Religion and Death in India

Religion and Death in India
Sonia Bhalotra (University of Bristol, UK) and Arthur van Soest (RAND and
University of Tilburg)
First Draft; to be updated before June.
Abstract
Muslim children in India face lower mortality risks than Hindu children. This is
surprising because one would have expected just the opposite: Muslims have, on
average, lower socio-economic status, higher fertility, shorter birth-spacing, relatively
low status of women, and are a minority group in India that, in principle, might live in
areas that have relatively poor public provision. Although higher fertility amongst
Muslims as compared with Hindus has excited considerable political and academic
attention in India, higher mortality amongst Hindus has gone largely noticed. This
paper investigates the seeming puzzle. We find that the largest part of the Muslim
advantage stems from where in India they live, in which state, and whether rural or
urban. Since fertility is higher amongst Muslims, their children tend to be born a bit
later which, given trend improvements in health technology and services, contributes
to their survival advantage. But, overall, only about 50% of the religion differential
can be explained by differences in characteristics. Using a supplementary data set, we
suggest that further explanation may be sought in considering religion-differences in
maternal health and breastfeeding behaviour.
1
Religion and Death in India
Sonia Bhalotra and Arthur van Soest
1. Introduction
In his keynote address at the annual conference of the European Society of
Population Economics in 2005, Timothy Guinnane argued that religion explains much
of the variation across Europe in the level and the timing of the onset of the fertility
transition; in particular, religion effects seem to dominate the effects of women’s
education. The extent to which culture effects are compositional is identified as a
“remaining puzzle”. He states that “this may be an uncomfortable fact for
economists”, but “religion really matters”. Although there is a recent surge of interest
amongst economists in ethnicity effects, especially in education (e.g. Fryer and Levitt
2004, Wilson et al 2005), there remains limited research on religion effects in
economics, and especially so in the area of health.
The analysis in this paper is motivated by the observation that Muslim children
in India face lower mortality risk than Hindu children. This is puzzling because
knowledge of religion-differences in the predictors of mortality risk suggests that the
opposite should be the case. In particular, Muslims have, on average, lower socioeconomic status, higher fertility and shorter birth intervals, and they are thought to
accord a lower status to women within the household.1 Moreover, as they are a
minority group in India, we might imagine that the areas in which they are
concentrated receive relatively poor public services. Although higher fertility amongst
Muslims as compared with Hindus has excited considerable political and academic
attention in India, higher mortality amongst Hindus has gone largely noticed. This
paper investigates this seeming puzzle, quantifying the extent to which the religiondifferential in childhood mortality risk can be explained by differences in observed
characteristics between the two communities. In this way, it estimates the size of the
(residual) “religion effect”. For instance, we may find that a part of the observed
survival advantage associated with being Muslim as opposed to Hindu is
compositional and can be captured by, for example, moving from one region to
1
There is considerable evidence that, where mothers have greater power in the household or
the community, children benefit in terms of their health and education (e.g. Maitra 2004,
Thomas 1990).
2
another, but that the rest can only be captured by adopting Muslim practices. The
latter would be described as the religion effect. It describes how, for given inputs, the
outcome is altered by agency, where, in this case, agency refers to the parents’
membership of a particular community. Agency is relevant to understanding how the
effectiveness of public resources may depend upon their distribution, in this case,
across religious communities.
The agency of women has attracted attention in the context of public policy in
both rich and poor countries. While most of the evidence pertains to women’s agency
within the household (e.g. Lundberg, Pollak and Wales 1997), there is also some
evidence of agency effects at the community or the state level (Edlund and Pande
2001, Chattopadhyay and Duflo 2004). Why does female agency matter? There are no
conclusive results here but the evidence suggests that women have different
preferences from men, they are more likely to spend household resources on food
rather than tobacco (e.g. Hoddinott and Haddad 1997), and more likely to spend
village resources on water rather than roads (Chattopadhyay and Duflo 2004). Agency
associated with religion raises a different set of questions but the idea is similar. If
Muslim parents are indeed better agents than Hindu parents for purposes of child
health, then the natural question of interest is why. Can we identify what practices
Muslims follow that, in principle, Hindus can emulate, such as early or more
sustained breastfeeding? Or are there inalienable characteristics of Muslim parents
such as stronger preferences for child health, possibly different by the gender of the
child?
To guide design and interpretation of the study, it is useful to consider essential
features of childhood death in developing countries. Most child deaths occur in the
first four weeks of life (the neonatal period), after which the risk of death tends to
decline continuously. Neonatal risk factors tend to be “biological”, the role of
environment and care rising as the child ages (e.g. Wolpin 1997). In the womb and at
birth, boys tend to be more vulnerable than girls, as a result of which childhood death
rates for boys tend to be higher in societies that do not exhibit any preference for
children of a particular gender. India is not such a society; there are numerous
accounts of son-preference amongst Indians (e.g. Arnold et al 19xx). Although girls
are, by nature, born with a survival advantage, this is seen to be eroded with age and,
for India as a whole, the data we use suggest that the risk of death for girls begins to
3
exceed that for boys by the age of one. The girl disadvantage has been observed to
increase with birth-order (e.g. DasGupta 1987), consistent with the notion that parents
the world over like to have at least one boy and one girl. It is the arrival of more and
more girls that appears, in India, to reduce the “value” of the further girls to their
parents. So, individual mortality risk varies with child age, gender and birth-order.
But does the difference in mortality risk between Hindus and Muslims within India
also vary along these dimensions? We find that it does.
The religion differential in mortality risk persists from birth till at least age five,
but the percentage (or relative) differential shrinks, especially after the age of one
(Figure 1). This suggests a plausible role for factors like maternal health, antenatal
care and delivery complications that influence risk earlier on. In particular, it is
possible that the factors that matter more as the child ages (e.g. sex-preference or
access to health clinics) are actually more favourable amongst Hindus, and that the
Hindu-disadvantage apparent in under-5 mortality rates is entirely on account of their
experiencing higher infant (i.e. under-1) mortality rates. The tendency for Hindus to
experience “excess” child mortality (i.e. greater than experienced by Muslims) is
greater amongst girls, and especially girls of higher birth-order. We therefore estimate
models of mortality risk by age, birth-order and religion.
The main results are as follows. On average, only about 50% of the religion
differential can be explained by characteristics, this percentage being greater for
higher-order births. For instance, for first-order births, the risk of under-5 death is
11.9% for Hindus and 10.0% for Muslims, a total differential of 1.9%-points.
Although some of the Hindu disadvantage is explained by their membership of
scheduled castes and tribes (low castes in India), mortality amongst high-caste Hindus
is also higher than amongst Muslims (a differential of 0.83%-points for first-borns).
The largest part of the Muslim advantage stems form where in India they live, in
which state, and whether rural or urban. Since fertility is higher amongst Muslims,
their children tend to be born a bit later which, given trend improvements in health
technology and services, contributes to their survival advantage. These factors
overwhelm the characteristics that predict an advantage for Hindus, namely their
higher levels of parental education, and their longer birth-intervals.
Lower mortality amongst Muslims in India has been noted by Shariff (1995),
Kulkarni (1996) and Bhat and Zavier (2005) The only multivariate analysis that we
are aware of appears in our earlier work (Bhalotra and van Soest (2004). Using data
4
from the state of Uttar Pradesh, where 17% of the population is Muslim, we find that
Muslim-status is associated with a 1.6%-point reduction in neonatal mortality. This is
a remarkably large religion effect, given that it obtains after controlling for birthspacing (and fertility), the survival status of the previous sibling, maternal age at birth,
education of the mother and father, caste, the gender and birth-order of the child, and
the child’s cohort.
Public awareness of the religion differential in mortality is limited and there
appears to be no previous research that has set out to explain it. In contrast, higher
fertility amongst Muslims as compared with Hindus has been the subject of recent
academic research (see Bhat and Zavier 2005 for an excellent survey), sparked by
political controversy and concerns that the Muslim population in India is growing
“too fast”. Census data show that, in 1951, Muslims constituted 9.9% of the
population of India and, in 2001, this percentage had risen to 13.4%. Over the same
period, the percentage of Hindus had declined from 85% to 80.5%. Higher fertility of
Muslims as compared with Hindus has been noted since before the partition of India
and the creation of Pakistan (e.g. Davis 1951). This differential has widened over time
and it is estimated that, in the late 1990s, the total fertility rate amongst Muslims was
29% higher than amongst Hindus. Lower infant and child mortality amongst Muslims
as compared with Hindus is also not a recent phenomenon (see Bhat and Zavier
2005).2 In the late 1990s, it was 23% lower but, as discussed in section 2 below, the
mortality differential appears to have narrowed over time. As there is considerable
evidence of mortality and fertility differentials being positively correlated over time
and across households, we may have expected a positive fertility differential between
two communities to be associated with a positive mortality differential. We find the
opposite. Our analysis shows that, while higher fertility amongst Muslims does tend
to raise mortality risk amongst their children, this effect is relatively small, and
outweighed by other factors.
2
See their Table 6, p.389, which reports the relevant means from the National Sample
Surveys of 1963/4 and 1965/6, the Sample Registration Survey of 1979, Census 1981 and
1991, and the National Family Health Surveys (NFHS) of 1991/2 and 1998/9. In this paper,
we use the NFHS survey for 1998/9, which contains information on births and child deaths
over a span of 39 years. While data on mortality from surveys such as the NFHS can be
subject to large sampling errors, SRS and Census data are not likely to suffer this problem.
The religion difference investigated here therefore seems real.
5
2. Data
The data are drawn from the second round of the National Family Health
Survey of India conducted in 1998/9 (see IIPS and ORC Macro 2000). This
interviewed ever-married women aged 15-49 in 1998-99 and obtained from them
complete fertility histories, including number of live births, birth intervals, and the
time and incidence of child deaths. The survey also contains information on relevant
individual, household and community characteristics.
Births in the sample occur during 1961-1999 and we use all of this
information, which provides not only the advantage of efficiency in estimation but
also a relatively long panel of children per mother. This allows us, in our one equation
model, to examine mortality risk by birth-order, with large samples available for
higher-order births. In the extended model that endogenises fertility, this has even
greater advantages (see section 4 below).
The data used in this paper are for 18 Indian states that, together, account for
more than 95% of India’s population. A list of the states with descriptive statistics
including their weight in the sample and the percentage of Hindus and Muslims in
each state is in Table xx. Not every state has enough Muslims to merit a state-wise
analysis, so we pool data from across the states and include state fixed effects. Our
objective is to address the broad question of how Muslims and Hindus in India as a
whole compare.
We drop the 11.05% of births in the survey that occur in households of
religions other than Muslim or Hindu. Amongst Muslims 1.35% of live births are
multiple (twin, triplet etc) and amongst Hindus the corresponding figure is 1.29%.
Rather than drop just these children, we drop all mothers who have a multiple birth,
which involves dropping 3.6% of children and 7.3% of mothers in the total sample.
The sample used has 202250 live births of 60622 mothers. It is standard practice in
the demographic literature to restrict the analysis to singletons as death risks are many
times higher for multiple births and, although relatively rare, they can skew the
statistics.
6
3. Descriptive Statistics
This section describes the religion-differential in mortality and the way in which it
varies with child age and birth-order. It also docuents a narrowing of this differential
over time. It then describes the explanatory variables used in the analysis.
The Dependent Variable
The sample analysed in this paper contains 86.9% Hindus and 13.1%
Muslims. Across the births in the data, which span the period 1961-99, average under5 mortality is 11.6% amongst Hindus and 9.4% amongst Muslims. So religion creates
a raw mortality differential of 2.2%-points, or 23.4% (Appendix Table 1). The
religion differential in infant and neonatal mortality is 1.5%-points and 1.0%-points
respectively (also see Figure 1). (These are big differentials relative, for example, to
the average annual rates of decline in mortality reported below). To allow the
parameters of the child health production function that is estimated to vary with child
age, the analysis is conducted for each of the neonatal (first month of life), infant (first
year of life) and under-5 (first five years of life) mortality rates.3 The difficulty with
estimating mortality risk year by year is that mortality risk in age 1-2 is conditional
upon survival to age 1.
