World Development Vol. 36, No. 1, pp.

128143, 2008
! 2007 Published by Elsevier Ltd.
0305-750X/$ - see front matter
www.elsevier.com/locate/worlddev

doi:10.1016/j.worlddev.2007.02.017

Gender Differentials in Literacy in India:

The Intriguing Relationship with Womens
Labor Force Participation
APARNA SUNDARAM and REEVE VANNEMAN
University of Maryland, College Park, USA

Summary. Contrary to expectations from either human capital or gender empowerment perspectives, analyses across 409 Indian districts show that girls have relatively lower literacy compared to
boys in areas where more women are in the labor force. The most likely explanation is that areas
with higher womens labor force participation are also areas with higher girls labor force participation; these higher rates of girls labor depress their literacy and education. Gender inequalities in
literacy are therefore an exception to the usual egalitarian impacts of womens labor force participation and remind us again of the multidimensionality of gender inequalities.
! 2007 Published by Elsevier Ltd.
Key words Asia, India, gender, womens labor force participation, literacy, spatial autocorrelation

1. INTRODUCTION
Womens participation in the labor force has
long been central to research on gender inequalities (Boserup, 1970). Much of this research has
sought to understand how labor force participation contributes to womens status and to the
reduction of gender inequalities. General models
of gender inequality (e.g., Chafetz, 1984) tend to
emphasize the importance of womens economic
roles in determining their position in other
spheres, from household bargaining to representation in state governance. Empirical research
has found that womens labor force participation is associated with less bias against girls in
child mortality (Kishor, 1993; Rosenzweig &
Schultz, 1982), better health for girls (Thomas,
1994), and with more say in some household
decision making (Dharmalingam & Morgan,
1996; Jejeebhoy & Sathar, 2001; Malhotra &
Mather, 1997). Of course, many conditions
may limit the liberating impacts of work outside
the household (e.g., who controls the income
from such work), and, even in the best of circumstances, outside work usually implies a dual burden for wives and mothers.
Nevertheless, it is less often asserted that
womens labor force participation actually
128

restrains womens progress toward gender

equality. The results we demonstrate below
illustrate one such circumstance when higher
rates of womens labor force participation contribute to less rather than more gender equality:
more womens labor force participation may
lead to girls being withdrawn from school and
put to work; the frequency of girls work may
restrict their schooling, which widens the gender gap in basic education. None of these linkages (from womens work to girls work, or
from girls work to less schooling, or even from
less girls schooling to wider gender gaps in
educational attainment) is inevitable. Under
* An earlier version of this paper was presented at the
annual meeting of the Population Association of America in Washington, DC, 2001. We gratefully acknowledge helpful comments from World Development
reviewers and from Sonalde Desai, Patricio Korzeniewicz, Harriet Presser, and participants at a seminar
at the University of Maryland. This research was supported in part by grants from the William and Flora
Hewlett Foundation to the Center on Population,
Gender, and Social Inequality. Please direct correspondence to Aparna Sundaram. Final revision accepted:
February 26, 2007.

GENDER DIFFERENTIALS IN LITERACY IN INDIA

129

Table 1. Literacy rates in percents for all ages combined, by sex: India, 19612001
Year Total percent Women percent Men percent Percentage point difference Difference in log odds ratios
1961
1971
1981
1991
2001*

22.2
27.8
34.6
42.1
65.4

11.6
17.2
23.1
31.1
54.2

32.1
37.7
45.4
52.3
75.8

Source: Vanneman and Barnes (2003).

* Census of India, 2001.

many conditions, we would expect womens

labor force participation to encourage more,
not less, education for girls. Nevertheless, the
general pattern we review across India provides
a cautionary message and reminds us once
again of the multidimensionality of gender
stratification.
Understanding the multidimensionality of
gender stratification (Mason, 1986) also helps
us think about other gendered consequences
of womens labor force participation. Similar
analyses to the one we describe below have
demonstrated that womens labor force participation rates lower gender gaps in child mortality (Kishor, 1993). However, while the
frequency of womens work may increase girls
economic value, which has positive benefits for
their survival (Rosenzweig & Schultz, 1982),
girls economic value may also increase their
parents incentives to keep them out of school
to maximize their immediate economic returns.
Each dimension of gender inequality requires
separate theoretical models; a global construct
of gender equality may only interfere with our
understanding of how gendered outcomes are
generated.
(a) Literacy
This analysis focuses on the gender gap in literacy in India. The importance of literacy need
not be stressedit is an end in itself. Moreover,
literacy and schooling are important determinants of economic growth (Barro, 2001) and
womens education may be especially important for future growth (Klasen, 2002; World
Bank, 2001). Within societies, education levels
are a principal determinant of adult outcomes:
if women are disadvantaged in schooling, that
disadvantage will ripple through the entire
system of gender stratification (Hill & King,
1993).
While overall literacy levels in India are low
(65% of the total adult population in India

20.5
20.5
22.3
21.2
21.7

!1.3
!1.1
!1.0
!0.9
!1.0

was literate in 2001), women fare worse than

men: in 2001, only 54% of the women were literate compared to 76% of the men. The gap between men and womens literacy has been
remarkably stable over the last forty years.
Table 1 presents Census data since 1961. The
two indices of the gender gap (the percentage
point difference and the logged gender odds
ratios) tell somewhat different stories. For
instance, the percentage point difference
indicates that the gender gap in literacy worsened in the 196171 period. However, the gender odds ratio indicates that the relative odds of
literacy improved for women in this time. The
absolute difference between male and female literacy percentages can be a misleading indicator
since it is necessarily small when literacy rates
are low (or high). The gender difference in odds
ratios avoids this problem. The use of the odds
ratio is also consistent with the general practice
of logistic regressions. For these reasons, gender differences are measured throughout this
analysis using the difference between girls log
odds of literacy and boys log odds of literacy
(see Tables 2 and 3).
Most states in India, including Kerala, have
more literate males than literate females. But
the gap between women and men (or between
girls and boys) varies widely across the country
so we can ask what is different about the areas
with smaller gaps from the areas with enormous gender gaps?
(b) The consequences of womens labor force
participation
Hypothesis 1. Areas with more equal labor
force participation by women will also be areas
with more equal literacy rates for girls.
At least three overlapping paths may link
womens labor force participation to reduced
gender inequalities in education. Education is,
at least in part, a family investment in their

130

WORLD DEVELOPMENT

Table 2. Gender difference in the log odds of literacy in

India: selected states, 1991
State

Gender gap in literacy

Kerala
Punjab
Assam
West Bengal
Tamil Nadu
Karnataka
Maharashtra
Andhra Pradesh
Orissa
Madhya Pradesh
Gujarat
Bihar
Himachal Pradesh
Haryana
Uttar Pradesh
Rajasthan
Source: Vanneman and Barnes (2003).

!0.13
!0.44
!0.45
!0.49
!0.74
!0.75
!0.77
!0.81
!0.85
!0.97
!0.98
!1.03
!1.05
!1.07
!1.08
!1.61

childrens human capital: where both women

and men are employed, there is an economic
incentive to educate both girls and boys. Even
if the income returns to human capital are
weaker for working women than for working
men, and even if there are other non-employment incentives for educating girls (e.g., their
increased value in a marriage market), the additional incentive of returns to schooling in the
labor market should lead to a positive association between womens labor force participation
and womens schooling, other factors being
equal.

