explanation for the
reduction might lie in the fact that the ST and SC were largely increasing the wealth
shares of the bottom rungs of the wealth distribution in the Urban Elite and Middle
groups. In contrast, the increase of the ST and SC shares in the ranks of the Urban
Manual group was much higher (from 10.8 to 12.3 percent) and the additional ST and SC
individuals were most likely drawn from the bottom portion of the urban ST and SC
distributions. We know (from Table 3) that the percentile cutoffs for the urban ST and SC
groups were much lower than those of the other groups in 2002 and hence will have the
effect of worsening the wealth distribution within the Urban Manual group. Indeed,
factors other than the greater presence of ST and SC in the Urban Manual population
must have exerted upward pressure on the Gini coefficient of the Urban Manual group
(an increase of 2.7 Gini points)—a set of issues that fall outside the scope of our study.
The increase in Gini coefficients between 1991 and 2002 for all the rural status
groups appears to be unrelated to any change in the SC and ST shares in their respective
populations since these shares actually declined or remained stable over the period. But,
the distribution of caste groups across the status groups did seem to affect the inequality
trends. The evidence presented in Table 8 showed that the rural OC group did experience
a modest increase of 1.4 points in their Gini coefficient, while that of the rural ST shot up
by 4.2 points. These two groups make up roughly 88 percent of the population in both the
Rural Elite and Rural Middle. Therefore, the increase in the Gini coefficients of the latter
groups appears to be a function of changing distribution within the status groups rather
than due to changes in their caste mix. It appears that the decline in the Gini coefficient
32
for the rural SC by 1.4 points might have had a moderating effect on the Gini coefficient
for the Rural Working group since the share of the rural SC is the highest (28 percent) in
this group (and that of the other two caste groups, ST and OC, the lowest) among all the
rural status groups.
The overlapping matrix for status groups in 2002 is shown in Table 11 (panel A).
Compared to the matrix for caste groups, the entries here are generally smaller, indicating
the lower degree of overlap among status groups relative to caste groups. The largest
single group, Rural Working, had fairly low values along its row, excepting the Rural
Middle, with which it had an overlapping index of 1. However, the overlapping of Rural
Working by Rural Middle is the lowest entry in the matrix, 0.440. The overlapping of
Rural Working by other groups is also fairly low with Urban Manual as the sole
exception (0.985). Rural Middle, the second largest status group in our scheme, also
displayed comparatively low values and, unlike other groups, it has no entry greater than
1 along its row. The overlapping of the rural and urban elites by each other were quite
high, but their overlapping of Urban Manual and Rural Working considerably low. With
the largest number of entries greater than 1 and the corresponding Oij's much smaller, the
Urban Middle and Urban Manual groups appear to have polarized distributions in the
sense of containing a number of distinct strata within them.
Our estimate of the matrix of ranks for the status groups in 2002 is shown in
Table 11 (panel B). Once again, some really small values (compared to what we found
for caste groups) can be observed along the rows of Rural Working and Urban Manual
groups (excluding their mutual intersections). The average rank of the Rural Working in
the wealth distribution of the Urban Elite was only at the 20th percentile while that of the
33
Urban Manual in the same distribution was somewhat higher at the 28th percentile. Ranks
of these two groups in the Rural Elite distribution were very similar to their ranks in the
Urban Elite distribution. The Urban and Rural Elites had ranks quite close to the middle
in each other’s distribution similar to what we observed for the urban and rural forward
caste Hindus.