The religion differential varies non-linearly with birth-order, tending to narrow
for second-births and then become larger for third and higher-order births than it is for
first-births. This pattern is evident in each of the three age-groups considered, though
most marked in the under-5 group (Table 3). Separate equations will be estimated in
this paper for first, second, third and higher-order births.
Over the 38-year period spanned by the data, under-5 mortality declined at an
average rate of 0.37%-points p.a. Separating the sample by religion, we find that the
average annual rate of decline in percent-points was 0.38 amongst Hindus and 0.30
amongst Muslims, indicates a gradual erosion of the Muslim advantage over time in
absolute terms. This is confirmed by comparing mortality rates for both religions for
the first and last decade of the sample-data. For births occuring during 1961-70,
Muslim children show a survival advantage of 1.45, 2.5 and 3.2 %-points in the
3
It has been estimated that the survival advantage exhibited by Muslims persists into
adulthood, with Muslims enjoying the greatest advantage in survival probabilities in the age
range 15-34 (see Bhat and Zavier 2005). As these authors remark, this phenomenon is underresearched. In this paper, the focus is entirely on childhood.
7
neonatal, infant and under5 age-groups respectively. The corresponding figures for
1990-99 are 0.92, 1.25 and 1.61%-points.
The religion differential has widened in relative terms. Consider the under-5
mortality rate. In the 1960s, this was 21.4% amongst Hindus and 18.2% amongst
Muslims, implying a differential of 17.6%. In the 1990s, the Hindu and Muslim rates
were 9.03% and 7.42%, which imply a differential of 21.7%.
The Independent Variables
These include the gender of the child, the birth-year of the child, the age of the
mother at the birth of the child, rural/urban location of the household, an indicator for
whether the household belongs to one of the “backward” castes, and a set of variables
indicating the educational level of the mother and of the father.4 All-India means by
religion are in Tables 4 and Figures 4-8. The corresponding state-specific means are
in Appendix Tables x. The sex ratio at birth is similar between the two communities.
Hindu mothers and fathers tend to be more educated and the religion gap in education
widens with the level of education. On its own, education would predict a Hinduadvantage in survival. However, other characteristics favour survival amongst Muslim
children. These include their being born a bit later in time (which follows from higher
fertility amongst Muslims), their being born to slightly older mothers on average (this
is again a function of fertility, but will vary by birth-order in our analysis), their being
less likely to live in rural areas, their being less likely to belong to a low-caste, and
their being concentrated in low-mortality states. The analysis to follow will indicate
the magnitude and significance of these tendencies.
4. The Econometric Model
The estimated equation is a reduced-form probit for mortality,
4
There is strictly no role for income or wealth in a health production function since incomes
and prices determine the level of inputs (e.g. Rosenzweig and Schultz 1983). However, it is
common to estimate a hybrid model that represents the reduced form of the technological
relation denoted by the health production function and the budget equation (see Strauss and
Thomas 1995). Data on household wealth are available only at the time of the survey and
since children in the sample are born over a long span of time, these data are not used.
Socioeconomic status is represented by the education of the mother and father of the child,
together with ethnicity and religion. It is worth noting that the level of per capita income or
expenditure (a common measure of income) depends on fertility which, in turn, is endogenous
to mortality. For this reason, it would probably be undesirable to include this variable even if
it were available for the entire sample of children.
8
(1)
Mij* =Φ( xi , xij,; θm) + umij ;
Mij=1 if Mij*>0 and Mij=0 if Mij*<0
Mij denotes an indicator variable with value 1 if child j in family i dies before a
specified age and 0 otherwise. For reasons discussed in section 3, we use three
alternative specifications of the dependent variable, with the cut-off age being one
month (neonatal), one year (infant) and five years (under-5) respectively. The vectors
xi and xij are exogenous explanatory variables that vary over children (xij, j=1,…,n)
and that do not vary over children (xi); these were described in section 3. The u term
denotes iid errors, and θ is a vector of parameters to be estimated. Equation (1) is
estimated for eight sub-samples: Hindu and Muslim children, of birth-orders one, two,
three, and four or higher. As we use information on all live births, we have the firstborn child of every mother, and first-borns constitute 30% of all births. Birth-orders
two, three and higher comprise 25%, 18% and 27% of the sample respectively.
Estimating the model by birth-order avoids having to introduce family-level
unobserved heterogeneity, and make the assumption that this is uncorrelated with the
included covariates.
Previous research on mortality that uses retrospective histories has tended to
left-truncate the data at ten or fifteen years prior to the survey date (e.g. Bhargava
2003, Bolstad and Manda 2001, Madise and Diamond 1995, Guo 1993, Curtis, et al
1993). One reason is that it is thought that measurement error in recollection of dates
of birth and death is likely to be greater the further back in time the event occurred.
We assume/expect that any recall error is similar across the religious communities.5 A
second reason that demographic research favours left-truncating the available birthhistory data is that they get more scarce the further back one goes in time, and will
contain a disproportionately large number of children born to young mothers. Thus, in
1998, the survey data will include births to mothers as young as 15 and as old as 49.
However, for 1968, the data will contain births to mothers who were 15 in 1968 and
45 in 1998, but it will exclude births to mothers who were 25 in 1968 because those
5
This issue was nevertheless examined in the following way for infant deaths. Since the data
reveal age-heaping at six-month intervals, the dependent variable was defined in terms of
death before or at 12 months of age, and the results compared with those obtained when the
dependent variable refers to death strictly before the age of 12 months. There were no
important differences (Arulampalam and Bhalotra 2004).
9
mothers are 55 in 1998 and too old to have been interviewed. To allow for this
triangularity in the data structure, we condition on mother’s age at birth.
Allowing Endogenous Fertility
A potential quibble with the specification (1) is that, if fertility and birthspacing are endogenous to mortality, then the age of the mother at the birth of the
child, the year of birth of the child and birth-order are endogenous variables (see
Rosenzweig and Wolpin (1995) for an essentially similar argument).
In an earlier paper that focused on neonatal mortality, we addressed this
problem by estimating mortality, birth-spacing and the probability of having another
child jointly using simulated maximum likelihood and exploiting the panel nature of
the data (Bhalotra and van Soest 2004). Mothers constitute the cross-sectional
dimension of the data. As mothers are observed repeatedly, in relation to every birth,
birth-order creates the time dimension of the panel. The model is dynamic in that it
includes “lagged mortality” (i.e. mortality of the preceding sibling), so that the effects
of mortality on birth-spacing and of birth-spacing on mortality are identified by
timing, avoiding the need to impose any arbitrary exclusion restrictions. An equation
for the probability of having a further child was introduced to allow for rightcensoring of the birth-interval data and it was identified using information on female
sterilization.6 A family-level random effect was introduced in every equation to
account for unobservables common to siblings, and these unobserved heterogeneity
terms were allowed to be correlated across equations. Children of different birth-order
were pooled, and birth-order included as a covariate. A reduced-form equation for
mortality of the first-born child was estimated together with the three main equations
to avoid the initial conditions problem that otherwise plagues dynamic models. An
advantage of this model is that endogeneity of not only the preceding birth-interval
but also of maternal age at birth, year of birth of child and birth-order, is taken
account of.
In this paper, we are not primarily interested in modelling the effects of birthspacing (or fertility) on mortality because, as discussed in section 1, we observe
higher death risks amongst Hindu children despite the lower fertility and shorter birthintervals of Hindus. Nevertheless, to investigate the importance of allowing for these
6
This the most common form of contraception in India and very difficult to reverse.
10
effects, and for the endogeneity of other fertility-related variables (maternal age, birth
year, birth-order), we estimated the four-equation model for neonatal mortality
described in our earlier paper, separately for Hindus and Muslims. We also
conducted, for this extended model, the decomposition exercise that we describe in
section x below. Any interesting insights into the central question of why Hindus have
higher mortality than Muslims that is yielded by moving to the extended model are
reported in discussion of the results (full results available on request). However, these
differences are, overall, not impressive. So, having investigated endogeneity and
concluded that it does not change our story about religion differences, we present the
simple model in equation (1) as the preferred model.
A further advantage of taking the single-equation approach here is that it can
be estimated for infant and under-5 mortality. This is relevant to our current project
because, as observed in section 3, the religion differential appears to increase with
child age in the range 0-5. It is not straightforward to incorporate these wider periods
of exposure to death risk into the extended model because, as explained above, it
relies upon the natural sequencing of the birth-spacing and neonatal mortality
processes for identification (details in Bhalotra and van Soest 2004).7
5. Results
The estimation results are in Tables 5-7, where marginal effect are presented
for each of the Hindu and the Muslim subsamples for under-5, infant and neonatal
mortality respectively. The discussion of effects of covariates below is, of course,
conditional on the other covariates in the model.
The difference in mortality risk faced by girls as opposed to boys varies with
religion and birth-order. Two fairly general patterns emerge in this respect. First,
across both communities and in every age-group, any excess risk faced by girls is
increasing in birth-order. This result cannot be explained in terms of increasing age
7
To repeat, the neonatal survival status of the preceding sibling (Mij-1) is necessarily revealed
before the birth of the index child. So, given controls for (correlated) unobserved
heterogeneity, Mij-1 is predetermined and its effects can be interpreted as causal. If neonatal is
replaced with infant survival status, this does not always work. The index child could be born
eight months after the preceding sibling and the preceding sibling could die between the
eighth and twelfth month of her life, which is after the birth of the index child. In this case,
Mij-1 is not predetermined. Replacing neonatal with under-5 is even harder. In our earlier
work, where the focus was on the bi-causal relation of mortality and birth-spacing, the choice
of the neonatal period was not so restrictive since mortality risk is more closely tied to birthspacing in the neonatal period than later.
11
of the mother at birth of the index child because it is obtained conditional upon
maternal age. It could nevertheless reflect biological factors such as maternal
depletion, which we would expect increaes with birth order, or else preferences in
which parents obtain diminishing returns from daughters at a more rapid rate than
from sons.8 Second, for under-5 and infant mortality, any excess risk faced by girls
(i.e. excess over the risk faced by boys) is smaller in Muslim households, whatever the
birth-order.9 It is noteworthy that this is not the case for neonatal mortality, for which
the effect of gender is not very different by religion. Girls face lower death risks in
every subsample, and these risks are significantly lower upto the third birth-order
amongst Hindus and upto the second birth-order amongst Muslims. This age-pattern
of the gender-effect suggests that the advantage that girls enjoy in Muslim households
may have more to do with postnatal care than with more biological variables like
birth-spacing. This is because neonatal death tends to be dominated by biological
causes (e.g. Wolpin 1997).
We find that the birth-order effect is different for the two communities,
implying that religion and birth-order interact in producing differential girl-boy
mortality. Consider the marginal effects of gender for under-5 mortality. First-born
girls are significantly less likely to die in both communities, though the marginal
effect is about twice as large among Muslims (-0.023 as compared with –0.012). In
the Muslim group, girls of higher birth-orders are no more or less likely to die than
boys. In the Hindu group, girls of birth-order three and higher are significantly more
likely to die than boys. The effect is substantial, being 0.010 percentage points at
birth-order three and 0.023 at higher birth-orders. In the case of infant mortality, the
female advantage at birth persists until birth-order two for both communities, and the
religion-differential is smaller. For first-borns, for example, it is –0.024 for Muslims
and –0.017 for Hindus. In no subsample do girls face a higher risk of death than boys.
The marginal effect of gender on neonatal mortality is of a similar magnitude. For
first-borns, it is –0.020 amongst Muslims and –0.016 amongst Hindus.
For every subsample considered, mortality risk is U-shaped in mother’s age at
birth, and the differences across subsamples are, in general, not significant. The
8
This is consistent with DasGupta (1990) who, based upon data from a region in the state of
Punjab, argued that “discrimination” was concentrated amongst later-born girls.
9
This is consistent with a similar finding reported for the case of infant mortality in Barooah
and Iyer (2005), using another Indian dataset.
12
turning points are reported in the Tables. The finding of a U-shaped relation is
common in the demography of developing countries. Although the coefficients on the
birth year of the child and its square are not individually significant in any of the
subsamples, they are jointly significant at 1% in every case, indicating that mortality
has been declining with time. The per annum rate of decline at the mean year is
reported in the Tables.