Second, general theories of gender stratification emphasize the strengthened bargaining position of women who have independent access
to economic resources, enabling them to better
resist some aspects of patriarchal domination.
Labor force participation usually implies expanded network ties and independent exposure
to the broader society that also raises womens
bargaining position both within and outside the
household. For instance, a study by Dharmalingam and Morgan (1996) on two villages in
South India showed that in the village where
womens labor force participation was high,
women were more likely to have greater mobility, have greater spousal communication, and
better control over their households resources
compared with the village where womens labor
force participation was low.
These conflict models apply best for adult
inequalities. Nevertheless, if we assume that
mothers are more apt than fathers to bargain
for their daughters schooling, then mothers
increased bargaining power through economic
empowerment might lead to more schooling
for their daughters (Thomas, 1994).
A broader, contextual argument traces gender inequalities in schooling to general cultural
evaluations of womens worth, which may be
improved by womens labor force participation. This more indirect influence depends
less on the interpersonal dynamics within
households and more on the general expectations of equal treatment in the society. To
the extent that these general expectations of
equality are influenced by economic roles such
as labor force participation, we would expect

Table 3. Means and standard deviations for the variables used in the analysis: districts in India, 1991
Variable
Gender difference in log literacy odds
Gender ratio adult main: non-main
Odds of women being in exogamous unions
Log ratio boys literacy/illiteracy 1014
Education effort
Housing index
Percent urban in a district
Percent landless in a district
Percent SC in a district
Percent ST in a district
Percent Muslims in a district
Southern states
Log ratio 014 boys, main/non-main
Log ratio 014 girls, main/non-main
Source: Vanneman and Barnes (2003).

Mean

Std. Dev.

Minimum

Maximum

409
409
409
409
409
409
409
409
409
409
409
409
409
409

!0.94
!2.26
1.66
1.47
27.22
35.33
21.73
36.34
16.13
10.47
11.08
0.20
!3.10
!3.77

0.44
0.95
0.92
0.99
11.84
14.38
16.66
18.11
7.56
17.88
11.50
0.40
0.70
1.06

!2.38
!4.98
!0.27
!0.44
8.52
4.77
0.00
1.03
0.00
0.00
0.07
0.00
!6.10
!6.47

0.16
!0.30
4.48
4.82
84.09
88.88
100.00
81.13
51.76
94.75
70.45
1.00
!1.56
!1.76

GENDER DIFFERENTIALS IN LITERACY IN INDIA

more equal access to schooling for girls and

boys.
On the other hand, womens labor force participation is undoubtedly itself influenced by
these broader cultural patterns. Thus, where
womens independent mobility outside the
home is culturally restricted, labor force participation is more unlikely. These mobility restrictions may also affect girls, especially adolescent
girls, access to schools. Thus, a positive relationship between womens labor force participation and girls education may result from
their common determination by cultural restrictions on womens free mobility.
(c) Child labor and schooling
Hypothesis 2a. Improved odds of female labor
force participation will increase rates of girl
child labor.
Hypothesis 2b. Areas with high rates of girls
labor force participation will have lower rates
of girls literacy.
While there may be several reasons to expect
womens labor force participation to increase
girls education, it could interfere with girls access to schools if womens labor force participation implies girls labor force participation or
more housework for girls, and if girls laborforce participation or increased housework reduces their schooling opportunities. There is
substantial debate about the relationship between child labor and access to schooling. After
an initial public outcry against child labor in
developing countries (for a discussion on this
see Anker, 2000; Basu, 1999), more careful
observers noted that child labor is not necessarily incompatible with schooling (Patrinos &
Psacharopoulos, 1997; Probe Team, 1999).
Even in the United States, adolescents often
are employed and there is little evidence that
this youthful employment has negative consequences for their adult outcomes (Carr, Wright,
& Brody, 1996).
On the other hand, some types of labor force
participation undoubtedly constrain childrens
access to school. Where girls labor can contribute to the households economic standing, parents, especially poor parents, will be tempted to
keep their children out of school in order to
maximize their immediate economic returns
(Basu & Van, 1998; Basu, 1999; Cigno &
Rosati, 2000). In this paper, we hypothesize

131

that womens participation in the labor force

does pull girls into the official labor force or
into more unrecorded domestic and productive
work and thus lowers their rate of schooling.
Previous research indicates that when women
work and earn a wage, girls labor force participation increases at the cost of their schooling
(Basu, 1992; Emerson & Souza, 2003; Levison,
Moe, & Knaul, 2000; Ray, 2000). Conversely,
Ravallion and Wodon (2000) show that subsidy-induced increases in Bangladeshi childrens schooling did result in some decline in
their likelihood of working. Thus, womens labor force participation could lead to greater
not smaller gender gaps in access to schooling
through its effects on girls labor force participation.
When children are not employed directly in
the paid labor force, they are generally expected
to help with (often gender segregated) household chores. Where mothers work outside the
home, it is likely that daughters are called on
to provide more household help, and this extra
household burden may also interfere with girls
education. Therefore, a more comprehensive
approach to the effect of womens labor force
participation would consider both daughters
household labor and their employment (Basu,
1992; Levison et al., 2000). Our data, however,
do not allow us to measure childrens domestic
work, but we would expect similar and probably more dramatic results if these measures
were available.
In addition, much of childrens non-domestic
work goes unrecorded in official censuses and
surveys. For instance, girls collection of fuel
wood and crop residues or boys supervision
of household animals is rarely counted in most
official statistics but contributes to the household economy and can require a substantial
time commitment that makes attending school
and completing schoolwork more difficult.
Even childrens work in their familys fields or
at the family store may be intermittent and go
unrecorded in government statistics but may
be important during peak demands. Family
expectations that children contribute to this
work may take precedence over attending
school.
While it would improve analyses if we had
better measures of childrens work, this problem may not be so great in area-level analyses
if the extent of this unrecorded and domestic
work co-varies with the officially recorded
employment. This seems likely to be the case
especially for girls work, where even moderate

132

WORLD DEVELOPMENT

rates of girls employment suggest an absence

of the cultural proscriptions against childrens
work that would affect unrecorded and perhaps
domestic work as well.
(d) Cultural determinants of gender gaps in
literacy
Hypothesis 3. Areas with patriarchal family
systems will have larger gender gaps in literacy.
Culture matters as well as economics, and because the two often covary, their roles need to
be evaluated independently. While the primary
focus of the analysis is on the relationship of
girls and womens labor force participation
with girls schooling, gender differentials in India also reflect family and marriage practices
that vary widely across the country. A general
cultural attitude endorsing patriarchy, based
for instance in religious traditions or in political
ideology, may influence gender gaps in literacy
as it would other gendered aspects of social life.
If cultural practices and beliefs consistently
emphasize womens subordinate position, parents will follow custom by educating boys first.
Often, specific cultural practices such as the
seclusion of women and prohibitions on their
free movement will directly limit girls ability
to attend schools (and to work outside the
home).
Indian gender stratification, in particular, is
interpreted with such cultural schema (Dyson
& Moore, 1983; Miller, 1981; Sopher, 1980).
The rough geographic division between North
and West versus South and East has long been
recognized as a cultural line dividing more
patriarchal from more egalitarian gender relations. This geographic division may reflect the
differential penetration of ancient patriarchal
cultures into the subcontinent. The North
South division also describes a variety of family
patterns that may underlie womens relative position. In India, patrilocal village exogamy may
be especially relevant for a broad range of gendered outcomes in India. When parents expect
to see their adult daughters only rarely and rely
on eventual support from their adult sons, village exogamy and patrilocal residence combine
to reinforce patriarchy (Kishor, 1993). It is not
surprising that in these circumstances parents
investments in their sons take priority over
investments in their daughters. Literacy may
be seen as dispensable for daughters but essential for sons. Therefore, areas with high levels