V. Conclusion
The average minority (SC or ST) person in India had a substantial disadvantage in
wealth relative to the average non-minority person in 2002. In the non-minority group,
the forward caste Hindus was the clear leaders in median wealth in both the rural and the
urban areas. The Other Backward Classes (OBC) and Non-Hindus (NH) occupied
positions that placed them noticeably above the minorities, but, significantly below the
FC in terms of median wealth values. In a worrisome trend, the relative median wealth of
the rural and urban ST was in fact lower in 2002 than in 1991. A similar picture of
minority disadvantage and forward caste advantage was evident throughout the
distributions in terms of gaps in percentile cutoffs. Estimates of the matrix of ranks for
caste groups also confirmed the existence of sizeable wealth gaps between the forward
castes and everyone else. Considered in conjunction with the findings documented in
other studies regarding the considerable shortfalls of the average minority person in
consumption, education, and development indices, the picture that emerges is one of
comprehensive and persistent minority disadvantage in contemporary India.
Our decomposition analysis showed that inequality between castes (between-
group inequality) accounted for as much as 13 percent of overall wealth inequality in
2002. The less elaborate caste schema (3 instead of 5) that we were forced to use in 1991
34
due to data limitations resulted in a lower share of between-group inequality (8 percent).
The major determinant of between-group inequality was the large gaps between
minorities (especially rural) and the forward castes (especially urban) in average wealth.
It would be interesting to compare this result to the results that would from using other
variables to classify the population (e.g. age or education). However, it is reasonable to
expect that irrespective of the “grouping variable” used, the share of within-group
inequality is likely to be the dominant factor in overall inequality. There are, inevitably,
other wide variations in the characteristics of households that, when taken together, are
likely to contribute more than the classifying variable itself to wealth differentials within
any group.
Results from our decomposition analysis also indicated that the forward caste
Hindus had fairly low degree of overlapping with the overall population and, especially,
with the minorities, i.e. they are more stratified in terms of their wealth distribution. The
other groups show fairly high degree of overlapping with the overall population as well
as with each other. Evidence of a polarized distribution could be detected for four
groups—Urban ST, Urban NH, Rural NH, and Urban SC (overlapping index greater than
1). The first three of these groups had within-group inequality that was much higher than
the overall inequality while the Gini coefficient for the last group was lower than the
overall Gini coefficient.
With the exception of the Rural SC, the other three minority caste groups, Urban
ST, Rural ST and Urban SC witnessed increases in within-group inequality between 1991
and 2002. This was especially striking for the ST. Given its occurrence along with the
deterioration in the median wealth of the group compared to the rest of the population, we
35
might be witnessing the emergence of a “nouveau rich” and growing income polarization
within the ST groups.
36
References
Barooah, Vani, K. 2005. “Caste, Inequality and Poverty in India,” Review of
Development Economics, 9(3), pp. 399-414.
Beteille, Andre. 2007. “Classes and Communities,” Economic and Political
Weekly Vol. 42(11), pp. 945-950.
Chatterjee, Partha. 1993. The Nation and its Fragments: Colonial and
Postcolonial Histories. Princeton: Princeton University Press.
Chaudhury, Pradipta. 2004. “The ‘Creamy Layer’: Political Economy of
Reservations,” Economic and Political Weekly May 15, pp. 1989-1991.
Deshpande, Ashwini. 2000. “Recasting Economic Inequality,” Review of Social
Economy, Vol. LVIII (3), pp. 382-399.
Deshpande, Ashwini. 2001. “Caste at Birth? Redefining Disparity in India,”
Review of Development Economics, 5(1), pp.130-144.
Dumont, Louis. 1970. Homo Hierarchicus: The Caste System and Its Implications
(Nature of Human Society) George Weidenfeld and Nicholson Ltd. and University of
Chicago.
Frick, Joachim R., Jan Goebel, Edna Schechtman, Gert G. Wagner and Shlomo
Yitzhaki. 2004. “Using Analysis of Gini (ANoGi) for Detecting Whether Two Sub-
Samples Represent the Same Universe: The SOEP Experience.” Discussion Paper No.
1049, The Institute for the Study of Labor (IZA), Bonn.
Gupta, Dipankar. 2000. Interrogating Caste: Understanding Hierarchy and
Difference in Indian Society, Penguin Books, New Delhi.