As discussed in section x, there are low-castes amongst Hindus, members of
which have been historically disadvantaged and, despite some positive discrimination,
continue to exhibit worse outcomes. This is confirmed by our finding that, for all
birth-orders, scheduled caste children suffer higher mortality risk. The marginal
effect is higher (0.02) for second and fourth or higher-order children than for first
and third borns (0.01). This somewhat odd pattern is repeated in the (smaller)
marginal effects associated with belonging to a scheduled tribe or to other backward
castes, but it is not clear what explains it. The effect of low-caste-membership on
mortality risk is weaker at younger ages. For infant and neonatal mortality, the effects
of SC and OBC membership are of similar size, the effect of ST-status being small
and mostly insignificant. The latter is somewhat surprising given that scheduled tribes
tend to live in geographically more isolated areas and to have weaker access to public
goods than other social groups (e.g. Banerjee and Somanathan 2005).
Consider the variables describing the level of education attained by each
parent where, in both cases, the omitted category refers to no education. The effect of
parental education on survival tends to increase with child age, consistent with an
increasing role for “nurture”. It is striking that a given level of maternal education has
much larger survival-raising effects amongst Hindus than amongst Muslims,
especially in the under-5 group. In this group, the effects of father’s education are
rather more similar between the religions. At low levels of maternal education, the
marginal effect tends to be larger at higher birth-orders. In most cases, the marginal
effect of a given level of education of the father is smaller than that flowing from the
same level of education of the mother, which is a result familiar from other studies of
the effects of parental education on health or educational outcomes for children in
developing countries.
In all but one of the eight samples for which results are reported in Table 3,
rural-residence is associated with higher under-5 mortality risk, and this added risk
tends to be greater in the Muslim community. In general, it does not vary with birth13
order, except that it is smaller for fourth and higher birth-orders in the infant and
under-5 groups. The marginal effects which, for first-borns, are of the order of 0.01
for Hindus and 0.02 for Muslims, do not change very much with child age. This
suggests that the rural effect probably denotes access to antental services, delivery and
neonatal care, such as iron-supplementation, tetanus for the pregnant mother, the
presence of a trained midwife at birth, or effective mangement of early complications.
The state effects are jointly significant in all subsamples but they are larger
and more significant for Hindus. The omitted state is Andhra Pradesh (see Table x). It
is only for the state of Rajasthan and at higher birth-orders that the state effects appear
significant for Muslims.
6. Decomposition of the Mortality Differential
(this section is to be completed and revised) The methods in this section
follow Fairlie (2006), who suggests an extension of the Oaxaca-Blinder procedures
commonly used in linear models to the case of non-linear models such as the probit or
the logit.
A decomposition of the observed differences in mortality risk between the
samples is presented in Tables x4-x13. Refer Table x4 which contains predicted
probabilities of mortality by birth order. Column 1 gives the average predicted
probability of neonatal mortality for Hindus. It takes the average of every variable
over the sample of Hindus and uses the parameter estimates obtained for the Hindu
sample (and reported in Table x1). Column 4 reports the average predicted probability
for Muslims. For every birth order, average predictions are much smaller in column 4
than in column 1, in line with the raw data. But these differences are on account of
both differences in characteristics (values of X) and differences in parameters (the
estimated coefficients). The rest of this section attempts to decompose the total
difference into these two categories.
Columns 2 and 3 of Table x4 report the "counterfactuals." Column 2 gives the
predictions for Hindus using the parameter estimates for Muslims. Similarly, column
3 gives the predictions for the Muslim sample using the parameters for Hindus. So,
comparing columns 1 and 2, we see that if Hindus retained their average
characteristics but had the parameters associated with Muslims, then the survival
chances of their children would improve. A similar result obtains comparing columns
14
3 and 4 for Muslims. These comparisons establish that the Hindu-Muslim difference
in mortality risk is not entirely explained by differences in characteristics. If, instead,
we compare columns 1 and 3 (or, instead, columns 2 and 4), we see that the religion
difference is not entirely explained by parameter-differences either. So, both matter,
but what is their relative explanatory power? The data that answer this question are in
Table x5.
For the second child for example, 0.23 percentage-points of the total mortality
difference of 0.81 percentage-points, a bit more than a fourth, is explained by
differences in the distribution of the explanatory variables in the Hindu and Muslim
samples. This is the difference between columns 1 and 3 in Table 1, that is,
parameters are kept constant at the Hindu estimates. The remaining 0.59 percentage
points is due to differences in the parameter estimates for Hindus and Muslims (the
difference between columns 3 and 4 in Table 1). For the fifth child, the total religion
differential in neonatal mortality is much larger, and a much larger share of it is
driven by sample composition differences. In total, combining all birth orders,
somewhat less than half of the differential is explained by differences in
characteristics (or sample composition) between the two communities.
Table x6 provides a more detailed decomposition. This comes with the health
warning that the decomposition of parameter-differences by variable seems to depend
strongly on the way in which variables are normalized, which benchmark categories
are used to construct dummies, etc., and the results seem hard to interpret
So,
although these results are presented for completeness in the right-hand panel of Table
x6, we will not discuss them. Instead, we focus on the contribution of particular sets
of variables to the role of observed characteristics in explaining the religion
differential in mortality.
For illustration, we present the detailed decomposition only for the fifth child.
(Results for other birth orders are available from the authors on request but would
take too much space to present here). As is clear from Table 2, the total differential
explained by characteristics for the fifth child is the sample composition is 0.64
percentage-points. Most of this difference (0.45 percentage-points) is explained by
state dummies. Muslims are better represented in the more developed states where
neonatal mortality is relatively low (Kerala, Assam, West Bengal; see Table xx which
shows percent of Muslims in each state). Similarly, Hindus are more likely than
Muslims to live in rural areas, by virtue of which, the estimates suggest that we
15
should expect Hindu mortality risk to be higher by 0.11 percentage-points. A further
0.22 percentage points of the religion-differential is explained by the fact that the
Hindu group includes scheduled and other backward castes which, as noted in the
previous section, experience higher mortality. Another important contributor to the
excess risk faced by Hindus is the fact that lagged mortality affects current morality
and is lower for Muslims. This persistence effect alone produces a religion-difference
in mortality risk of 0.14 percentage-points. As a result of higher total fertility amongst
Muslims, their children are born somewhat later than the Hindu children, on average.
This explains about 0.09 percentage-points of the mortality difference because of an
overall negative trend in neonatal mortality. Differences between the communities in
the proportion of girls in the sample, and in the age of the mother at the birth of the
index child contribute marginally to excess mortality amongst Hindus, the
contribution of each being of the order of 0.01 percentage-points.
Some of the observed differences in characteristics between the religious
communities predict lower mortality amongst Hindus. In this sense, they make a
negative contribution to explaining excess mortality amongst Hindus. For example,
educational attainment amongst fathers and mothers is higher in the Hindu than in the
Muslim community. Our estimates show that this would predict neonatal mortality
amongst Hindus to be 0.30 (mothers 0.14, fathers 0.16) percentage-points lower than
that amongst Muslims, other things equal. In other words, the distribution of parental
education in the sample contributes -0.30%-points to explaining higher mortality rates
amongst Hindus. A similar counteracting effect is found for the length of the birth
interval. Since birth intervals are shorter for Muslims than for Hindus, the estimates
imply that Muslims would have more neonatal mortality if other factors were kept
constant. The predicted effect is about -0.10 percentage-points.
Results from Extended Model
As discussed in section 1, Muslims have higher fertility than Hindus. A
seeming paradox is that this would have led us to expect higher childhood mortality
amongst Muslims, when in fact we find the converse. To what extent do our estimates
contribute to understanding/resolving this seeming paradox?
Higher fertility amongst Muslims is reflected in the specification for mortality
(equation 1) through a number of (associated) variables: shorter preceding birth
intervals, calendar year of birth, younger age of mother at first birth, and birth order of
16
child. A further association arises from the permitted correlation between the familyspecific unobserved heterogeneity terms in the mortality, birth-spacing and “fertility”
equations (equations 1-3). As we have seen, shorter birth-intervals amongst Muslims
would lead to neonatal mortality in this group being 0.10 percentage-points higher, a
result that underlines the paradox. However, Muslim children being born later (in
calendar time) on average, and the distribution of age at birth of Muslim mothers
lying to the left of the distribution for Hindu mothers contribute 0.09 and 0.01
percentage-points respectively to the finding of higher mortality amongst Hindu
children. As we have presented results by birth-order, the contribution of birth-order
to the religion differential in mortality is not explicit in the decomposition. However,
the birth order terms in both equations in Table x1 are insignificant, consistent with
Figure xx which shows that Hindu’s have higher neonatal mortality than Muslims at
every birth-order. Table x5 shows that the contribution of sample composition to the
total religion differential tends to increase with birth-order [might this follow from
there simply being more Muslim children at higher birth-orders because of higher
fertility? I should think about this later]. Looking at the correlations of the unobserved
heterogeneity terms across equations 1-3 (Table x3’), we find that Hindu families
with higher frailty also have a tendency towards longer birth intervals, while there is
no such relation for Muslims. However, amongst Muslims, but not Hindus, families
with higher frailty tend to be those with lower fecundity. [Estimates for both religions
reveal a negative association of the unobserved heterogeneity terms in the birthspacing and fertility equations.] Thus, while fertility and mortality interact in a
number of ways, some of which predict a positive association of these variables, the
religion differential in mortality seems to arise largely from factors that are unrelated
to the fertility differential.
7. Discussion
We find that a large part of the religion differential is unexplained by the
covariates in the estimated model. Even under the category of variation explained by
included characteristics, a large part of the variation is attributed to state fixed effects,
which are basically unobservables. What might be going on that we have not
captured?
17
There may be differences in cultural practices between Hindus and Muslims
that explain the observed differences in mortality. Thus, Hindus have a ritual of
offering the first-milk (colostrum) to the earth, despite scientific evidence that
colostrum is rich in antibodies that prevent recurrent infections. Another possibility is
that there are differences in diet between the two communities that might impact on
child health and survival; in particular, Muslims are more likely to eat meat
That Hindus are relatively less careful about their daughter’s health than are
Muslims is potentially consistent with the greater role of dowry amongst Hindus, and
the importance they attach to having sons perform religious rites. However, this
hypothesis has yet to be carefully investigated.
The wider set of questions put to women who had had a recent birth include
some questions that might provide insight into the puzzling fact of higher mortality
rates amongst Hindu as compared with Muslim children. Figure 5 shows the number
of hours from the time of birth to the first-breastfeed reported by women. The entire
curve suggests that Muslims put the baby to the breast sooner. This is especially
marked for the important first-hour following birth, when the nutrient-rich colostrum
milk is available to the child. Thus 16.9% of Muslim women breastfeed immediately
on birth, compared with 14.48% of Hindu women. Notice how small these fractions
are overall.
Another statistic that appears to shed light on the religion difference in
mortality relates to maternal health. Taking the conventional definition of undernourishment as indicated by a body mass index (BMI) smaller than 18.5, 36% of
Hindu women were undernourished at the time of the survey, compared with 31.4%
of Muslim women. Although there is no marked overall difference in socio-economic
status between the religious groups (see Bhat 2005), the higher BMI of Muslim
women may be a reflection of their diet (they are more likely to eat meat and so
possibly to have a more protein-rich diet at a similar SES-status), or else of their
phenotype (Indian Muslims originated North of the Indian subcontinent, where the
racial stock is typically taller and of bigger build). We have confirmed in our data that
low-BMI has a significant and large risk-increasing effect on neonatal (and later)
mortality. Indeed, we find that not only are fewer Muslim women undernourished but
the effect of low BMI on death risk seems to be smaller amongst Muslims.
An advantage of comparing two religious groups residing within the same
country is that there is less unaccounted heterogeneity. As we are drawing inferences
18
about religion here, it may be instructive to compare Muslims with Hindus in other
contexts, such as Pakistan or Indonesia. If we were to find that Muslim children
experience lower mortality than Hindu children in other countries too, we would have
greater reason to belief that there are “cultural” reasons behind the religion effect.