of patrilocal exogamy should be characterized

by larger gender gaps in literacy.
Besides exogamy, dowry practices, decision
making within the family, and seclusion of women are other aspects of a patriarchal culture
that could have broad implications for gender
relations, including educational outcomes.
These patterns of gender relations within the
family are difficult to measure with existing
data and so are not specified as separate
hypotheses here. We assume that the elements
of this patriarchal family pattern tend to cluster
together so that our measure of family exogamy probably proxies for the entire set of family differences. Thus, the relationships with
marriage exogamy should not be interpreted
too narrowly as the outcome of one custom
alone.
(e) Development effects on gender gaps in
literacy
Hypothesis 4. More developed areas will have
more gender equal literacy rates.
Neo-classical and modernization theorists often posit that inequalities between men and women will erode with development (Knodel &
Jones, 1996; Wils & Goujon, 1998). Beutel
and Axinn (2002), for instance, hypothesize
that family roles become less important with
the spread of modern institutions (schools,
wage labor, transportation infrastructure) so
that gender differentiation declines with the decline of the family. These modernization theories have been controversial and several
studies have shown that gender inequalities
are resilient to economic development (Beneria
& Sen, 1981; Forsythe, Korzeniewicz, & Durrant, 2000; Kabeer, 1994). If we make a distinction between womens (absolute) status and
their (relative) position in the gender stratification system, there may be little expectation of
reduced gender gaps with increased development, at least at the early stages of economic
growth (Dollar & Gatti, 1999). Gender inequalities may be embedded in the social and cultural
structure of the society that is largely unaffected
by economic wealth. In fact, because increasing
wealth permits more families to enact patriarchal ideals in their daily lives, increasing wealth
can sometimes aggravate gender gaps in the
society. In India, for instance, the wealthiest
areas have experienced the biggest increases in
sex selective abortions because those are the

GENDER DIFFERENTIALS IN LITERACY IN INDIA

places where sonograms are affordable and

most widespread (Arnold, Kishor, & Roy,
2002).
Even when development may not imply any
direct changes in the gender system itself, the
additional resources available in wealthier societies may still affect the gender gaps in schooling. When families and educational systems
follow a gender queuing model of access to
schools, more resources may lead to a declining
gender gap once the male queue diminishes.
The queuing effect suggests that even in areas
where son preference is strong and patriarchal
norms are entrenched, there might exist a queue
for resources with boys at the head of the queue
and girls at the back. As resources expand with
economic development, girls move to the head
of the queue after the boys queue is exhausted.
Since families reach this stage at different times
and since other factors enter into the schooling
decision, a queuing model does not imply that
all boys are educated before any girls. However, the queuing model does suggest that the
extent of boys schooling will influence how
many girls gain access. While early in the process, the gender gap may widen because boys
are the head of the queue, as queues shorten,
more education will imply more girls education.
A model in which girls education is determined largely by demands of the marriage
market predicts a similar result: male collegeeducated engineers should not marry illiterate
wives. So, as boys education increases, families
will be under pressure to educate daughters
(although not as much as sons) in order to find
suitable matches. In addition, mothers hometeaching roles supplement school based education for their sons and raise the demand for
female education (Behrman, Foster, Rosenzweig, & Vashishtha, 1999). Again, girls educational levels would be driven in part by boys
levels in these models.
The lack of effort at improving educational
access may also have an impact on the gender
gap in literacy (Banerji, 1997; Dreze & Saran,
1995; Visaria, Gumber, & Visaria, 1993). When
there is a scarcity of school resources, it is likely
that boys receive access to these scarce resources ahead of girls. However, as educational
resources expand, girls begin to benefit. This
suggests that in areas where the extent of education is more widespread, the gender gap in literacy will be narrower.
Finally, development may imply not only a
higher standard of living but also a closer inte-

133

gration into a world cultural system (Meyer,

Ramirez, & Soysal, 1992). To the extent that
the world cultural system assumes gender equal
access to schools, then educational systems in
developing nations will be constrained to incorporate both boys and girls as they become more
integrated into the global culture.
2. METHODS
Our hypotheses concern systems of gender
equality reflecting macrolevel influences that
may or may not be mediated through household level differences. For example, in societies
where most women are employed, parents may
recognize the advantages of educating daughters, even if the particular mothers of those
daughters are not themselves employed. The
household characteristics may intensify the
relationships, but may not exhaust them.
The macrolevel relationships are probably
some mix of contextual and household compositional effects. It is not our purpose at this
stage to evaluate the relative contributions of
those two types of effects but rather to identify
the overall macrolevel relationships themselves.
We therefore use macrolevel data, not as a substitute for household level data, but as the
appropriate level of analysis for our system level hypotheses.
We use district level data from the 1991
Indian Census 1 (Vanneman & Barnes, 2003).
Indian states are subdivided into administrative
units called districts. In 1991, a district had an
average population of about 2,000,000 persons
and was about 5,000 km2 in size. Nine small
states and union territories are treated as single
districts. 2 The census was not taken in Jammu
and Kashmir in 1991. This leaves 409 districts
in the analysis.
Spatial autocorrelation is a substantial problem for Indian districts as it is for most geographic analyses. Literacy rates in the
Gangetic plain are likely to be similar but different from those in Kerala or those in the
Himalayas, even after adjusting for the measurable determinants. Adjacent Indian districts are
not independent units, as OLS would assume.
The principal dependent variable analyzed here
has a spatial autocorrelation, Morans I, of 0.68
across adjacent districts. Consequently, the results reported here incorporate a control for
this spatial autocorrelation using maximum
likelihood estimation techniques (Doreian,
1981). We adopt a spatial disturbance model

134

WORLD DEVELOPMENT

using methods specified by Ord (1975) that calculate a log likelihood function incorporating
alternative estimates of the spatial autocorrelation of the residuals across adjacent districts.
The autocorrelation estimate that minimizes
this function is selected and then used to calculate the coefficients and their standard errors.
(a) Dependent variables
The Indian census defines a literate person as
one who can both read and write with understanding in any language. . . In operation, the
census mostly accepts the households own
claims to literacy without testing for understanding. Partly because of this minimal
requirement, we supplement our analyses with
an examination of gender differentials in
matriculation from secondary school. There
may be additional reasons why the gender
inequalities at matriculation may differ from
the literacy differentials. Cultural restrictions
on adult womens mobility might restrict adolescent girls school attendance but have less
impact on younger girls primary education.
We have measured the gender gap in literacy
by comparing the logged odds of girls 1014
being literate with the logged odds of boys
1014 being literate:
"
#
F literate =F illiterate j
Lj Ln
M literate =M illiterate j
where
Lj = The sex difference in literacy.
Fliterates = Number of literate girls in age
groups 1014.
Filliterates = Number of illiterate girls in age
groups 1014.
Mliterates = Number of literate boys in age
groups 1014.
Milliterate = Number of illiterate boys in age
groups 1014.
As noted in the discussion of national literacy
rates (Table 1), the difference in logged odds
has some advantages over the difference in percentages since it is not mathematically constrained to be small when literacy rates are
very low or high. These constraints could be
important for Indian data since literacy rates
vary so widely across the country. For example,
if boys are 95% literate and girls 90% literate,
there is only a five percentage point difference,
but illiteracy is still twice as common for girls.
The difference in logged odds captures these
gender inequalities better at the high and low