37
Hasan, Rana and Mehta, Aashish, 2006. “Under-representation of Disadvantaged
Classes in India,” Economic and Political Weekly, 41(35), pp. 3791-3796.
Jayadev, Arjun, Motiram, Sripad, Vakulabharanam, Vamsi. 2007. “Patterns of
Wealth Disparities in India During Liberalization,” Mimeo. Dalhousie University,
Canada.
Kojima, Yoko. 2006. “Caste and Tribe Inequality: Evidence from India, 1983-
1999,” Economic Development and Cultural Change, 54(2), pp. 369-404.
Mehrotra, Santosh. 2006. “Well-Being and Caste in Uttar Pradesh,” Economic
and Political Weekly, 41(40), pp. 4261-4271.
Milanovic, Branko and Shlomo Yitzhaki. 2002. “Decomposing World Income
Distribution: Does the World Have a Middle Class?” Review of Income and Wealth,
Series 48, No.2, pp.155-178.
Mohanty, Mritiunjoy, 2006. “Social Inequality, Labour Market Dynamics and
Reservation,” Economic and Political Weekly, 41(46), pp. 3777-3789.
Munshi, Kaivan, and Rosenzweig, Mark. 2006. "Traditional Institutions Meet the
Modern World: Caste, Gender, and Schooling Choice in a Globalizing Economy,"
American Economic Review, American Economic Association, 96(4), pages 1225-1252.
National Sample Survey 2005. “Household Assets and Liabilities in India (as on
30.06.2002),” Report, November 2005.
Srinivas, M.N. 2000. Social Change in Modern India, Orient Longman, India.
Srinivasan, K. and Mohanty, S.K. “Deprivation of Basic Amenities by Caste and
Religion,” Economic and Political Weekly, 39(07), pp. 728-735.
38
Subramanian, S., and Jayaraj D. 2006. “The Distribution of Household Wealth in
India,” paper prepared for UNU-WIDER project meeting, 4-6 May, WIDER: Helsinki.
Sundaram, K. 2006. “On Backwardness and Fair Access to Higher Education in
India: Results from NSS 55th Round Surveys, 1999-2000,” Economic and Political
Weekly, 41(50), pp. 5173-5182.
Yitzhaki, Shlomo. 1994. “Economic Distance and Overlapping Distributions,”
Journal of Econometrics, 61, pp. 147-159.
39
Table 1 Joint Distribution of Wealth and Consumption, 2002
Consumption Wealth 1991 q1 q2 q3 q4 q5
q1 7.2 4.8 3.4 2.4 2.2q2 5.5 5.2 4.2 3.0 1.9q3 3.8 4.7 4.7 4.0 2.8q4 2.6 3.6 4.7 5.1 4.0q5 0.9 1.7 3.0 5.4 9.1
Total 20.0 20.0 20.0 20.0 20.0 2002
q1 7.3 4.8 3.7 2.4 1.4q2 5.8 5.1 4.2 3.0 1.5q3 3.9 4.6 4.7 4.2 2.5q4 2.3 3.7 4.6 5.3 4.2q5 0.7 1.8 2.8 5.1 10.4
Total 20.0 20.0 20.0 20.0 20.0 Note: Consumption is measured as per capita consumption expenditures (MPCE),
i.e., total household consumption expenditures divided by the number of persons in the household.