However, if we were to find instead that there is no survival advantage amongst
Muslim children, then we cannot reject the role of “culture” because culture may
operate very differently for Muslims in different countries. It has, for instance, been
recognised in discussion of Muslim-Hindu fertility differences, that this is the case
(e.g. Bhat and Zavier 2005)
8. Conclusions
Why do Hindu children face higher mortality risks than Muslim children in India?
To the extent that the religion-difference is attributable to observed characteristics, it
is overwhelmingly the geographical distribution of Muslims that confers a survival
advantage upon their children, with further contributions in this direction flowing
from persistence-effects and from the fact that Muslim children tend to be born later
in time, an artefact of higher Muslim fertility. These factors swamp the effects of birth
interval length and parental education that, on their own, predict higher mortality
amongst Muslim children.
To the extent that the religion-difference can be attributed to parameter
differences, it seems that, for a given maternal age at birth, the deleterious effect on
mortality is much larger amongst Hindus than amongst Muslims. This is consistent
with our suggestion that under-nourishment is less prevalent amongst Muslim
mothers. Another parameter difference that contributes to higher mortality amongst
Hindus is that the mortality-reducing effect of mother’s education is smaller amongst
Hindus. However this is small relative to the maternal-age effect, which is large
enough that it overwhelms a number of the counteracting influences evident in our
results.
On average, and for first to fourth children in the family, observed characteristics
explain less than half of the total religion differential. In the illustrated case of the
fifth child, characteristics explain 0.64 percentage-points, and parameter differences
explain 0.54 percentage-points. The residual “religion” effect will, of course, shrink
as the set of observable covariates available expands. In work in progress, we
19
investigate with a richer dataset, the inclusion of indicators of maternal health, and
maternal behaviours such as breastfeeding and hygiene into the model.
20
References
Arulampalam, W. and Bhalotra, S. (2004), Sibling death clustering in India:
genuine scarring vs unobserved heterogeneity. Revised June 2005. Available as
Discussion Paper 03/552, Department of Economics, University of Bristol.
Banerjee, A. and R. Somanathan (2005), The political economy of public
goods: Some evidence from India, Mimeograph, Cambridge MA: MIT.
Barooah, V. and S. Iyer (2005), Religion, literacy and the female-to-male
ratio, Economic and Political Weekly, 29 January: 419-427.
Bhalotra, S. and A. van Soest (2004), Neonatal mortality and birth-spacing in
India: Dynamics, frailty and fecundity, Working Paper, 04/567, University of Bristol,
December.
Bhargava, A. (2003) Family planning, gender differences and infant mortality:
evidence from Uttar Pradesh, India. Journal of Econometrics, 112, 225-240.
Bolstad, W. M. and Manda, S. O. (2001) Investigating child mortality in
Malawi using family and community random effects: a Bayesian analysis. Journal of
the American Statistical Association, 96, 12-19.
Curtis, S. L., Diamond, I., and McDonald, J. W. (1993) Birth interval and
family effects on postneonatal mortality in Brazil. Demography, 33(1), 33-43.
Davis, K. (1951), The Population of India and Pakistan, Princeton University
Press: Princeton, NJ.
Guo, G. (1993) Use of sibling data to estimate family mortality effects in
Guatemala. Demography, 30(1), 15-32.
Kulkarni, P.M. (1996)
Livi Bacci, M. (1967), Modernization and Tradition in the Recent History
of Italian Fertility, Demography, Vol. 4, No. 2: 657-672.
Maitra, P. (2004), Parental Bargaining, Health Inputs and Child Mortality in
India. Journal of Health Economics. 23(2), 2004, 259 – 291.
Madise, N. J. and Diamond, I. (1995) Determinants of infant mortality in
Malawi: an analysis to control for death clustering within families. Journal of
Biosocial Science, 27(1), 95-106.
Mari Bhat, P.N. and AJ Francis Zavier (2005), Role of religion in fertility
decline: The case of Indian Muslims, Economic and Political Weekly, 29 January:
385-402.
21
Mosher, WD, L.B. Williams and D.P. Johnson (1992), Religion and
Fertility in the United States: New Patterns, Demography, Vol. 29, No. 2. (May):
199-214.
Rosenzweig, M. and T.P. Schultz (1983b), Estimating a household production
function: Heterogeneity, the demand for health inputs, and their effects on birth
weight, The Journal of Political Economy, Vol. 91, No. 5, 723-746.
Rosenzweig, M. and K. Wolpin (1995), Sisters, siblings and mothers: the
effect of teenage childbearing on birth outcomes in a dynamic family context,
Econometrica, Vol. 63, No. 2, 303-326.
Shariff, A. (1995)
Strauss, J. and D. Thomas (1995), Human resources: Empirical modeling of
household and family decisions, in J. Behrman and T.N. Srinivasan (eds.), Handbook
of Development Economics, Vol 3A, pp. 1883-2023, Amsterdam: North-Holland
Press.
Thomas, D. (1990), Intra-household resource allocation: An inferential
approach”, Journal of Human Resources, 25(4), 635-664.
Wilson, D., S. Burgess and A. Briggs (2005), The dynamics of school
attainment of England’s ethnic minorities, Working Paper 05/130, CMPO, University
of Bristol.
Wolpin, K. (1997), Determinants and consequences of the mortality and health
of infants and children, Chapter 10 in Mark Rosenzweig and Oded Stark (eds.),
Handbook of Population and Family Economics, Volume 1A, Amsterdam: North
Holland.
Yun, M-S. (2004), Decomposing differences in the first moment, Economics
Letters 82, pp.275-80.
22
Table 1
Geographic Dispersion of Hindus & Muslims in India
State name
andhra pradesh
assam
bihar
gujarat
haryana
himachal pradesh
karnataka
kerala
madhya pradesh
maharashtra
orissa
punjab
rajasthan
tamil nadu
uttar pradesh
west bengal
new delhi
north-east
ALL INDIA
Number of % India’s %Muslims %Of all
observations population in state
Muslims
from this population that live in
state
each state
9088
8498
20211
9943
7422
6955
10373
4869
20134
11077
11115
3220
19772
9499
9835
26844
5845
8548
4.47
4.18
9.94
4.89
3.65
3.42
5.10
2.40
9.91
5.45
5.47
1.58
9.73
4.67
4.84
13.21
2.88
4.21
0.09
0.32
0.17
0.07
0.06
0.04
0.15
0.44
0.07
0.17
0.03
0.07
0.10
0.07
0.25
0.16
0.11
0.10
3.09
10.11
12.82
2.66
1.60
0.94
5.65
8.03
5.40
6.96
1.10
0.84
7.64
2.63
9.14
15.89
2.41
3.14
203248
100.00
0.13
100.00
Notes: Authors’ calculations from NFHS2 data
23
Table 2
Mortality Rates by Religion, Age and Gender
Mortality rates Muslim
Males
Neonatal
Infant
Under-5
Females
Neonatal
Infant
Under-5
Female vs male
(% points)
Neonatal
Infant
Under-5
Hindu
High-caste
Hindus
Muslim vs
Hindu (%
points)
Muslim vs
hc-Hindu
(% points)
0.047
0.071
0.093
0.056
0.085
0.111
0.054
0.080
0.102
-0.90
-1.40
-1.80
-0.69
-0.88
-0.90
0.037
0.062
0.091
0.047
0.079
0.116
0.045
0.075
0.106
-1.00
-1.70
-2.50
-0.80
-1.25
-1.50
-1.01
-0.85
-0.22
-0.96
-0.65
0.50
-0.89
-0.53
0.40
Notes: Excluding scheduled caste and scheduled tribes from amongst Hindus
generates the category defined as high-caste (hc) Hindus. Negative signs in the last
two columns indicate that Muslims have a survival advantage over Hindus in both
gender groups and in each of the three age groups, although the advantage is greater
amongst girls and increasing in child age. The last three rows show that females have
a survival advantage over males until the age of 1 year in both communities. After
that, the female advantage is gradually eroded and, in the case of Hindus, reversed.
24
Table 3
Religion Differences in Mortality by Birth-Order
Hindu, Muslim, Hindu, Muslim, Hindu, Muslim, Hindu, Muslim,
Child-1 Child-1 Child-2 Child-2 Child-3 Child-3 Child>= Child>=
4
4
mean of neonatal
t-test
0.064
0.053
3.53 (0.00)
0.045
0.040
1.96 (0.05)
0.042
0.033
3.04 (0.00)
0.053
0.041
4.60 (0.00)
mean of infant
t-test
0.093
0.079
3.65 (0.00)
0.075
0.064
2.91 (0.00)
0.071
0.052
4.77 (0.00)
0.088
0.069
5.89 (0.00)
mean of under5
t-test
0.119
0.100
4.42 (0.00)
0.104
0.286
3.40 (0.00)
0.309
0.272
5.49 (0.00)
0.337
0.098
8.54 (0.00)
Notes: p-values in brackets
25
Table 4
Means of Variables by Religion
birth year of child
proportion female
age ma at birth
prop rural
caste:
proportion SC
proportion ST
proportion OBC
mother’s education
no education
less than primary
primary
incomplete secondary
complete secondary & higher
father’s education
no education
less than primary
primary
incomplete secondary
complete secondary
higher
number of obs
Hindu
Muslim
t-test
85.882
86.558
-13.895
0.479
0.477
0.388
0.738
0.603
-12.146
45.768
0.216
0.106
0.335
0.000
0.000
0.255
85.421
55.965
25.989
0.619
0.090
0.072
0.120
0.099
0.648
0.123
0.067
0.108
0.053
-9.177
-17.459
3.148
5.483
23.953
0.316
0.112
0.094
0.211
0.119
0.148
0.392
0.148
0.087
0.199
0.096
0.078
-24.854
-16.957
3.609
4.831
10.653
30.849
175745
26505
Notes: Col. 3 reports a t-test of the null that the means by religion (col.1 & col. 2) are
equal. The p-values associated with the t-test is 0.00 in every row except proportion
female. SC= scheduled caste, ST= scheduled tribe, OBC= other backward caste.