literacy levels than does the percentage difference. For example, the 9095% difference is
equivalent in log odds to a 5068% difference
at lower literacy levels.
The 1014-age category specifies a group who
would have acquired literacy recently, reflecting
current conditions in the district. Yet, this is
still an age by which most people are likely to
have acquired literacy. Across India, 69% of
the 1014 age group are literate compared to
57% in the 79 age group, 64% in the 1519
age group, and 56% in the 2024 age group.
In 1991, 60% of Indian girls 1014 were literate, compared to 77% of Indian boys. The
girls odds of being literate (1.48:1) were less
than half the boys odds (3.35:1). The odds varied widely across India. In several Kerala districts where childrens literacy was nearly
universal, girls had a slight advantage over
boys. In the Northeast predominantly tribal
state of Meghalaya, girls literacy, 59%, also exceeded boys literacy, 57%. But in most districts, girls literacy lagged behind boys. The
largest gaps were found in Rajasthan; in Jalaur
district, for instance, 63% of boys were literate
compared to only 14% of girls.
(b) Independent variables
(i) Female share of the labor force
Calculations of womens labor force participation rates are sensitive to the definitions of
the labor force. The Indian Census includes
two measures of labor force participation.
Main workers are defined as those who were
economically productive (i.e., not household
work or production for household consumption) for the major part of the year. Marginal
workers are those who worked during the past
year but for less than six months. The marginal
worker category typically includes a higher proportion of women and children than does the
main worker category. We report the results
of the analyses using the main worker definition. We also re-calculated the analyses using
the broader definition that includes both main
and marginal workers; the results are almost
identical.
On average, 21% of the labor force is female,
but this varies from about 2% in western Uttar
Pradesh (UP) districts to over 45% in the hill
districts of Uttaranchal. We use a ratio measure
of womens share of the labor force rather than
the simpler labor force participation rate. In
practice, the two are highly correlated. Because
mens labor force participation rates vary only

GENDER DIFFERENTIALS IN LITERACY IN INDIA

slightly across India, the gender ratio (or female

share) is determined primarily by womens
odds of labor force participation. But a few
areas do have somewhat lower rates of male labor force participation (e.g., Kerala) than others, so a simple measure of womens odds of
labor force participation does not reflect the
relative gender equality in participation rates
in those areas. A gender ratio therefore captures somewhat better the relative importance
of women in the labor force. Our measure is
the ratio of adult (i.e., age 15 and over) womens odds of being main workers relative to
adult mens odds.
Relative odds of female labor force participation =
ln

!
"
adult female workersmain =adult female non-workers
adult male workersmain =adult male non-workers

(ii) Child labor

The Indian Census reports the number of
children 514 who are full-year or part-year
(marginal) workers. We calculate the odds
of girls and of boys being full-year workers,
but have also tested the broader, marginal
worker, measure. Children were not defined as
either main or marginal workers if they were
primarily a student, even if such a person
helped in the family economic activity. Thus,
the Census measures are quite conservative.
Across India, only 4% of children were recorded as being full-year workers and an additional 1% as marginal workers. Boys (mean =
5%) are more likely to be working than girls
(3.5%).
Child labor rates vary across the country.
The highest levels of girl child labor are in Bellary in Karnataka (15%) and in Karimnagar in
Andhra Pradesh (14%), while the lowest levels
of girl child labor are in Lakshadweep (0.1%)
and in Mainpuri in UP (0.1%). The highest
levels of boy child labor are in Jhabua in
Madhya Pradesh (17%), while the lowest levels
of boy child labor are in Kannur in Kerala
(0.2%).
While these census rates are low, we suspect
the variation in these measures across India
may also proxy for a range of work demands
on children not captured in full-time labor
force participation statistics. Part-time employment, unrecorded work, and perhaps household labor may co-vary with the census
measures and these, too, may interfere with
childrens schooling.

135

(iii) Patrilocal exogamy

The Census provides no direct measure of
exogamy. It is estimated by comparing the log
odds for women migrating from their birthplace to the log odds for men
Exogamy
!
"
female migrants=female non-migrants
ln
male migrants=male non-migrants
The counts of migrants do not distinguish by
age or marital status, so both the numerators
and denominators include young children who
are unlikely to have moved. Nevertheless, the
odds of women (and girls) having moved from
their birthplace are on average four times as
great as mens odds of moving. Exogamy is
especially high in Bihar where several districts
report over half of the women living in places
other than their birthplaces (compared to only
12% of men). It is low in the Northeast where
in several districts more men have moved than
women. The exogamy measure is also low in
the major metropolitan areas such as Calcutta,
Mumbai, and Delhi reflecting the numbers of
male migrants in those cities.
(iv) Educational development
Like gender stratification, development is a
multidimensional concept. We have included
two educational indicators and three economic
indicators to capture various aspects of the
developmental level of the district
Boys literacy rate :
We measure boys literacy rate as
#literate boys1014
ln
#illiterate boys1014
This ratio is also the denominator of the left
side dependent variable and so may at first
seem questionable. By including the boys ratio
on both sides of the equation, we are effectively
regressing girls literacy rates, the numerator of
the gender literacy ratio, on womens labor
force participation and the other independent
variables, holding constant boys literacy. The
coefficients and standard errors for womens labor force participation are identical if we include boys literacy only on the right side or if
we also incorporate it as the denominator on
the left side. 3 We expect boys literacy rates
to affect the gender gap by affecting the queue
for school resources, so boys literacy may
have a curvilinear impact on the gender gap

136

WORLD DEVELOPMENT

in literacy. Girls may benefit only after most

boys have become literate. Hence, we also include the variable in its quadratic form.
(v) Educational effort
We have used the ratio of teachers per 1,000
children of school going age (514) as a measure of educational effort. On average, there
are about 27 teachers for every 1,000 potential
students in a district but this varies from 8 to
84 across the country.
(vi) Economic development
Two indicators of economic development are
included in this analysis. First, the level of
urbanization is measured as the number of persons living in an urban area as a proportion of
the total district population
Level of urbanization
#persons in urban areas in a district

total persons in district

On average about 22% of the people in each
district live in urban areas, but this varies from
no urban population in Lahul Spiti and Kinnaur in Himachal Pradesh to 100% urban in
Calcutta, Mumbai, Hyderabad, and Chennai.
In some preliminary models, we included the
proportion of workers employed in manufacturing as a measure of economic development,
but the proportion manufacturing is highly correlated with the level of urbanization and results suggested keeping the urbanization
variable and omitting the proportion of workers in manufacturing.
Second, we constructed an index of housing
quality by averaging the proportions of households with electricity, modern cooking fuel, finished floors, roofs and walls, toilets, and indoor
taps. This index resembles a consumption
goods ownership index that has been shown
to reflect family wealth (Filmer & Pritchett,
2001). It is correlated 0.77 with urbanization.
We also experimented with a direct measure
of poverty incomes as reported by the National
Sample Survey (NSS) (see Murthi, Guio, &
Dreze, 1995). The NSS reports these poverty
rates for 75 regions, each of which combines
several districts (Dubey & Gangopadhyay,
1998). We assigned the NSS region poverty
score to each of these districts, so this is a less
geographically detailed measure than the other
development indices. Across districts, regional
poverty is correlated !0.57 with housing facili-