40
Table 2 Wealth by Caste Groups (in thousands of 2006 Rupees)
A. Scheme I 1991 2002 Percent change
Mean MedianShare in
Population Mean MedianShare in
Population Mean MedianUrban ST 38.5 19.5 0.7 67.1 20.8 0.7 74% 6%Urban SC 31.1 16.1 3.1 46.0 22.6 3.8 48% 40%Urban OC 88.4 33.1 20.6 123.4 47.1 20.9 40% 42%Rural ST 24.6 15.8 8.0 34.1 19.0 7.3 39% 21%Rural SC 23.3 13.0 15.3 30.9 18.0 15.9 33% 39%Rural OC 58.8 31.1 52.2 77.6 38.9 51.4 32% 25%All 55.7 24.9 100.0 75.3 32.0 100.0 35% 28% B. Scheme II (2002 only)
Mean MedianShare in
Population Urban ST 67.1 20.8 0.7 Urban SC 46.0 22.6 3.8 Urban OC 123.4 47.1 20.9
OBC 85.5 34.8 9.0 Hindu Others (FC) 169.3 77.7 8.4 Non-Hindu Others 109.5 34.5 3.4
Rural ST 34.1 19.0 7.3 Rural SC 30.9 18.0 15.9 Rural OC 77.6 38.9 51.4
OBC 62.0 34.2 31.3 Hindu Others (FC) 105.4 60.1 14.1 Non-Hindu Others 93.9 25.9 6.0
41
Table 3 Percentile Cutoffs for Scheme II, 2002 (in thousands of 2006 Rs)
Percentile Urban ST Urban SC Urban OBC Urban FC
Urban NH Rural ST Rural SC
Rural OBC Rural FC Rural NH All
5 0.8 0.7 1.1 2.5 1.0 2.7 2.1 3.4 6.1 2.4 2.410 1.8 2.0 3.1 6.0 2.1 4.8 4.0 6.6 12.0 4.5 5.115 3.3 3.7 5.8 11.0 4.7 6.3 5.5 9.5 18.1 5.9 7.620 5.2 5.6 8.8 17.3 8.3 7.7 7.0 12.3 22.8 7.3 10.225 7.0 7.9 12.3 24.5 11.2 9.3 8.6 15.2 27.7 9.5 12.930 9.8 10.6 16.1 33.1 14.5 11.0 10.1 18.3 33.0 11.8 16.035 12.0 13.4 20.1 42.2 18.0 12.6 11.9 21.6 39.2 14.8 19.340 14.7 16.3 24.2 52.8 22.8 14.7 13.6 25.6 45.5 17.9 23.045 18.2 19.4 29.0 64.6 27.9 16.6 15.8 29.7 52.5 21.4 27.350 20.8 22.6 34.8 77.7 34.5 19.0 18.0 34.2 60.1 25.9 32.055 23.6 27.2 41.4 93.1 42.0 21.6 20.3 39.1 68.9 31.0 37.660 28.9 31.0 49.0 111.7 52.1 24.7 23.3 45.0 78.9 38.1 44.465 38.0 35.1 58.4 131.7 65.3 28.5 26.6 51.3 89.9 46.8 52.370 49.0 41.1 70.5 155.5 84.5 32.0 31.0 60.1 104.2 59.0 62.875 61.4 50.4 83.9 191.8 107.7 37.3 36.2 70.8 122.4 76.7 76.380 77.3 63.6 106.9 235.0 139.1 43.8 43.2 83.9 145.3 104.1 94.585 102.9 81.8 143.2 290.9 189.9 52.8 52.2 102.8 181.2 145.1 122.290 144.9 109.8 192.3 379.3 273.9 71.6 66.5 136.1 230.5 220.8 170.195 220.6 168.8 313.1 562.9 429.0 110.4 98.3 205.9 331.0 402.3 272.3
42
Table 4 Caste and Occupational Status (percentage distribution across status) 1991 2002 OC Subgroups
ST SC OC ST SC OC
OBC Hindu FC Non-Hindu
Others Urban Elite 0.7 1.9 5.8 1.2 1.9 6.0 3.6 10.0 7.1Urban Middle 1.5 3.1 7.9 1.9 3.5 8.0 5.4 11.9 9.8Urban Manual 6.0 11.9 14.7 5.8 14.2 14.9 13.4 15.7 19.4Rural Elite 13.5 8.7 17.5 10.3 8.6 15.7 14.3 19.9 11.9Rural Middle 29.4 13.8 21.9 25.9 10.3 18.9 19.7 19.8 13.1Rural Working 48.8 60.7 32.3 55.0 61.6 36.6 43.7 22.9 38.8All 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0
43
Table 5 Caste, Occupational Status and Wealth A. Mean Values in Thousands of 2006 Rs.