26
Table 5
Under-5 Mortality
Probits by Religion and Birth-Order
1 if female
age ma at birth of child x10
square of age ma x100
child's birth year x10
square of birth year x100
1 if currently rural resident
scheduled caste
scheduled tribe
other backward caste
ma incomplete primary
1
Hindu,
Child-1
-0.012**
[0.003]
-0.320**
[0.003]
0.066**
[0.000]
-0.013
[0.031]
-0.001
[0.002]
0.015**
[0.003]
0.012**
[0.004]
-0.001
[0.005]
0.007
[0.004]
-0.005
[0.005]
2
Muslim,
Child-1
-0.023**
[0.007]
-0.252**
[0.008]
0.054**
[0.000]
-0.03
[0.080]
0
[0.005]
0.020*
[0.008]
0.007
[0.009]
-0.005
[0.011]
3
Hindu,
Child-2
0.004
[0.003]
-0.356**
[0.003]
0.065**
[0.000]
0.004
[0.037]
-0.002
[0.002]
0.013**
[0.003]
0.020**
[0.004]
0.014*
[0.005]
0.009*
[0.004]
-0.009*
[0.005]
27
4
Muslim,
Child-2
-0.001
[0.007]
-0.446**
[0.007]
0.087**
[0.000]
-0.016
[0.095]
-0.001
[0.006]
0.012
[0.009]
-0.005
[0.009]
-0.016
[0.011]
5
Hindu,
Child-3
0.010**
[0.003]
-0.297**
[0.004]
0.049**
[0.000]
0.026
[0.052]
-0.003
[0.003]
0.016**
[0.004]
0.011*
[0.005]
0.009
[0.006]
0.005
[0.004]
-0.013*
[0.006]
6
7
8
Muslim,
Hindu,
Muslim,
Child-3 Child>=4 Child>=4
0.004
0.023**
0.009
[0.007]
[0.003]
[0.006]
-0.248** -0.268** -0.145**
[0.008]
[0.003]
[0.005]
0.041*
0.044**
0.026**
[0.000]
[0.000]
[0.000]
0.009
0.03
-0.074
[0.117]
[0.063]
[0.121]
-0.002
-0.004
0.002
[0.007]
[0.004]
[0.007]
0.022*
0.019**
0.032**
[0.009]
[0.004]
[0.007]
0.019**
[0.005]
0.013*
[0.006]
0.012
0.011*
0.017*
[0.011]
[0.004]
[0.008]
-0.022*
-0.014* -0.043**
[0.011]
[0.006]
[0.008]
ma complete primary
ma incomplete secondary
ma sec or higher
pa incomplete primary
pa complete primary
pa incomplete secondary
pa complete secondary
pa higher than secondary
assam
bihar
gujarat
haryana
himachal pradesh
karnataka
-0.028**
[0.005]
-0.025**
[0.004]
-0.050**
[0.005]
0.001
[0.005]
-0.001
[0.005]
-0.017**
[0.004]
-0.025**
[0.004]
-0.031**
[0.005]
-0.032**
[0.008]
0.005
[0.008]
0.030**
[0.010]
0.002
[0.009]
-0.025**
[0.008]
0.007
-0.007
[0.014]
-0.029**
[0.011]
-0.035*
[0.014]
-0.017
[0.010]
-0.025*
[0.011]
-0.023*
[0.009]
-0.025*
[0.011]
-0.030*
[0.012]
-0.003
[0.024]
0.022
[0.027]
0.049
[0.040]
0.046
[0.047]
-0.021
[0.035]
0.007
-0.013*
[0.005]
-0.025**
[0.004]
-0.030**
[0.005]
-0.002
[0.005]
-0.013**
[0.005]
-0.022**
[0.004]
-0.033**
[0.004]
-0.028**
[0.005]
-0.015
[0.009]
0.011
[0.008]
0.035**
[0.010]
0.011
[0.010]
-0.020*
[0.008]
-0.009
28
-0.018
[0.014]
-0.002
[0.013]
0.004
[0.020]
-0.008
[0.011]
-0.014
[0.012]
-0.015
[0.010]
-0.027*
[0.011]
-0.01
[0.014]
0.011
[0.026]
0.032
[0.029]
0.052
[0.041]
0.044
[0.048]
-0.034
[0.032]
0.003
-0.016**
[0.006]
-0.018**
[0.006]
-0.033**
[0.007]
0.004
[0.006]
-0.008
[0.006]
-0.017**
[0.004]
-0.028**
[0.005]
-0.032**
[0.006]
-0.022*
[0.011]
0.011
[0.010]
0.02
[0.012]
0
[0.012]
-0.029**
[0.010]
0.01
-0.027*
[0.013]
-0.038**
[0.011]
-0.016
[0.019]
0.006
[0.012]
-0.004
[0.014]
-0.004
[0.011]
-0.002
[0.015]
-0.01
[0.016]
0.019
[0.032]
0.018
[0.031]
0.07
[0.053]
-0.024
[0.031]
-0.015
[0.042]
0.032
-0.015*
[0.007]
-0.020**
[0.007]
-0.034**
[0.011]
-0.004
[0.005]
-0.017**
[0.005]
-0.017**
[0.004]
-0.034**
[0.005]
-0.049**
[0.006]
-0.034**
[0.011]
0.020*
[0.010]
0.011
[0.012]
-0.002
[0.012]
-0.042**
[0.011]
-0.003
-0.016
[0.012]
-0.016
[0.013]
-0.053**
[0.017]
-0.029**
[0.008]
-0.020*
[0.009]
-0.035**
[0.007]
-0.035**
[0.009]
-0.02
[0.013]
0.102*
[0.042]
0.081*
[0.037]
0.160**
[0.062]
0.076
[0.050]
-0.036
[0.048]
0.075
kerala
madhya pradesh
maharashtra
orissa
punjab
rajasthan
tamil nadu
west bengal
uttar pradesh
new delhi
north-east
F tests:
(state dummies)
(child's birth year quadratic)
[0.008]
-0.041**
[0.010]
0.076**
[0.010]
0.01
[0.009]
0.033**
[0.009]
-0.012
[0.012]
0.048**
[0.009]
0
[0.008]
0.001
[0.009]
0.060**
[0.009]
0.033**
[0.012]
-0.001
[0.009]
[0.027]
-0.006
[0.025]
0.051
[0.036]
-0.03
[0.021]
0.134*
[0.065]
0.04
[0.056]
0.076*
[0.037]
0.008
[0.033]
0.032
[0.029]
0.084*
[0.034]
-0.023
[0.028]
0.071
[0.041]
[0.008]
-0.036**
[0.010]
0.048**
[0.009]
-0.004
[0.008]
0.036**
[0.010]
0.018
[0.014]
0.046**
[0.009]
0.005
[0.009]
-0.009
[0.009]
0.054**
[0.009]
0.006
[0.011]
-0.004
[0.009]
[0.027]
-0.033
[0.019]
0.036
[0.034]
0.003
[0.026]
0.066
[0.058]
-0.023
[0.040]
0.055
[0.035]
0.03
[0.038]
0.006
[0.026]
0.04
[0.029]
0.004
[0.034]
0.053
[0.040]
[0.011]
-0.033*
[0.015]
0.055**
[0.011]
-0.024*
[0.009]
0.035**
[0.012]
0.008
[0.016]
0.044**
[0.011]
0.003
[0.011]
-0.004
[0.011]
0.066**
[0.012]
-0.001
[0.014]
0.002
[0.012]
[0.037]
-0.033
[0.021]
0.049
[0.041]
0.015
[0.033]
0.132
[0.080]
-0.037
[0.034]
0.07
[0.043]
0.011
[0.040]
0.028
[0.034]
0.051
[0.036]
0.024
[0.043]
0.031
[0.042]
[0.011]
-0.062**
[0.019]
0.050**
[0.011]
-0.014
[0.011]
0.028*
[0.012]
-0.017
[0.016]
0.034**
[0.011]
-0.008
[0.012]
-0.001
[0.012]
0.070**
[0.011]
-0.004
[0.015]
-0.019
[0.011]
[0.042]
0.082
[0.044]
0.180**
[0.053]
0.089
[0.046]
0.133*
[0.067]
0.077
[0.068]
0.133**
[0.045]
0.112
[0.060]
0.075
[0.039]
0.137**
[0.040]
0.120*
[0.056]
0.189**
[0.059]
410.97
(0.000)
286.18
78.23
(0.000)
27.68
265.61
(0.000)
217.37
34.33
-(0.008)
46.68
238.65
(0.000)
191.31
33.80
-(0.009)
13.22
323.22
(0.000)
222.40
62.94
(0.000)
42.83
29
(maternal age quadratic)
(low castes)
(ma education)
(pa education)
rate of decline p.a.
turning point for age-ma
mean of under5
Observations
(0.000)
249.58
(0.000)
12.60
-(0.006)
99.77
(0.000)
62.75
(0.000)
-1.733
24.402
0.119
53792
(0.000)
19.28
(0.000)
0.58
-(0.447)
8.03
-(0.091)
10.83
-(0.055)
-0.030
23.420
0.100
6830
(0.000)
334.15
(0.000)
24.14
(0.000)
37.93
(0.000)
73.51
(0.000)
-3.435
27.260
0.104
45298
(0.000)
57.03
(0.000)
0.27
-(0.603)
3.26
-(0.515)
5.54
-(0.353)
-1.736
25.516
0.286
5810
(0.000)
181.22
(0.000)
5.83
-(0.120)
22.51
(0.000)
46.99
(0.000)
-5.133
30.065
0.309
32204
P-values in brackets; Standard errors in square brackets; * significant at 5%; ** significant at 1%
30
-(0.001)
26.50
(0.000)
1.35
-(0.245)
10.60
-(0.031)
0.99
-(0.964)
-3.430
30.259
0.272
4548
(0.000)
93.68
(0.000)
16.40
-(0.001)
17.99
-(0.001)
77.51
(0.000)
-6.848
30.533
0.337
44451
(0.000)
8.02
-(0.018)
4.76
-(0.029)
23.19
(0.000)
28.65
(0.000)
3.365
28.248
0.098
9317
Table 6
Infant Mortality
Probits by Religion and Birth-Order
1 if female
age ma at birth of child x10
square of age ma x100
child's birth year x10
square of birth year x100
1 if currently rural resident
scheduled caste
scheduled tribe
other backward caste
1
Hindu,
Child-1
-0.017**
[0.002]
-0.262**
[0.002]
0.054**
[0.000]
-0.013
[0.028]
0
[0.002]
0.010**
[0.003]
0.007*
[0.004]
-0.004
[0.004]
0.006
[0.003]
2
Muslim,
Child-1
-0.024**
[0.006]
-0.200**
[0.007]
0.043*
[0.000]
0.029
[0.073]
-0.003
[0.004]
0.021**
[0.007]
0.005
[0.008]
3
Hindu,
Child-2
-0.005*
[0.002]
-0.276**
[0.002]
0.052**
[0.000]
-0.032
[0.032]
0.001
[0.002]
0.010**
[0.003]
0.010**
[0.004]
0
[0.004]
0.008*
[0.003]
31
4
Muslim,
Child-2
-0.014*
[0.006]
-0.381**
[0.006]
0.075**
[0.000]
-0.067
[0.080]
0.003
[0.005]
0.006
[0.007]
-0.009
[0.008]
5
Hindu,
Child-3
-0.002
[0.003]
-0.205**
[0.003]
0.034**
[0.000]
-0.058
[0.043]
0.002
[0.003]
0.010**
[0.004]
-0.003
[0.004]
-0.003
[0.005]
0.001
[0.004]
6
7
8
Muslim,
Hindu,
Muslim,
Child-3 Child>=4 Child>=4
0.001
0.005
0.002
[0.006]
[0.003]
[0.005]
-0.189** -0.184** -0.090*
[0.006]
[0.003]
[0.005]
0.034**
0.030**
0.016*
[0.000]
[0.000]
[0.000]
-0.02
-0.011
-0.175
[0.093]
[0.053]
[0.100]
0
-0.001
0.009
[0.005]
[0.003]
[0.006]
0.015*
0.009*
0.029**
[0.007]
[0.004]
[0.006]
0.010*
[0.004]
0.003
[0.005]
0.008
0.006
0.014*
[0.009]
[0.004]
[0.007]
ma incomplete primary
ma complete primary
ma incomplete secondary
ma sec or higher
pa incomplete primary
pa complete primary
pa incomplete secondary
pa complete secondary
pa higher than secondary
assam
bihar
gujarat
haryana
himachal pradesh
-0.003
[0.004]
-0.020**
[0.004]
-0.016**
[0.004]
-0.039**
[0.004]
0.003
[0.004]
0.006
[0.005]
-0.003
[0.003]
-0.013**
[0.004]
-0.015**
[0.004]
-0.033**
[0.007]
-0.007
[0.006]
0.017*
[0.008]
-0.008
[0.008]
-0.021**
0
[0.010]
0.006
[0.014]
-0.009
[0.011]
-0.018
[0.014]
-0.029**
[0.008]
-0.020*
[0.010]
-0.021**
[0.008]
-0.019
[0.010]
-0.026*
[0.011]
-0.003
[0.021]
0.011
[0.023]
0.028
[0.033]
0.013
[0.036]
-0.041
0.002
[0.004]
-0.008
[0.004]
-0.016**
[0.004]
-0.017**
[0.005]
0
[0.004]
-0.009*
[0.004]
-0.013**
[0.003]
-0.016**
[0.004]
-0.016**
[0.004]
-0.004
[0.008]
0
[0.006]
0.028**
[0.009]
0.009
[0.009]
-0.023**
32
-0.017*
[0.009]
-0.015
[0.011]
0
[0.011]
-0.008
[0.014]
-0.001
[0.009]
-0.006
[0.011]
-0.003
[0.009]
-0.013
[0.010]
0.013
[0.015]
0.014
[0.023]
0.015
[0.023]
0.049
[0.037]
0.057
[0.048]
-0.014
-0.009*
[0.005]
-0.009
[0.005]
-0.011*
[0.005]
-0.018**
[0.006]
0.005
[0.005]
0
[0.005]
-0.005