ties and !0.30 with urbanization. NSS poverty

rates have no effect on literacy rates once the
other development measures are controlled, so
we have not reported these results here.
(vii) Control variables
In addition to the gender and development
measures described above, we have included
four other variables as controls in our model.
Each variable represents a segment of the population (scheduled castes, tribals, Muslims, and
landless) that might be related to gender-gaps
in literacy at the household level. 4 It should
be remembered that district-level data might
not reflect household effects. Scheduled castes
and tribals, being outside the strict Hindu codes
of purity and patriarchy, may be freer to adopt
gender equal norms within their households (for
scheduled castes, see Srinivas, 1989; for tribals,
see Agarwal, 1994; Beteille, 1986). Similarly,
landless households may be too poor to observe
the restrictions on womens mobility typical of
more landed households, so this relative freedom may spill over into more gender equality
for childrens education as well. In contrast, it
is sometimes asserted that Muslims are more
conservative on gender issues than even higher
caste Hindus, although this may depend on
whether other household characteristics are
held constant. There is no great variation across
districts on any of these control variables except
for the proportion tribal. Twenty-two districts
out of the 409 have tribal majorities. Five districts are Muslim majority districts.
3. RESULTS
(a) Independent variables
Table 4 presents results for two models: with
and without the controls for the extent of child
labor. In the first model, without controls for
child labor, girls in areas with a high female
share of the labor force have a lower chance
of becoming literate than in areas where women
are less likely to be working. This is the opposite of hypothesis one and counter to what most
theories of gender stratification would predict.
Higher rates of womens labor force participation should protect girls from discrimination;
instead, they appear to reduce girls chances
for literacy.
The paradox is clarified in the second model
once controls are entered for the chance of boys
and girls working. In areas where girls below 15

GENDER DIFFERENTIALS IN LITERACY IN INDIA

Table 4. Maximum likelihood estimates of selected independent variables on the gender gaps in literacy: districts
in India, 1991
Variable

Model 1

Model 2

Constant

!1.349***

!1.072***
(!6.80)
0.191***
!4.47
!0.146***
(!5.81)
!0.175**
(!3.31)
0.041***
(3.82)
0.007***
(3.65)
0.004!
(1.96)
!0.000
(!0.30)
0.006***
(5.40)
0.000
(0.18)
0.002!
(1.89)
!0.001
(!0.77)
!0.024
(!0.32)
!0.336***
(!7.68)
0.285***
(5.63)
0.769***
(21.31)

Gender ratio adult

main:non-main
Odds of women being
in exogamous unions
Log ratio boys literacy/
illiteracy 1014
(Log ratio boys literacy/
illiteracy 1014)2
Education effort
Housing index
Percent urban in a
district
Percent landless in a
district
Percent SC in district
Percent ST in district
Percent Muslims in
a district
Southern districts
Log ratio 514 girls
main/non-main
Log ratio 514 boys
main/non-main
Rho (Spatial
autocorrelation)

(!9.34)
!0.078**
(!3.12)
!0.141***
(!5.19)
!0.233***
(!4.39)
0.052***
(4.62)
0.008***
(3.98)
0.005*
(2.13)
!0.001
(!0.32)
0.006***
(5.18)
0.002
(0.90)
0.002!
(1.81)
0.001
(0.54)
!0.079
(!1.07)

0.818***
(26.37)

Not applicable.
Note: Numbers in parentheses are t-statistics.
Source: Vanneman and Barnes (2003).
!
p < 0.1.
* p < 0.05.
** p < 0.01.
*** p < 0.001.

enter the labor force, they have lower chances

of becoming literate. Once girls labor force
participation rates are held constant, the coefficient for adult womens labor force participation flips to the expected positive sign and is
statistically significant.
Girls work primarily in areas that already
have high labor force participation rates for
adult women (r = 0.8). This correlation between
girls and adult womens labor force participation is responsible for the negative association
of adult womens work and girls literacy as

137

shown in model 1. That is, adult womens labor

force participation appears to hurt girls chances
of literacy because it opens up the possibility for
girls to work. The close association between
adult womens and girls employment probably
results from joint determination by cultural
restrictions on female mobility and by patterns
of industrial demand for female labor in a gender segregated labor force. However, where
adult women work and young girls do not
(model 2), girls actually have a better chance
of going to school and becoming literate than
in areas where women do not work. This is what
we would expect from most theories of gender
stratification and from human capital models
of returns to schooling. However, this effect is
not strong enough to counterbalance the negative impact of girls work that tends to come together with adult women working.
We need to be cautious in interpreting the negative coefficient for girls work on girls literacy
as a causal effect. As we noted above, many have
argued that it may be higher rates of girls working are a consequence of girls not going to school.
Probably some of the negative relationship between girls work and girls literacy is a result
of this effect of low schooling on more work.
However, the high correlation between adult
womens labor force participation and girls labor force participation suggests that girls work
rates are determined primarily by cultural and
economic factors and not from their schooling
opportunities. The chance of girls working is
predicted quite well by the rates of adult womens work: where women work, girls are more
likely to work too. We might even interpret the
negative effects of adult womens labor force
participation as shown in model 1 as an instrument for the effect of girls labor force participation. It appears that girls work rates reduce their
literacy rates more than the reverse. Parents are
more likely to keep their daughters at home from
school if they can put them to work. As a result,
model 1 documents an unusual finding in the
gender stratification literature: higher rates of
womens labor force participation actually increase some gender inequalities.
Alternative causal interpretations of the negative association of womens or girls labor
force participation and girls schooling seem
harder to sustain. Unmeasured cultural proscriptions on girls or womens mobility outside
the house would produce a positive, not a negative, association between work and schooling.
Poor schooling opportunities that would reduce girls education and increase girls work

138

WORLD DEVELOPMENT

would not likely increase adult womens labor

force participation. 5 The most plausible interpretation of the negative association of womens labor force participation and girls
education would seem to us the greater work
(and household) demands on girls where their
mothers are in the labor force.
The other indicator of gender stratification in
the model, patrilocal exogamy, has the expected negative sign and is statistically significant. In areas of high patrilocal exogamy,
girls are much less likely to be literate than
boys. Parents investment in a daughters human capital is more likely when she lives in
the same village after marriage.
As expected, girls equality with boys literacy
rates follows a U-shaped relationship. Initially,
as the odds of male literacy improve, girls odds
fall further behind and the gender gap widens.
However, eventually, when boys literacy rates
pass a certain level, the odds of girls literacy
begin to catch up with boys and the gender
gap in literacy narrows. The point at which
the gender gap begins to narrow can be estimated as the inflection point of the quadratic
curve. These calculations reveal that when
boys literacy levels have reached 93%, the gender gap in literacy begins to decline. This result
suggests that the queuing effect may be very
strong. Only when almost all boys are literate
does the gender gap begin to narrow.
Fortunately, other development effects tend to
narrow the gender gap in literacy. Most interestingly, educational effort, as measured by the
number of teachers in the district, has an important positive association with girls literacy. The
more teachers available, the lower the gender
gap in literacy rates. Girls education appears
to be especially responsive to educational effort,
a result that confirms findings from other countries (Filmer, 1999; Lloyd, El Tawila, Clark, &
Mensch, 2001). Again, we need to be cautious
about inferring a causal effect here since it is
quite plausible that where daughters are sent to
school, more teachers are required. But if some
part of this association is due to the causal impact of schooling opportunities on girls equality, this association has important policy
implications: investments in education (teacher
salaries are a majority of education expenses)
not only raise overall literacy levels, but also
tend to reduce gender inequalities in literacy.
District wealth also helps to reduce gender
inequalities in literacy. Even controlling for
investments in education and for the education
of sons, households in wealthier areas are more