1991 2002 OC Subgroups
ST SC OC ST SC OC
OBC Hindu FC Non-Hindu
Others Urban Elite 82.1 64.7 180.6 144.0 104.3 231.6 178.9 270.9 211.7Urban Middle 64.7 39.3 87.4 69.6 57.7 116.3 81.9 144.4 115.7Urban Manual 30.8 23.4 52.7 50.4 35.5 83.4 62.1 123.4 69.0Rural Elite 49.4 52.2 112.6 72.9 62.3 151.5 122.4 167.9 235.1Rural Middle 31.1 43.8 67.1 50.1 65.2 99.4 85.5 111.3 145.4Rural Working 13.7 14.4 24.2 19.3 20.8 34.6 31.7 45.9 33.2
B. Status Disparity in Wealth by Caste (Ratio to Rural Working) 1991 2002 OC Subgroups
ST SC OC ST SC OC
OBC Hindu FC Non-Hindu
Others Urban Elite 5.99 4.48 7.47 7.45 5.02 6.68 5.65 5.91 6.38Urban Middle 4.72 2.72 3.62 3.60 2.77 3.36 2.59 3.15 3.48Urban Manual 2.25 1.62 2.18 2.61 1.71 2.41 1.96 2.69 2.08Rural Elite 3.60 3.62 4.66 3.77 3.00 4.37 3.86 3.66 7.08Rural Middle 2.27 3.03 2.77 2.59 3.14 2.87 2.70 2.43 4.38Rural Working 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
C. Caste Disparity in Wealth by Status (Ratio to OC)
1991 2002 OC Subgroups
ST SC OC ST SC OC
OBC Hindu FC Non-Hindu
Others Urban Elite 0.45 0.36 1.00 0.62 0.45 1.00 0.77 1.17 0.91Urban Middle 0.74 0.45 1.00 0.60 0.50 1.00 0.70 1.24 0.99Urban Manual 0.59 0.44 1.00 0.60 0.43 1.00 0.74 1.48 0.83Rural Elite 0.44 0.46 1.00 0.48 0.41 1.00 0.81 1.11 1.55Rural Middle 0.46 0.65 1.00 0.50 0.66 1.00 0.86 1.12 1.46Rural Working 0.57 0.60 1.00 0.56 0.60 1.00 0.91 1.32 0.96
44
Table 6 Occupational Status and Wealth A. Mean Values in Thousands of 2006 Rs.
Mean Median Percent Change Proportion 1991 2002 1991 2002 Mean Median 1991 2002 Urban Elite 170.8 220.2 64.3 104.5 28.9 62.4 4.6 4.8Urban Middle 82.4 109.1 38.3 53.1 32.5 38.5 6.4 6.6Urban Manual 47.0 72.8 19.9 27.1 54.7 36.2 13.4 14.0Rural Elite 101.5 136.0 56.6 71.7 33.9 26.7 15.5 13.9Rural Middle 59.8 89.7 41.1 60.2 50.0 46.5 21.0 17.7Rural Working 20.3 29.2 12.4 17.6 44.0 41.6 39.0 43.0All 55.7 75.3 24.9 32.0 35.1 28.4 100.0 100.0
B. Status Disparities in Wealth (Ratio to Rural Working)
Mean Median 1991 2002 1991 2002Urban Elite 8.43 7.55 5.19 5.95Urban Middle 4.07 3.74 3.09 3.02Urban Manual 2.32 2.49 1.60 1.54Rural Elite 5.01 4.66 4.56 4.08Rural Middle 2.95 3.08 3.32 3.43Rural Working 1.00 1.00 1.00 1.00
45
Table 7 Within and Between Group Inequality by Caste and Status
A. Caste Gini points Percentage shares 2002 2002 1991 Scheme I Scheme II 1991 Scheme I Scheme II Overall Gini 0.648 0.655 0.655 100.0 100.0 100.0 Within group 0.595 0.599 0.572 91.9 91.4 87.4 Between group 0.053 0.056 0.083 8.1 8.6 12.6 B. Status Gini points Percentage shares 1991 2002 1991 2002 Overall Gini 0.648 0.655 100.0 100.0 Within group 0.480 0.490 74.1 74.9 Between group 0.168 0.164 25.9 25.1