[0.004]
-0.018**
[0.004]
-0.015**
[0.005]
-0.019*
[0.008]
-0.008
[0.007]
0.002
[0.009]
-0.016*
[0.008]
-0.029**
-0.013
[0.009]
-0.017
[0.010]
-0.023*
[0.009]
-0.01
[0.014]
0.002
[0.010]
0.006
[0.013]
0.002
[0.009]
0.006
[0.013]
0.014
[0.016]
0.023
[0.031]
0.015
[0.027]
0.071
[0.053]
-0.008
[0.031]
0.018
-0.004
[0.005]
-0.012*
[0.006]
-0.007
[0.006]
-0.020*
[0.009]
-0.003
[0.004]
-0.008
[0.004]
-0.009*
[0.004]
-0.017**
[0.005]
-0.026**
[0.005]
-0.027**
[0.008]
-0.004
[0.008]
-0.004
[0.009]
-0.01
[0.009]
-0.029**
-0.035**
[0.006]
-0.005
[0.011]
-0.012
[0.011]
-0.026
[0.016]
-0.015*
[0.007]
-0.005
[0.009]
-0.019**
[0.006]
-0.024**
[0.008]
-0.008
[0.011]
0.065
[0.034]
0.038
[0.027]
0.110*
[0.052]
0.027
[0.035]
-0.021
karnataka
kerala
madhya pradesh
maharashtra
orissa
punjab
rajasthan
tamil nadu
west bengal
uttar pradesh
new delhi
north-east
F tests:
(state dummies)
[0.007]
-0.003
[0.007]
-0.035**
[0.008]
0.044**
[0.008]
-0.003
[0.007]
0.023**
[0.008]
-0.015
[0.010]
0.025**
[0.008]
-0.008
[0.007]
-0.001
[0.008]
0.039**
[0.008]
0.011
[0.010]
-0.012
[0.007]
[0.022]
-0.003
[0.023]
-0.01
[0.020]
0.051
[0.033]
-0.031
[0.017]
0.126*
[0.062]
0.047
[0.053]
0.046
[0.030]
-0.006
[0.026]
0.017
[0.025]
0.055
[0.029]
-0.011
[0.026]
0.043
[0.035]
[0.007]
-0.01
[0.007]
-0.027**
[0.008]
0.027**
[0.008]
-0.006
[0.007]
0.034**
[0.009]
0.012
[0.012]
0.032**
[0.008]
0.002
[0.007]
-0.006
[0.007]
0.039**
[0.008]
0.006
[0.010]
-0.005
[0.008]
[0.031]
0.003
[0.022]
-0.02
[0.017]
0.021
[0.028]
0.002
[0.022]
0.064
[0.053]
0.002
[0.042]
0.019
[0.026]
0.015
[0.031]
-0.003
[0.020]
0.026
[0.024]
0.003
[0.028]
0.036
[0.034]
[0.007]
-0.006
[0.008]
-0.038**
[0.009]
0.013
[0.008]
-0.026**
[0.006]
0.019*
[0.009]
-0.001
[0.012]
0.016
[0.008]
-0.009
[0.008]
-0.005
[0.009]
0.026**
[0.008]
-0.012
[0.010]
-0.009
[0.008]
[0.048]
0.058
[0.043]
-0.024
[0.017]
0.051
[0.041]
0.025
[0.033]
0.149
[0.086]
-0.008
[0.039]
0.049
[0.039]
-0.02
[0.022]
0.032
[0.033]
0.045
[0.034]
0.044
[0.046]
0.023
[0.037]
[0.009]
-0.017*
[0.008]
-0.059**
[0.012]
0.013
[0.008]
-0.019*
[0.009]
0.011
[0.009]
-0.006
[0.013]
0.012
[0.008]
-0.014
[0.009]
-0.001
[0.010]
0.036**
[0.009]
-0.009
[0.011]
-0.018*
[0.008]
[0.039]
0.027
[0.029]
0.044
[0.034]
0.098*
[0.041]
0.035
[0.033]
0.058
[0.051]
0.054
[0.057]
0.064
[0.033]
0.075
[0.049]
0.042
[0.030]
0.079*
[0.031]
0.044
[0.040]
0.134**
[0.051]
291.14
(0.000)
62.77
(0.000)
209.98
(0.000)
21.9
(0.189)
137.94
(0.000)
31.91
(0.015)
186.99
(0.000)
47.91
(0.000)
33
(child's birth year quadratic)
(maternal age quadratic)
(low castes)
(ma education)
(pa education)
rate of decline p.a.
turning point for age-ma
mean of infant
Observations
126.3
(0.000)
196.66
(0.000)
9.16
(0.027)
71.65
(0.000)
22.52
(0.000)
-0.013
24.165
0.093
53792
14.07
(0.001)
14.82
(0.001)
0.3
(0.585)
2.32
(0.678)
15.52
(0.008)
-5.13
23.339
0.079
6830
89.86
(0.000)
244.96
(0.000)
11.8
(0.008)
22.14
(0.000)
29.82
(0.000)
1.688
26.746
0.075
45298
16.68
(0.000)
58.26
(0.000)
1.1
(0.295)
4.83
(0.305)
3.53
(0.619)
5.092
25.428
0.064
5810
94.4
(0.000)
122.58
(0.000)
1.29
(0.733)
11.22
(0.024)
22.14
(0.000)
3.381
29.964
0.071
32204
P-values in brackets; Standard errors in square brackets; * significant at 5%; ** significant at 1%
34
11.17
(0.004)
14.59
(0.001)
0.92
(0.339)
6.08
(0.194)
1.15
(0.949)
-0.02
27.711
0.052
4548
89.86
(0.000)
61.58
(0.000)
6.55
(0.088)
7.98
(0.092)
28.62
(0.000)
-1.731
30.482
0.088
44451
27.48
(0.000)
4.42
(0.110)
4.58
(0.032)
19.39
(0.001)
13.14
(0.022)
15.301
28.234
0.069
9317
Table 7
Neonatal Mortality
Probits by Religion and Birth-Order
1 if female
age ma at birth of child x10
square of age ma x100
child's birth year x10
square of birth year x100
1 if currently rural resident
scheduled caste
scheduled tribe
other backward caste
ma incomplete primary
1
Hindu,
Child-1
-0.016**
[0.002]
-0.191**
[0.002]
0.040**
[0.000]
0.003
[0.023]
-0.001
[0.001]
0.006*
[0.003]
0.003
[0.003]
-0.001
[0.004]
0.004
[0.003]
-0.001
[0.004]
2
Muslim,
Child-1
-0.020**
[0.005]
-0.133*
[0.005]
0.031*
[0.000]
-0.043
[0.058]
0.002
[0.003]
0.021**
[0.006]
0.004
[0.007]
-0.002
[0.008]
3
Hindu,
Child-2
-0.010**
[0.002]
-0.187**
[0.002]
0.036**
[0.000]
-0.033
[0.024]
0.002
[0.001]
0.006*
[0.002]
0.005
[0.003]
0.001
[0.003]
0.007**
[0.003]
0.002
[0.003]
35
4
Muslim,
Child-2
-0.007
[0.005]
-0.227**
[0.004]
0.045**
[0.000]
-0.049
[0.062]
0.002
[0.004]
0.015**
[0.006]
0.002
[0.007]
-0.008
[0.007]
5
Hindu,
Child-3
-0.007**
[0.002]
-0.159**
[0.002]
0.027**
[0.000]
-0.044
[0.032]
0.002
[0.002]
0.006*
[0.003]
-0.001
[0.003]
-0.001
[0.004]
0.004
[0.003]
-0.009**
[0.003]
6
7
8
Muslim,
Hindu,
Muslim,
Child-3 Child>=4 Child>=4
-0.004
-0.002
-0.005
[0.005]
[0.002]
[0.004]
-0.092** -0.117** -0.070*
[0.006]
[0.002]
[0.003]
0.014**
0.019**
0.013*
[0.000]
[0.000]
[0.000]
-0.045
-0.008
-0.191*
[0.076]
[0.041]
[0.075]
0.002
0
0.010*
[0.004]
[0.002]
[0.004]
0.017**
0.007*
0.017**
[0.006]
[0.003]
[0.004]
0.008*
[0.003]
0.004
[0.004]
0.002
0.009**
0.007
[0.007]
[0.003]
[0.005]
-0.005
-0.001
-0.021**
[0.008]
[0.004]
[0.005]
ma complete primary
ma incomplete secondary
ma sec or higher
pa incomplete primary
pa complete primary
pa incomplete secondary
pa complete secondary
pa higher than secondary
assam
bihar
gujarat
haryana
himachal pradesh
karnataka
-0.010**
[0.004]
-0.006
[0.003]
-0.024**
[0.004]
0.003
[0.004]
0.004
[0.004]
0
[0.003]
-0.005
[0.004]
-0.007
[0.004]
-0.024**
[0.005]
-0.002
[0.005]
0.012
[0.007]
-0.009
[0.006]
-0.021**
[0.005]
0
0.007
[0.011]
-0.007
[0.009]
-0.007
[0.012]
-0.019**
[0.006]
-0.01
[0.008]
-0.012
[0.006]
-0.011
[0.008]
-0.016
[0.009]
-0.02
[0.012]
-0.001
[0.016]
0.007
[0.022]
-0.016
[0.018]
-0.035**
[0.011]
-0.01
-0.005
[0.003]
-0.007*
[0.003]
-0.009*
[0.004]
0.003
[0.003]
-0.002
[0.003]
-0.005
[0.003]
-0.007*
[0.003]
-0.006
[0.003]
-0.011*
[0.005]
-0.003
[0.005]
0.009
[0.006]
-0.002
[0.006]
-0.021**
[0.004]
-0.011**
36
-0.014
[0.008]
-0.007
[0.008]
-0.015
[0.009]
0.003
[0.008]
-0.005
[0.009]
-0.001
[0.007]
-0.004
[0.009]
0.027
[0.014]
-0.002
[0.015]
-0.006
[0.014]
0.054
[0.036]
0.021
[0.032]
-0.016
[0.020]
0.004
-0.007
[0.004]
-0.004
[0.004]
-0.009
[0.005]
0.005
[0.004]
-0.004
[0.004]
-0.002
[0.003]
-0.009**
[0.004]
-0.003
[0.004]
-0.01
[0.006]
-0.003
[0.005]
0.004
[0.007]
-0.008
[0.006]
-0.022**
[0.005]
-0.006
-0.007
[0.010]
-0.009
[0.008]
0.012
[0.017]
0
[0.008]
0.013
[0.012]
-0.001
[0.007]
-0.009
[0.009]
0.007
[0.013]
-0.004
[0.016]
0
[0.017]
0.028
[0.033]
0.025
-0.007
[0.004]
-0.001
[0.005]
-0.014*
[0.007]
-0.004
[0.003]
-0.003
[0.003]
-0.004
[0.003]
-0.006
[0.004]
-0.013**
[0.004]
-0.021**
[0.006]
-0.005
[0.006]
-0.01
[0.006]
-0.016**
[0.006]
-0.020**
[0.007]
-0.008
0
[0.009]
-0.004
[0.009]
-0.017
[0.012]
-0.004
[0.006]
0.005
[0.007]
-0.010*
[0.005]
-0.016**
[0.006]
0.004
[0.010]
0.033
[0.024]
0.019
[0.020]
0.061
[0.040]
0.001
[0.022]
-0.004
[0.033]
0.017
kerala
madhya pradesh
maharashtra
orissa
punjab
rajasthan
tamil nadu
west bengal
uttar pradesh
new delhi
north-east
F tests:
(state dummies)
(child's birth year quadratic)
[0.006]
-0.022**
[0.007]
0.030**
[0.007]
-0.005
[0.006]
0.013*
[0.007]
-0.01
[0.008]
0.014*
[0.006]
-0.004
[0.006]
0
[0.006]
0.026**
[0.007]
0.003
[0.008]
-0.01
[0.006]
[0.015]
-0.015
[0.013]
0.032
[0.025]
-0.026*
[0.011]
0.035
[0.039]
0.029
[0.040]
0
[0.017]
-0.016
[0.016]
0.001
[0.016]
0.017
[0.019]
-0.026
[0.014]
0.023
[0.025]
[0.004]
-0.016**
[0.006]
0.012*
[0.006]
-0.005
[0.005]
0.013*
[0.006]
-0.006
[0.008]
0.011
[0.006]
-0.002
[0.005]
-0.010*
[0.005]
0.017**
[0.006]
-0.013*
[0.006]
-0.01
[0.005]
[0.018]
-0.011
[0.013]
0.009
[0.021]
0.017
[0.022]
-0.017
[0.019]
0.016
[0.039]
0.009
[0.019]
0.004
[0.022]
-0.008
[0.014]
0.017
[0.019]
0.008
[0.025]
-0.002
[0.019]
[0.006]
-0.016*
[0.008]
0.006
[0.006]
-0.012*
[0.005]
0.014
[0.007]
-0.004
[0.009]
0.005
[0.006]
-0.004
[0.006]
-0.006
[0.006]
0.013*
[0.006]
-0.005
[0.008]
-0.005
[0.006]
[0.027]
-0.024**
[0.008]
0.018
[0.025]
0.012
[0.022]
0.021
[0.040]
-0.002
[0.031]
0.012
[0.022]
201.78
(0.000)
75.44
(0.000)
45.43
(0.000)
9.96
-(0.007)
135.09
(0.000)
41.38
(0.000)
22.24
-(0.176)
7.52
-(0.023)
68.84
(0.000)
60.10
(0.000)
37
[0.022]
0.011
[0.021]
0.043
[0.028]
0.017
[0.023]
0
[0.027]
0.007
[0.019]
0.005
[0.018]
0.006
[0.025]
0
[0.020]
[0.006]
-0.041**
[0.007]
0.005
[0.006]
-0.012
[0.007]
0.002
[0.007]
-0.005
[0.010]
0.011
[0.007]
-0.007
[0.007]
-0.003
[0.008]
0.023**
[0.007]
-0.01
[0.008]
-0.008
[0.007]
15.89
-(0.320)
7.98
-(0.019)
146.93
(0.000)
43.95
(0.000)
26.95
-(0.042)
30.78
(0.000)
0.036
[0.025]
0.033
[0.034]
0.017
[0.020]
0.035
[0.022]
0.017
[0.027]
0.081*
[0.040]
(maternal age quadratic)
(low castes)
(ma education)
(pa education)
rate of decline p.a.