likely to educate their daughters. These results

are similar to the development effects on girls
school attendance found in household surveys
in rural areas (Dreze & Kingdon, 2001). It is
not clear what mechanism would be linking district wealth and gender equality once boys literacy and educational effort are controlled. The
positive coefficient may even reflect the endogeneity of wealth: educating women produces more
wealth (World Bank, 2001), so the causal direction may flow from gender equality to wealth.
The only development measure that is unrelated to gender equality is urbanization. While
it is true that girls in urban areas do have higher
literacy rates that are closer to boys rates, this
is a function largely of greater wealth and secondarily of more teachers. Beyond these greater
resources found in cities and towns, there is no
tendency toward more equal schooling of boys
and girls because of urban life. 6
(b) Control variables
Of the remaining variables, percent landless
has the strongest relationship with the gender
gap in literacy. As predicted, areas with high
landlessness are associated with more equal odds
of boys and girls literacy. In India, groups that
are more dominant tend to be more patriarchal,
so it is not surprising that districts with more
landed households have more unequal access
to schooling. Tribal areas also tend to have more
equal literacy rates, although this effect is observed only after controls for wealth (tribal areas
tend to be poorer) and child labor (girls work
more often in tribal areas and this suppresses
their literacy rates). The coefficients for percent
scheduled caste and percent Muslim are not statistically significant. However, these proportions
do not vary dramatically across this sample of
districts, so it may be difficult to detect whether
they have any general social structural effects.
In addition, it is worth noting again that these
district-level coefficients should not be interpreted as if they were household level measures.
Finally, South India is not more gender equal for
literacy rates, as it tends to be for other dimensions of gender equality. 7
So far, the analysis has investigated gender
differences in literacy only. Because literacy
rates approach 100% in some districts, gender
differences in these districts may be muted by
ceiling effects. Thus, it is useful to test whether
the results reported in Table 4 also hold for
higher levels of education achieved by only a
minority of Indians. Table 5 reports results

GENDER DIFFERENTIALS IN LITERACY IN INDIA

139

Table 5. Maximum likelihood estimates of selected independent variables on the gender gap in literacy and
matriculation ages 2024: districts in India, 1991
Variable

Literacy without
child labor

Constant

!2.225***
(!11.45)
!0.109***
(!4.15)
!0.141***
(!4.88)
!0.073
(!0.41)
0.077
(0.82)

!1.751***
(!7.28)
0.111*
(2.42)
!0.143***
(!5.20)
0.061
(0.32)
0.106
(1.13)

Gender ratio adult

main:non-main
Odds of women being
in exogamous unions
Log ratio boys literacy/
illiteracy 2024
(Log ratio boys literacy/
illiteracy 2024)2
Log ratio boys
matriculates/nonmatriculates 2024
(Log ratio boys
matriculates/nonmatriculates 2024)2
Education effort
Housing index
Percent urban in
a district
Percent landless in
a district
Percent SC in district
Percent ST in district
Percent Muslims in
a district
Southern districts
Log ratio 514 girls
main/non-main
Log ratio 514 boys
main/non-main
Rho (Spatial
autocorrelation)

Literacy with
child labor

0.007***
(4.41)
0.010***
(3.65)
!0.225
(!1.41)
0.006***
(4.44)
!0.001
(!0.33)
0.002!
(1.65)
!0.001
(!0.61)
0.095
(1.08)

0.817***
(!26.31)

0.006***
(4.14)
0.009***
(3.62)
!0.244
(!1.60)
0.005***
(4.32)
!0.002
(!1.00)
0.001
(1.35)
!0.003!
(!1.82)
0.133
(1.62)
!0.274***
(!5.85)
0.270***
(5.26)
0.796***
(23.91)

Matriculates without
child labor
!2.299***
(!7.63)
!0.092***
(!3.64)
!0.176***
(!6.33)

Matriculates with
child labor
!2.305***
(!7.04)
0.007
(0.16)
!0.179***
(!6.49)

0.473*
(!2.00)

!0.523*
(!2.20)

!0.082
(!1.33)

!0.092
(!1.50)

0.006***
(4.31)
0.011***
(4.42)
0.082
(0.52)
0.003**
(2.79)
0.000
(0.05)
!0.001
(!0.49)
!0.002
(!1.35)
!0.040
(!0.47)

0.006***
(3.96)
0.011***
(4.44)
0.089
(0.57)
0.003**
(2.71)
0.000
(!0.20)
!0.001
(!0.54)
!0.003!
(!1.80)
!0.014
(!0.180)
!0.122*
(!2.58)
0.080
(1.63)
0.787***
(22.97)

0.802***
(24.54)

Not applicable.
Note: Numbers in parentheses are t-statistics.
Source: Vanneman and Barnes (2003).
!
p < 0.1.
* p < 0.05.
** p < 0.01.
*** p < 0.001.

for gender differences in Matriculation (10

years of education in the Indian school system).
The appropriate age range has been adjusted to
men and women 2024 years old and the comparisons with literacy for this age group are
also reported. In 1991, only 33% of Indian
men 2024 had matriculated (compared to

72% who were literate), and only 17% of Indian

women of that age had matriculated (compared
to 44% who were literate).
The determinants of gender differences in
matriculation are quite similar to the determinants of literacy. Most importantly, districts
with higher adult womens labor force

140

WORLD DEVELOPMENT

participation have lower levels of girls matriculation relative to boys just as they have lower
relative levels of girls literacy. 8 Similarly, most
of the other coefficients in the equation for
girls relative matriculation are quite similar
to the coefficients for girls relative literacy.
The factors that impede girls literacy appear
to impede girls higher levels of education.
The effects of girls and boys work are similar
for gender differentials in matriculation as for
differentials in literacy. The effect sizes are, surprisingly, somewhat smaller for the matriculation differentials when one might have expected
that work would have been more of an obstacle
for adolescent girls schooling than for young
girls. Nevertheless, districts where more girls
work are districts where fewer girls matriculate
relative to boys. Adding the effects of girls work
again changes the sign of the coefficient for
womens work to positive, although not statistically significant as was the case for girls literacy.
On the whole the results for matriculates reinforce the results obtained for literacy.
4. DISCUSSION
The results of our analyses are interesting and
rather unexpected. A widening gender gap in
education is associated with higher proportions
of women in the labor force. Theories as disparate as human capital and feminist empowerment would suggest that adult womens
employment should encourage girls education.
In India, this is not the case. In fact, girls literacy is further behind boys literacy in districts
with more adult womens employment. The best
explanation for this seems to be that the more
adult women who work, the more girls who
work. Where girls are in the labor force, they
have less of a chance of attending school and
learning to read and write. We suspect,
although we could not test, that the greater demands for daughters housework in households
where their mothers work outside the house also
helps mediate this effect of womens labor force
participation on lower girls education.
The linkage between child labor and schooling has been much debated recently (Anker,
2000; Basu, 1999). Work does not necessarily
imply withdrawal from school. Many children
both attend school and engage in economically
productive work before and after school hours,
or during school vacations. Sometimes the
additional income from child labor can finance
school expenses (Patrinos & Psacharopoulos,