46
Table 8 Within Group Inequality and Overlapping by Caste
1991 2002
Population
share Wealth share Gini Overlap
Population share
Wealth share Gini Overlap
Urban ST 0.007 0.005 0.628 1.049 0.007 0.006 0.725 1.137Urban SC 0.031 0.017 0.627 1.056 0.038 0.023 0.632 1.051Urban OC 0.206 0.327 0.700 0.993 0.209 0.342 0.683 0.966
Urban OBC 0.090 0.102 0.677 1.016Urban FC 0.085 0.190 0.648 0.840Urban NH 0.034 0.050 0.713 1.054
Rural ST 0.080 0.035 0.526 0.913 0.073 0.033 0.568 0.969Rural SC 0.153 0.064 0.573 0.973 0.159 0.065 0.557 0.947Rural OC 0.522 0.551 0.595 0.918 0.514 0.530 0.609 0.929
Rural OBC 0.313 0.258 0.580 0.932Rural FC 0.141 0.197 0.563 0.791Rural NH 0.060 0.075 0.734 1.095
All 1 1 0.648 1 1 1 0.655 1
47
Table 9 Matrices of Overlapping and Ranks for Caste Groups, 2002
A. Overlapping
Urban ST
Urban SC
Urban OBC
Urban FC
Urban NH
Rural ST
Rural SC
Rural OBC
Rural FC
Rural NH
Urban ST 1 1.045 1.107 1.131 1.052 1.065 1.042 1.193 1.266 1.051Urban SC 0.938 1 1.009 0.928 0.933 1.051 1.032 1.119 1.111 0.963Urban OBC 0.881 0.916 1 1.062 0.951 0.905 0.885 1.068 1.179 0.928Urban FC 0.716 0.722 0.842 1 0.827 0.681 0.662 0.866 1.037 0.776Urban NH 0.915 0.944 1.041 1.133 1 0.928 0.906 1.100 1.230 0.970Rural ST 0.855 0.925 0.918 0.809 0.842 1 0.977 1.040 1.000 0.879Rural SC 0.852 0.924 0.889 0.739 0.812 1.021 1 1.017 0.934 0.868Rural OBC 0.794 0.849 0.908 0.903 0.838 0.851 0.831 1 1.070 0.826Rural FC 0.654 0.678 0.792 0.903 0.750 0.625 0.608 0.836 1 0.697Rural NH 0.937 0.973 1.075 1.163 1.029 0.971 0.945 1.148 1.277 1
B. Ranks
Urban ST
Urban SC
Urban OBC
Urban FC
Urban NH
Rural ST
Rural SC
Rural OBC
Rural FC
Rural NH
Urban ST 0.5 0.502 0.419 0.298 0.417 0.522 0.536 0.410 0.301 0.448Urban SC 0.498 0.5 0.410 0.282 0.409 0.526 0.541 0.402 0.283 0.445Urban OBC 0.581 0.590 0.5 0.362 0.492 0.622 0.634 0.502 0.381 0.531Urban FC 0.701 0.718 0.638 0.5 0.619 0.748 0.758 0.653 0.544 0.656Urban NH 0.582 0.590 0.508 0.381 0.5 0.616 0.628 0.509 0.400 0.534Rural ST 0.478 0.473 0.378 0.251 0.383 0.5 0.517 0.361 0.237 0.423Rural SC 0.463 0.459 0.366 0.242 0.371 0.483 0.5 0.348 0.227 0.409Rural OBC 0.590 0.598 0.497 0.347 0.491 0.639 0.652 0.5 0.363 0.538Rural FC 0.699 0.717 0.618 0.456 0.600 0.763 0.773 0.637 0.5 0.650Rural NH 0.551 0.555 0.469 0.344 0.466 0.576 0.591 0.461 0.350 0.5
48
Table 10 Within Group Inequality and Overlapping by Status
1991 2002
Populationshare
Wealth share Gini Overlap