turning point for age-ma
mean of neonatal
observations
138.56
(0.000)
2.69
-(0.441)
37.47
(0.000)
8.11
(0.150)
-1.717
23.746
0.064
53792
6.64
(0.036)
0.31
-(0.579)
1.45
(0.836)
8.95
(0.111)
3.396
21.060
0.053
6830
167.94
(0.000)
9.79
-(0.020)
10.28
(0.036)
10.18
(0.070)
3.406
26.183
0.045
45298
33.10
(0.000)
0.07
-(0.791)
3.84
(0.428)
7.95
(0.159)
3.390
25.184
0.040
5810
106.30
(0.000)
3.39
-(0.335)
9.14
(0.058)
11.08
(0.050)
3.395
29.065
0.042
32204
P-values in brackets; Standard errors in square brackets; * significant at 5%; ** significant at 1%
38
11.33
(0.004)
0.07
-(0.792)
2.62
(0.623)
3.58
(0.611)
3.394
31.956
0.033
4303
44.36
(0.000)
10.86
(0.013)
5.13
(0.274)
10.12
(0.072)
-0.008
30.747
0.053
44451
5.46
(0.065)
1.72
(0.190)
11.84
(0.019)
9.79
(0.081)
17.004
27.579
0.041
9248
Fig. 1
0.00
0.25
Risk of death
0.50
0.75
1.00
Kaplan-Meier survival estimates
0
20
40
Age of the Child/ months
Hindu
39
Muslim
60
Fig. 2
40
Fig. 3
41
Fig. 4
0
Frequency density
.02
.04
.06
Year of birth of child religion comparison
60
70
80
Year of birth of child
Hindu
90
100
Muslim
Fig. 5
0
.02
Frequency density
.04
.06
.08
.1
Age ma at birth of index child
10
20
30
40
Age ma at birth of index child (years)
Hindu
42
Muslim
50
Fig. 6
Education of Mother
Hindu
Muslim
5.35%
10.84%
9.93%
12.01%
6.68%
7.21%
12.32%
61.88%
8.97%
64.81%
No education
Complete primary
Complete secondary and higher
Incomplete primary
Incomplete secendary
Fig. 7
Education of Father
Hindu
Muslim
7.81%
9.63%
14.83%
31.58%
39.24%
11.88%
19.85%
21.15%
11.20%
8.68%
9.37%
N o education
C omplete primary
C omplete secondary
14.79%
Incomplete primary
Incomplete secendary
Higher
43
Fig. 8
Proportion of families in Rural areas
Hindu
Muslim
26.23%
39.7%
60.3%
73.77%
Not Rural
Rural
Fig. 9
Proportion of females
Hindu
47.85%
Muslim
47.73%
52.15%
Male
Female
44
52.27%
Appendix Table 1
Mortality differences by Religion & Child Age
State name
andhra pradesh
assam
bihar
gujarat
haryana
himachal pradesh
karnataka
kerala
madhya pradesh
maharashtra
orissa
punjab
rajasthan
tamil nadu
uttar pradesh
west bengal
new delhi
north-east
ALL INDIA
Neo-Diff
Neonatal
Hindu
Muslim
0.059
0.033
0.032
0.042
0.049
0.049
0.052
0.046
0.037
0.031
0.022
0.015
0.047
0.041
0.022
0.024
0.069
0.053
0.040
0.025
0.062
0.036
0.033
0.034
0.060
0.047
0.046
0.028
0.072
0.052
0.044
0.044
0.032
0.025
0.037
0.056
0.052
0.042
0.026
0.010
0.000
0.005
0.007
0.008
0.006
0.002
0.016
0.015
0.026
0.001
0.013
0.017
0.020
0.000
0.007
0.018
0.010
Inf-Diff
Infant
Hindu Muslim
0.089
0.043
0.053
0.070
0.075
0.073
0.082
0.063
0.064
0.066
0.041
0.029
0.073
0.058
0.030
0.038
0.109
0.082
0.061
0.035
0.099
0.100
0.056
0.055
0.096
0.077
0.067
0.050
0.114
0.088
0.071
0.067
0.054
0.043
0.059
0.086
0.083
45
0.068
0.046
0.017
0.003
0.020
0.002
0.012
0.015
0.008
0.027
0.026
0.001
0.001
0.019
0.017
0.026
0.004
0.010
0.027
0.015
Under5Diff
Under5
Hindu Muslim
0.114
0.054
0.068
0.087
0.112
0.104
0.113
0.086
0.088
0.094
0.053
0.036
0.105
0.079
0.039
0.049
0.163
0.111
0.083
0.048
0.128
0.126
0.071
0.067
0.136
0.123
0.091
0.071
0.157
0.122
0.090
0.092
0.070
0.059
0.081
0.114
0.116
0.094
0.060
0.018
0.008
0.027
0.006
0.017
0.027
0.010
0.052
0.035
0.002
0.004
0.013
0.020
0.035
0.002
0.011
0.033
0.022
Appendix Table 2
Distribution of children by birth-order for each religion
Child birth-order % for Hindus % for Muslims
1
30.61
25.77
2
25.77
21.92
3
18.32
17.16
4
11.44
12.61
5
6.59
8.66
6
3.67
5.78
7
1.90
3.61
8
0.94
2.21
9
0.43
1.27
10
0.20
0.60
11
0.07
0.28
12
0.02
0.08
13
0.01
0.02
14
0.00
0.01
15
0.00
0.01
16
0.00
0.00
46
Appendix Table 3
Under-5 Mortality Differences by Religion and Birth Order
State name
andhra pradesh
assam
bihar
gujarat
haryana
himachal pradesh
karnataka
kerala
madhya pradesh
maharashtra
orissa
punjab
rajasthan
tamil nadu
uttar pradesh
west bengal
new delhi
north-east
ALL INDIA
Child-1
Hindu Muslim
0.120
0.065
0.063
0.079
0.116
0.117
0.119
0.095
0.089
0.126
0.056
0.050
0.113
0.087
0.041
0.058
0.194
0.112
0.098
0.034
0.136
0.183
0.061
0.105
0.150
0.148
0.094
0.064
0.094
0.113
0.164
0.163
0.084
0.041
0.088
0.137
0.119
0.100
diff
0.054
0.017
0.000
0.024
0.038
0.006
0.026
0.017
0.081
0.063
0.047
0.044
0.003
0.031
0.020
0.002
0.043
0.049
0.018
Child-2
Hindu Muslim
0.108
0.071
0.067
0.089
0.108
0.109
0.114
0.100
0.083
0.107
0.050
0.030
0.088
0.086
0.034
0.037
0.155
0.100
0.076
0.063
0.124
0.133
0.075
0.038
0.136
0.127
0.086
0.076
0.077
0.090
0.144
0.117
0.058
0.063
0.072
0.112
0.104
0.090
diff
0.037
0.022
0.002
0.014
0.025
0.020
0.003
0.003
0.055
0.013
0.009
0.037
0.010
0.010
0.013
0.027
0.005
0.040
0.014
47
Child-3
Hindu Muslim
0.106
0.049
0.064
0.078
0.101
0.090
0.104
0.094
0.079
0.043
0.050
0.039
0.108
0.085
0.047
0.025
0.154
0.098
0.062
0.050
0.123
0.143
0.072
0.023
0.130
0.121
0.090
0.056
0.088
0.086
0.149
0.107
0.057
0.060
0.079
0.076
0.107
0.080
diff
0.057
0.014
0.010
0.011
0.036
0.011
0.023
0.021
0.056
0.012
0.020
0.049
0.009
0.033
0.002
0.041
0.004
0.003
0.026
Child>=4
Hindu Muslim
0.128
0.032
0.077
0.095
0.128
0.103
0.121
0.112
0.104
0.097
0.065
0.020
0.118
0.070
0.053
0.072
0.161
0.133
0.094
0.051
0.135
0.088
0.085
0.072
0.139
0.114
0.110
0.091
0.114
0.076
0.167
0.128
0.079
0.085
0.084
0.117
0.130
0.098
diff
0.096
0.018
0.025
0.009
0.007
0.045
0.048
0.018
0.028
0.043
0.046
0.013
0.024
0.019
0.038
0.039
0.006
0.033
0.032
Appendix Table 4
Infant Mortality Differences by Religion and Birth Order
State name
andhra pradesh
assam
bihar
gujarat
haryana
himachal pradesh
karnataka
kerala
madhya pradesh
maharashtra
orissa
punjab
rajasthan
tamil nadu
uttar pradesh
west bengal
new delhi
north-east
ALL INDIA
Child-1
Hindu Muslim
0.100
0.056
0.048
0.069
0.087
0.091
0.094
0.071
0.068
0.084
0.050
0.025
0.087
0.066
0.036
0.048
0.145
0.101
0.076
0.025
0.112
0.169
0.050
0.105
0.113
0.107
0.074
0.047
0.079
0.088
0.132
0.126
0.065
0.041
0.067
0.104
0.093
0.079
diff
0.044
0.021
0.005
0.023
0.016
0.025
0.022
0.013
0.044
0.050
0.057
0.056
0.006
0.027
0.009
0.006
0.024
0.037
0.014
Child-2
Hindu Muslim
0.078 0.055
0.055 0.071
0.070 0.068
0.084 0.083
0.062 0.095
0.032 0.030
0.061 0.064
0.026 0.028
0.102 0.070
0.054 0.050
0.097 0.117
0.055 0.038
0.096 0.073
0.063 0.051
0.057 0.058
0.106 0.083
0.045 0.049
0.052 0.084
0.075 0.064
diff
0.024
0.016
0.002
0.001
0.033
0.002
0.003
0.002
0.032
0.005
0.019
0.017
0.024
0.012
0.001
0.023
0.004
0.032
0.011
48
Child-3
Muslim
0.084 0.028
0.047 0.050
0.063 0.052
0.070 0.070
0.049 0.029
0.034 0.039
0.071 0.074
0.026 0.013
0.092 0.071
0.042 0.035
0.091 0.122
0.054 0.023
0.087 0.069
0.060 0.016
0.067 0.058
0.098 0.070
0.040 0.052
0.055 0.042
0.071 0.052
diff
0.056
0.003
0.012
0.000
0.020
0.005
0.003
0.013
0.022
0.007
0.032
0.030
0.018
0.044
0.009
0.028
0.012
0.013
0.019
Child>=4
Hindu Muslim
0.096 0.028
0.056 0.075
0.080 0.072
0.080 0.089
0.073 0.063
0.050 0.020
0.074 0.044
0.027 0.055
0.099 0.088
0.064 0.030
0.094 0.053
0.073 0.058
0.094 0.071
0.075 0.077
0.086 0.058
0.116 0.090
0.060 0.047
0.061 0.096
0.088 0.069
diff
0.067
0.019
0.008
0.009
0.010
0.031
0.030
0.028
0.011
0.034
0.041
0.015
0.023
0.002
0.029
0.026
0.014
0.035
0.019
Appendix Table 5
Neonatal Mortality by Religion & Birth Order