1997) although household strategies that send

girls to work to finance boys schooling would
increase gender inequalities (Greenhalgh,
1985). Some of the association between child labor and school withdrawal undoubtedly reflects
a causal flow from school withdrawal to work,
rather than the other way around. However,
even with these reservations, the Indian data
provide evidence that a higher proportion of
girls in the labor force reduces their opportunities for literacy.
As expected, indicators of male patriarchy
like patrilocal exogamy increase the gender
gap in education. Unlike the results for womens labor force participation, this finding is
consistent with prior research on other gender
inequalities. Areas where marriage patterns require wives leaving their natal communities,
the gender gap in child mortality is wide
(Kishor, 1993). Thus, gender gaps in education
are part of patriarchal systems that generate
gender inequalities in other areas. Educating
girls is not simply a question of more universal
education so that girls can catch up with boys.
Educational attainment is gendered; in some
places, girls start out further behind boys,
regardless of the resources available. How much
further behind they are depends on the
more general patterns of gender values in the
culture.
A second surprise, although others might not
find this so unexpected, is the consistent positive
relationship between economic development
and more gender equality in literacy. The availability of teachers and household wealth are
associated with both boys and girls education,
but the association is stronger with girls education so that economic development reduces
the gap between boys and girls outcomes. The
development effects are also consistent with the
strong queuing effects: the gap between boys
and girls literacy does not begin to narrow until
almost all boys are literate. In that environment,
more family and community resources differentially benefit girls whose education is a lower priority when resources are scarce.
The research literature is mixed on the relationship between development and gender outcomes. For example, Indian district level
analyses on child survival do not find that girls
chances for survival are any better in more
developed districts (Kishor, 1993) nor is there
evidence of gender queuing in child survival.
For literacy, however, more developed areas
have lower gender gaps. This analysis therefore
provides some support for a modernization the-

GENDER DIFFERENTIALS IN LITERACY IN INDIA

sis for the reduction of gender inequality at

least in the area of literacy. While past work focuses more on the role of urbanization and
industrialization as agents of modernization
(Kabeer, 1994; Kuznets, 1955), our analysis
finds neither to be important once educational
effort and wealth are controlled. This seems to
indicate that narrowing the gender gap in education is primarily a question of the availability
of economic resources.
This study lies at the crossroads of the gender
inequality discourse on one hand and the literacy discourse on the other. We feel that these
results will help further our understanding of
both areas. Gender stratification research is
again reminded that outcomes are different,
and gender gaps in educational attainment are

141

generated quite differently than gender gaps in

other areas. In particular, womens access to
work does not reduce all gender inequalities.
For young girls, adult womens participation
in the labor force sometimes means that girls
will also work, thus reducing their opportunities for school. Educational research is reminded that decisions to educate children
remain a gendered process. In most part of
India, girls schooling remains a second priority. As a result, womens ability to read and
write is poorly developed. Without these skills,
their access to and understanding of the many
changes surrounding their lives remains limited.
These limitations will be a persistent drag on
attaining gender equalities in all other aspects
of life.

NOTES
1. The 1991 data offer some advantages in completeness. We believe these relationships between womens
labor force participation and gender differentials in
literacy reflect enduring constraints on Indian households. We have tested a similar model with 1981 data,
although with a somewhat different measure of economic development, and find a similar pattern of results.
(Results available on the website.)
2. These include Arunachal Pradesh, Delhi, Goa,
Manipur, Meghalaya, Nagaland, Pondicherry, Sikkim,
and Tripura. The two island territories, Lakshadweep,
and the Andaman and Nicobar Islands, have been
dropped because their lack of contiguous neighbors
precludes their inclusion in the spatial autocorrelation
adjustments. In addition, the small disconnected district
of Daman and Diu is not included in the analysis.
3. This can be shown algebraically (where G = the
ratio of literate to illiterate girls and B = the ratio of
literate to illiterate boys):
X
bi xi :
1
lnG=B b0 b1 ln B

Adding ln B to both sides:

X
ln G b0 b1 ln B ln B
bi xi
X
b0 b1 1 ln B
bi xi :

4. These controls are also related to womens labor

force participation rates in India. Analyses available on
the website show that adult women tend to work in
tribal areas, in the South, and areas with low Muslim
populations and less landlessness. Girls labor force
participation is related to much the same set of factors.

If the (strong) relationship with womens labor force

participation rates is held constant, poverty emerges as
an additional factor predicting girls work.
5. It is still possible that this interpretation has the
causal order reversed. We might imagine that girls
illiteracy could cause higher womens labor force
participation: low chances for girls schooling could
cause high work rates for girls, and these higher rates as
young girls could cause higher work rates for adult
women because of their childhood experience. But this
scenario seems less plausible.
6. We had speculated that economic development and
educational effort might affect the gender gap in literacy
by affecting how quickly the schooling queue was
educated so that there might be curvilinear associations
of development with gender equality. In analyses not
reported in detail here, we entered a quadratic term for
each of these development variables. These models failed
to show any curvilinear relationship of wealth or
urbanization with gender equalities of literacy. Unlike
the Dollar and Gatti (1999) crossnational curvilinear
results, Indian districts have only a restricted range of
development that corresponds better with the earlier
development stages where they found a weaker positive
relationship. There was some evidence of a curvilinear
effect of educational effort: the positive relationship
became weaker at high levels of educational effort.
7. Although these other factors reviewed above are also
associated with gender differentials in literacy, they do
not affect the labor force participation relationships that
are the focus of these analyses. In a test for the

142

WORLD DEVELOPMENT

robustness of our conclusions, we calculated a series of

models that sequentially dropped each control variable.
In all results, womens labor force participation had a
statistically significant negative association with girls
relative literacy before controls for child labor and a
statistically significant positive association after controls
for child labor. (Results available on the website.)
8. Some of this effect on gender differences in matriculation rates is a result of the gender differences in

literacy rates. Girls who never become literate have no

chance of becoming matriculates. Therefore, if womens
labor force participation reduces girls literacy, it will
reduce girls matriculation as well. In analyses not
reported in detail here, we investigated the relative odds
of girls matriculation among literates only, that is, girls
relative odds of going beyond literacy to matriculation.
For these analyses too, adult womens labor force
participation was associated with reduced chances of
girls matriculation.

REFERENCES
Agarwal, B. (1994). A field of ones own: Gender and land
rights in South Asia. Cambridge: Cambridge University Press.
Anker, R. (2000). Conceptual and research frameworks
for the economics of child labour and its elimination.
Geneva: International Labour Office.
Arnold, F., Kishor, S., & Roy, T. K. (2002). Sexselective abortions in India. Population and Development Review, 28(4), 759785.
Banerji, R. (1997). Why dont children complete primary
school? A study of a low income neighborhood in
Delhi. Economic and Political Weekly, 32, 20532063.
Barro, R. J. (2001). Human capital and economic
growth. American Economic Review, 91(2), 1217.
Basu, A. (1992). Family size and child welfare in an
urban slum: some disadvantages of being poor but
modern. In Cynthia Lloyd (Ed.), Fertility, family
size, and structure: Consequences for families and
children. New York: The Population Council.
Basu, K. (1999). Child labor: Cause, consequence, and
cure, with remarks on international labor standards.
Journal of Economic Literature, 37(3), 10831119.
Basu, K., & Van, P. H. (1998). The economics of child
labor. American Economic Review, 88(3), 412427.
Behrman, J. R., Foster, A. D., Rosenzweig, M. R., &
Vashishtha, P. (1999). Womens schooling, home
teaching, and economic growth. Journal of Political
Economy, 107(4), 682714.
Beneria, L., & Sen, G. (1981). Accumulation, reproduction, and womens role in economic development:
Boserup revisited. Signs: Journal of Women in
Culture and Society, 7(2), 279298.
Beteille, A. (1986). The concept of tribe with special
reference to India. European Journal of Sociology, 27,
297318.
Beutel, A. M., & Axinn, W. G. (2002). Gender, social
change, and educational attainment. Economic
Development and Cultural Change, 51(1), 109134.
Boserup, E. (1970). Womens role in economic development. New York: St. Martins Press.
Carr, R. V., Wright, J. D., & Brody, C. J. (1996). Effects
of high school work experience a decade later:
Evidence from the National longitudinal survey.
Sociology of Education, 69(1), 6681.
Chafetz, J. S. (1984). Sex and advantage: A comparative,
macrostructural theory of sex stratification. Totowa,
NJ: Rowman and Allenheld.