Populationshare
Wealth share Gini Overlap
Urban Elite 0.046 0.142 0.694 0.763 0.048 0.141 0.640 0.728Urban Middle 0.064 0.095 0.640 0.948 0.066 0.096 0.620 0.921Urban Manual 0.134 0.113 0.674 1.089 0.140 0.135 0.701 1.083Rural Elite 0.155 0.283 0.578 0.764 0.139 0.250 0.599 0.798Rural Middle 0.210 0.226 0.453 0.679 0.177 0.211 0.464 0.642Rural Working 0.390 0.142 0.542 0.917 0.430 0.166 0.547 0.930All 1 1 0.648 1 1 1 0.655 1
49
Table 11 Matrices of Overlapping and Ranks for Status Groups, 2002
A. Overlapping
Urban Elite
Urban Middle
Urban Manual
Rural Elite
Rural Middle
Rural Working
Urban Elite 1 0.880 0.688 1.021 0.971 0.527Urban Middle 1.026 1 0.856 1.126 1.210 0.749Urban Manual 1.029 1.075 1 1.152 1.308 0.985Rural Elite 0.925 0.855 0.681 1 1.027 0.522Rural Middle 0.758 0.730 0.604 0.865 1 0.440Rural Working 0.654 0.796 0.843 0.773 1.009 1
B. Ranks
Urban Elite
Urban Middle
Urban Manual
Rural Elite
Rural Middle
Rural Working
Urban Elite 0.5 0.605 0.714 0.533 0.599 0.803Urban Middle 0.395 0.5 0.623 0.415 0.466 0.719Urban Manual 0.286 0.377 0.5 0.289 0.315 0.588Rural Elite 0.467 0.584 0.711 0.5 0.573 0.817Rural Middle 0.401 0.534 0.685 0.427 0.5 0.829Rural Working 0.197 0.281 0.412 0.183 0.171 0.5
50
Figure 1 Disparity in Wealth by Caste, 1991 and 2002 (Ratio to Mean or Median Values of Rural OC)
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
1.60
1.80
Mean Median Mean Median
1991 2002
Rat
io
Urban STUrban SCUrban OCRural STRural SC
51
Figure 2 Disparity in Wealth by among OC groups, 2002 (Ratio to Mean or Median Values of Rural OC)
0.00
0.50
1.00
1.50
2.00
2.50
OBC HinduOthers
Non-HinduOthers
OBC HinduOthers
Non-HinduOthers
Rat
io Mean
Median
Urban Rural
52
Figure 3 Deviation from Overall Percentile Cutoffs by Caste at Selected Percentiles, 2002 (in thousands of 2006 Rs.)
-200-175-150-125-100
-75-50-25
0255075
100125150175200225250275300325
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
Percentiles
Thou
sand
s of
Rs
Urban STUrban SCUrban OBCUrban FCUrban NHRural STRural SCRural OBCRural FCRural NH
53
Figure 4 Urban-Rural Wealth Gap (as a Percent of Rural Wealth) by Caste at Selected Percentiles
-80%
-60%
-40%
-20%
0%
20%
40%
60%
80%
100%
120%
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
Percentile
Urb
an-R
ural
Gap ST
SCOBCFCNH
54
Figure 5 Percent Change in Wealth at Selected Percentiles by Caste Group, 1991 to 2002
-40%
-20%
0%
20%
40%
60%
80%
100%
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95
Percentiles
Perc
ent
Urban ST
Urban SC
Urban OC
Rural ST
Rural SC
Rural OC