State name
andhra pradesh
assam
bihar
gujarat
haryana
himachal
karnataka
kerala
madhya
pradesh
maharashtra
orissa
punjab
rajasthan
tamil nadu
uttar pradesh
west bengal
new delhi
north-east
ALL INDIA
Child-1
diff
Hindu
Muslim
0.069
0.047
0.032
0.038
0.062
0.070
0.066
0.048
0.044
0.042
0.029
0.013
0.062
0.045
0.028
0.036
0.099
0.080
0.051
0.075
0.035
0.075
0.053
0.056
0.090
0.043
0.044
0.064
0.016
0.085
0.088
0.054
0.030
0.061
0.080
0.018
0.080
0.053
0.022
0.005
0.008
0.019
0.002
0.017
0.017
0.008
0.020
0.034
0.010
0.052
0.021
0.023
0.005
0.010
0.025
0.036
0.011
Child-2
diff
Hindu
Muslim
0.053
0.033
0.029
0.039
0.045
0.036
0.050
0.072
0.036
0.060
0.016
0.015
0.036
0.043
0.020
0.021
0.065
0.043
0.037
0.059
0.027
0.055
0.042
0.032
0.065
0.020
0.030
0.045
0.021
0.009
0.009
0.022
0.024
0.001
0.007
0.001
0.022
0.039
0.017
0.038
0.044
0.030
0.034
0.057
0.035
0.034
0.040
0.002
0.043
0.011
0.011
0.011
0.001
0.008
0.016
0.004
0.006
49
Child-3
diff
Hindu
Muslim
0.052
0.028
0.030
0.030
0.040
0.035
0.045
0.047
0.030
0.000
0.015
0.000
0.040
0.056
0.021
0.008
0.054
0.047
0.028
0.056
0.029
0.047
0.038
0.036
0.057
0.026
0.032
0.042
0.029
0.041
0.023
0.039
0.000
0.043
0.037
0.026
0.028
0.033
0.024
0.001
0.005
0.002
0.030
0.015
0.015
0.013
0.007
0.001
0.016
0.005
0.008
0.038
0.007
0.020
0.000
0.004
0.010
Child>=4
diff
Hindu
Muslim
0.059
0.021
0.029
0.046
0.048
0.044
0.041
0.061
0.034
0.029
0.028
0.020
0.047
0.032
0.011
0.030
0.058
0.050
0.038
0.053
0.044
0.061
0.048
0.049
0.074
0.033
0.040
0.053
0.020
0.018
0.000
0.047
0.049
0.035
0.049
0.028
0.065
0.041
0.038
0.017
0.003
0.020
0.006
0.009
0.015
0.019
0.009
0.018
0.035
0.044
0.014
0.000
0.013
0.024
0.005
0.026
0.012
Appendix Table 6
Means of Independent Variables
State name
andhra pradesh
assam
bihar
gujarat
haryana
himachal
pradesh
karnataka
kerala
madhya
pradesh
maharashtra
orissa
punjab
rajasthan
tamil nadu
uttar pradesh
west bengal
new delhi
north-east
ALL INDIA
Prop female births birth year of child Age ma at birth
Prop rural
SC
ST
OBC
Hindu
Muslim Hindu Muslim Hindu Muslim Hindu Muslim Hindu Muslim Hindu Muslim Hindu Muslim
0.484
0.458 84.881 85.056 20.818 22.849
0.774
0.318
0.184
0.000
0.054
0.000
0.505
0.114
0.478
0.471 86.171 87.815 22.663 22.289
0.731
0.936
0.169
0.000
0.267
0.000
0.174
0.007
0.478
0.478 86.851 87.374 22.778 23.593
0.912
0.869
0.251
0.000
0.074
0.000
0.531
0.575
0.480
0.478 85.739 86.201 22.516 22.725
0.646
0.268
0.176
0.000
0.226
0.000
0.251
0.264
0.458
0.493 86.275 88.221 22.884 24.351
0.739
0.888
0.236
0.000
0.000
0.000
0.222
0.389
0.479
0.422 86.029 87.847 22.938 22.671
0.771
0.644
0.241
0.000
0.004
0.000
0.186
0.197
0.491
0.495
0.478
0.468
0.471
0.493
85.457
84.858
86.654
86.565
85.340
86.293
21.303
23.364
22.116
21.695
22.474
22.757
0.728
0.673
0.793
0.502
0.814
0.302
0.217
0.168
0.177
0.000
0.000
0.000
0.077
0.019
0.232
0.000
0.000
0.000
0.474
0.518
0.407
0.011
0.459
0.539
0.479
0.483
0.460
0.476
0.485
0.488
0.471
0.470
0.488
0.486
0.478
0.365
0.475
0.496
0.477
0.486
0.470
0.494
85.920
86.451
86.195
86.955
85.581
86.885
85.344
86.658
87.206
87.642
87.042
86.092
87.165
86.047
87.482
86.519
88.120
89.076
21.526
22.377
23.524
22.962
22.165
21.790
23.061
23.224
23.536
21.923
23.570
24.081
23.883
22.477
22.334
23.597
22.988
22.685
0.532
0.817
0.526
0.820
0.619
0.851
0.540
0.078
0.799
0.125
0.385
0.244
0.455
0.257
0.617
0.802
0.136
0.657
0.107
0.215
0.343
0.204
0.272
0.317
0.234
0.261
0.190
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.110
0.187
0.000
0.168
0.010
0.057
0.026
0.008
0.140
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.000
0.244
0.319
0.187
0.231
0.700
0.051
0.317
0.157
0.232
0.045
0.065
0.734
0.258
0.966
0.005
0.182
0.287
0.128
0.479
0.477
85.882
86.558
22.494
22.897
0.738
0.603
0.216
0.000
0.106
0.000
0.335
0.255
50
Appendix Table 6 cont.
Fathers’ education
State name
andhra pradesh
assam
bihar
gujarat
haryana
himachal pradesh
karnataka
kerala
madhya pradesh
maharashtra
orissa
punjab
rajasthan
tamil nadu
uttar pradesh
west bengal
new delhi
north-east
ALL INDIA
No education
Hindu Muslim
0.479
0.304
0.295
0.493
0.438
0.582
0.274
0.171
0.326
0.630
0.164
0.418
0.403
0.410
0.058
0.118
0.354
0.255
0.201
0.190
0.332
0.290
0.223
0.495
0.395
0.480
0.234
0.109
0.271
0.402
0.296
0.492
0.135
0.334
0.244
0.354
0.316
0.392
Incomplete
Complete primary
Incomplete
Complete
Higher
primary
secondary
Secondary
Hindu Muslim Hindu Muslim Hindu Muslim Hindu Muslim Hindu Muslim
0.096
0.078
0.117
0.101
0.126
0.186
0.081
0.152
0.100
0.180
0.183
0.199
0.046
0.034
0.259
0.165
0.086
0.056
0.132
0.053
0.063
0.062
0.055
0.046
0.173
0.141
0.147
0.121
0.124
0.048
0.166
0.178
0.076
0.065
0.235
0.326
0.090
0.138
0.159
0.123
0.041
0.047
0.065
0.094
0.201
0.092
0.217
0.085
0.148
0.052
0.062
0.036
0.099
0.056
0.235
0.241
0.264
0.165
0.175
0.084
0.117
0.158
0.064
0.083
0.175
0.194
0.124
0.101
0.117
0.054
0.163
0.276
0.118
0.129
0.343
0.319
0.197
0.112
0.120
0.046
0.141
0.183
0.137
0.148
0.189
0.191
0.039
0.068
0.140
0.155
0.151
0.176
0.055
0.083
0.271
0.323
0.149
0.138
0.174
0.089
0.156
0.198
0.115
0.137
0.228
0.276
0.086
0.065
0.082
0.034
0.037
0.000
0.097
0.095
0.228
0.171
0.241
0.180
0.174
0.059
0.075
0.099
0.104
0.082
0.194
0.212
0.097
0.068
0.135
0.059
0.143
0.140
0.148
0.203
0.250
0.254
0.101
0.164
0.125
0.130
0.214
0.245
0.056
0.072
0.236
0.169
0.080
0.049
0.142
0.064
0.063
0.086
0.113
0.100
0.206
0.137
0.133
0.096
0.188
0.089
0.029
0.066
0.069
0.097
0.183
0.200
0.198
0.128
0.385
0.176
0.201
0.229
0.076
0.098
0.268
0.218
0.076
0.044
0.136
0.058
0.112
0.148
0.094
51
0.087
0.211
0.199
0.119
0.096
0.148
0.078
Appendix Table 6 cont.
Mother’s education
State name
andhra pradesh
assam
bihar
gujarat
haryana
himachal pradesh
karnataka
kerala
madhya pradesh
maharashtra
orissa
punjab
rajasthan
tamil nadu
uttar pradesh
west bengal
new delhi
north-east
ALL INDIA
No education
Hindu Muslim
0.685
0.508
0.497
0.670
0.797
0.888
0.582
0.429
0.657
0.950
0.402
0.719
0.603
0.635
0.093
0.172
0.728
0.618
0.431
0.370
0.599
0.604
0.415
0.806
0.809
0.865
0.426
0.269
0.464
0.597
0.737
0.827
0.366
0.713
0.482
0.634
0.619
0.648
Incomplete
Complete primary
Incomplete
Secondary or
primary
secondary
Higher
Hindu Muslim Hindu Muslim Hindu Muslim Hindu Muslim
0.080
0.108
0.073
0.072
0.096
0.146
0.065
0.166
0.135
0.166
0.040
0.025
0.223
0.108
0.105
0.032
0.037
0.048
0.028
0.023
0.081
0.025
0.057
0.016
0.105
0.194
0.059
0.083
0.135
0.222
0.120
0.072
0.029
0.007
0.097
0.024
0.085
0.002
0.131
0.017
0.057
0.016
0.199
0.112
0.151
0.092
0.191
0.060
0.100
0.100
0.038
0.104
0.143
0.109
0.116
0.051
0.164
0.294
0.083
0.104
0.313
0.318
0.348
0.112
0.093
0.133
0.067
0.110
0.061
0.069
0.051
0.070
0.161
0.176
0.043
0.101
0.221
0.231
0.144
0.123
0.130
0.147
0.088
0.092
0.131
0.140
0.053
0.017
0.033
0.018
0.159
0.063
0.123
0.041
0.271
0.072
0.038
0.047
0.053
0.039
0.059
0.037
0.041
0.012
0.146
0.173
0.125
0.214
0.193
0.234
0.110
0.110
0.191
0.229
0.057
0.044
0.178
0.105
0.110
0.026
0.047
0.031
0.078
0.067
0.066
0.035
0.072
0.040
0.044
0.030
0.100
0.076
0.171
0.087
0.319
0.094
0.190
0.195
0.060
0.052
0.179
0.104
0.089
0.015
0.090
0.123
0.072
52
0.067
0.120
0.108
0.099
0.053