Cigno, A., & Rosati, F.C. (2000). Why do Indian

Children Work, and is it Bad for Them? Discussion
Paper No. 115, Bonn, Germany: Institute for the
Study of Labor (IZA).
Dharmalingam, A., & Morgan, S. P. (1996). Womens
work, autonomy, and birth control: Evidence from
two South Indian villages. Population Studies, 50(2),
187201.
Dollar, D., & Gatti, R. (1999). Gender inequality,
income, and growth: are good times good for
women. Policy research report on gender and development: Working Paper Series No. 1, Washington,
DC: World Bank.
Doreian, P. (1981). Estimating linear models with
spatially distributed data. In S. Leinhardt (Ed.),
Sociological methodology (pp. 359388). San Francisco: Jossey-Bass.
Dreze, J., & Kingdon, G. G. (2001). School participation
in rural India. Review of Development Economics,
5(1), 124.
Dreze, J., & Saran, M. (1995). Choice welfare and
development: A Festschrift in honor of Amartya K.
Sen. In K. Basu, P. Pattanaik, & K. Suzumura
(Eds.), Primary education and economic development in China and India: Overview and two case
studies (pp. 182241). Oxford: Oxford University
Press.
Dubey, A., & Gangopadhyay, S. (1998). Counting the
Poor: Where are the Poor in India. Sarvekshana,
Analytical Report No. 1. New Delhi: Department of
Statistics, Government of India.
Dyson, T., & Moore, M. (1983). On kinship structure,
female autonomy, and demographic behavior in
India. Population and Development Review, 9(1),
3560.
Emerson, P. M., & Souza, A. P. (2003). Is there a child
labor trap? Intergenerational persistence of child
labor in Brazil. Economic Development and Cultural
Change, 51(2), 375398.
Filmer, D. (1999). The Structure of Social Disparities in
Education: Gender and Wealth. Policy Research
Report on Gender and Development: Working Paper
Series No. 5, Washington, DC: World Bank.
Filmer, D., & Pritchett, L. H. (2001). Estimating wealth
effects without expenditure dataor tears: An application to educational enrollments in states of India.
Demography, 38(1), 115132.

GENDER DIFFERENTIALS IN LITERACY IN INDIA

Forsythe, A., Korzeniewicz, R. P., & Durrant, V. (2000).
Gender inequalities and economic growth: A longitudinal evaluation. Economic Development and Cultural Change, 48(3), 573617.
Greenhalgh, S. (1985). Sexual stratification: The other
side of growth with equity in East Asia. Population
and Development Review, 11(2), 265314.
Hill, M. A., & King, E. M. (1993). Womens education in
developing countries: An overview. In E. M. King, &
M. A. Hill (Eds.), Womens education in developing
countries: Barriers, benefits, and policies (pp. 150).
Baltimore: Johns Hopkins University Press for the
World Bank.
Jejeebhoy, S. J., & Sathar, Z. (2001). Womens autonomy in India and Pakistan: The influence of religion
and region. Population and Development Review,
27(4), 687712.
Kabeer, N. (1994). Reversed realities: Gender hierarchies
in development thought. London: Verso.
Kishor, S. (1993). May god give sons to all: Gender and
child mortality in India. American Sociological
Review, 58(2), 247265.
Klasen, S. (2002). Low schooling for girls, slower
growth for all? Cross-country evidence on the effect
of gender inequality in education on economic development. World Bank Economic Review, 16(3),
345373.
Knodel, J., & Jones, G. (1996). Post-Cairo population
policy: Does promoting girls schooling miss the Mark?
Population and Development Review, 22(4), 683702.
Kuznets, S. (1955). Economic growth and income inequality. The American Economic Review, 45(1), 128.
Levison, D., Moe, K. S., & Knaul, F. M. (2000). Youth
education and work in Mexico. World Development,
29(1), 167188.
Lloyd, C. B., El Tawila, S., Clark, W. H., & Mensch, B.
S. (2001). Determinants of Educational Attainment
among Adolescents in Egypt: Does School Quality
Make a Difference? New York: Population Council,
Working Paper Number 150.
Malhotra, A., & Mather, M. (1997). Do schooling and
work empower women in developing countries?
Gender and domestic decisions in Sri Lanka. Sociological Forum, 12(4), 599630.
Mason, K. O. (1986). The status of women: Conceptual
and methodological issues in demographic studies.
Sociological Forum, 1(2), 284300.
Meyer, J. W., Ramirez, F. O., & Soysal, Y. N. (1992).
World expansion of mass education, 18701980.
Sociology of Education, 65(April), 128149.

143

Miller, B. D. (1981). The endangered sex: Neglect of

female children in rural North India. Ithaca: Cornell
University Press.
Murthi, M., Guio, A., & Dreze, J. (1995). Mortality,
fertility, and gender bias in India: A district-level
analysis. Population and Development Review, 21(4),
711744.
Ord, K. (1975). Estimation methods for models of
spatial interaction. Journal of American Statistical
Association, 70(349), 120126.
Patrinos, H. A., & Psacharopoulos, G. (1997). Family size,
schooling, and child labor in peruan emprical
analysis. Journal of Population Economics, 10(4),
387405.
Probe Team (1999). The public report on basic education
in India. New Delhi: Oxford University Press.
Ravallion, M., & Wodon, Q. (2000). Does child labor
displace schooling? Evidence on behavioral responses
to an enrollment subsidy. The Economic Journal,
110(462), C158C175.
Ray, R. (2000). Analysis of child labor in Peru and
Pakistan. Journal of Population Economics, 13(1),
319.
Rosenzweig, M., & Schultz, T. P. (1982). Market
opportunities, genetic endowments, and intrafamily
resource distribution: Child survival in rural India.
The American Economic Review, 72(4), 803815.
Sopher, D. E. (1980). An exploration of India: Geographical perspectives on society and cultures. Ithaca,
NY: Cornell University Press.
Srinivas, M. N. (1989). Cohesive role of sanskritization
and other essays. Delhi: Oxford University Press.
Thomas, D. (1994). Like father, like son; like
mother, like daughter: Parental resources and child
height. Journal of Human Resources, 29(4), 950989.
Vanneman, R., & Barnes, D. (2003). Indian Development
District Database. Center for Population, Gender
and Social Inequality, University of Maryland,
College Park. Unpublished Codebook.
Visaria, P., Gumber, A., & Visaria, L. (1993). Literacy
and primary education in India 198081 to 1991.
Journal of Educational Planning and Administration,
7(1), 1362.
Wils, A., & Goujon, A. (1998). Diffusion of education in
six world regions: 196090. Population and Development Review, 24(2), 357368.
World Bank (2001). Engendering development: Through
gender equality in rights, resources, and voice. New
York and Washington, DC: Oxford University Press
and the World Bank.

Available online at www.sciencedirect.com