ISSN (Print) - 0012-9976 | ISSN (Online) - 2349-8846

A+| A| A-

Religious Differentials in Fertility in India: Is There a Convergence?

Analysis of the data from the three rounds of India's National Family Health Survey shows that fertility transition is continuing in all the major religious groups of India. Substantial declines have occurred in fertility over the period encompassing the three rounds. The use of contraceptives has become more widely prevalent with a majority of couples wanting to stop childbearing at two or three children. Spatial variation is also noticeable with religious differentials being very small or negligible in some states. At the national level, religious differentials are narrowing though it is difficult to say when a convergence could occur. However, fertility for all religious groups is expected to fall further towards the replacement level and possibly to below this level.


Religious Differentials in Fertility in India: Is There a Convergence?

Manoj Alagarajan, P M Kulkarni

Analysis of the data from the three rounds of India’s National Family Health Survey shows that fertility transition is continuing in all the major religious groups of India. Substantial declines have occurred in fertility over the period encompassing the three rounds. The use of contraceptives has become more widely prevalent with a majority of couples wanting to stop childbearing at two or three children. Spatial variation is also noticeable with religious differentials being very small or negligible in some states. At the national level, religious differentials are narrowing though it is difficult to say when a convergence could occur. However, fertility for all religious groups is expected to fall further towards the replacement level and possibly to below this level.

An earlier draft of this paper was discussed in a workshop at the Centre for Development Studies, Thiruvananthapuram. Detailed comments from U S Mishra and suggestions from K Srinivasan, Leela Visaria, Narayanan Nair, K S James, and S Irudaya Rajan were valuable. We are indebted to the various organisations that were engaged in the NFHS, particularly to the International Institute for Population Sciences, Mumbai, and Macro International, for facilitating access to the data.

Manoj Alagarajan ( is at the International Institute for Population Sciences, Mumbai and P M Kulkarni ( is with the Centre for the Study of Regional Development, School of Social Sciences, Jawaharlal Nehru University, New Delhi.

ne of the remarkable changes to have occurred in Indian society in the last generation is a shift from largely u nregulated fertility to fairly regulated fertility within marriages. Both registration (Sample Registration System) and survey (National Family Health Survey-3) data show that the t otal fertility rate (TFR), or the number of births a woman is e xpected to have in a lifetime, has fallen to well below three (India, Registrar General 2007; IIPS and Macro International 2007). This change has occurred because many couples in the reproductive ages exercise the choice of regulation, with nearly half using some method of modern contraception. Fertility in India can now be said to be within the “calculus of conscious choice” and the country is well into fertility transition.

But the timing and pace of the shift from unregulated to regulated fertility has not been the same throughout the country. R egional variations have been conspicuous; the southern and some western states have led the move and progressed at a rapid rate while the north-central states have been late and slow. There are notable variations across social and economic classes as well. While differentials by education and place of residence are well recognised and discussed, those by religion are debated rather contentiously. As religious affiliation is primarily ascribed than acquired, the differentials assume political overtones in multireligious societies. Social scientists have studied the role of religion in fertility behaviour quite extensively and analyses of the empirical evidence on Christian-Jewish, Catholic-Protestant, Christian-Muslim, Hindu-Muslim, and Buddhist-Muslim differences in various populations have led to a number of competing explanations. In addition to religion per se as a factor, with precepts and injunctions influencing fertility behaviour (called the particularised theology hypothesis), differences in other socioeconomic conditions (characteristics hypothesis), being a minority (minority population hypothesis), and the effects of interactions (interaction hypothesis) have been put forward (Freedman and Whelpton 1961; Goldscheider and Uhlenberg 1969; Chamie 1977; Day 1984; for an excellent review, see McQuillan 2004).

Religious differentials have long been noted in India (Visaria 1974; Balasubramanian 1984; Das and Pandey 1985) with religious influences, differences in educational level, and minority status cited as possible reasons. When data from large surveys, especially the National Family Health Surveys (NFHS), became available, it was possible to examine the various hypotheses systematically. The NFHS is a nationwide household survey covering a wide range of issues on health, in particular, maternal and child health, fertility, contraceptive knowledge and use, and desired

november 29, 2008

family size. Detailed fertility history is collected from all women Muslims, Hindus, Christians and Sikhs, respectively (IIPS and of childbearing ages in the sample. Adequate information on Macro International 2007). For various reasons, it is the higher their socio-economic background is obtained and this enables than average fertility among Muslims that attracts the most the kind of analysis required to study differentials and answer a ttention. But fertility has fallen substantially among Muslims too; questions on relationships. Three rounds of the survey have been over the 13 years between the NFHS-1 and the NFHS-3 (1992-93 to carried out, the NFHS-1 during 1992-93, the NFHS-2 during 1998-2005-06), the TFR for Muslims declined steeply from 4.41 to 3.09 99, and the NFHS-3 during 2005-06. The sample size of women of (IIPS 1995; IIPS and Macro International 2007). Fertility among childbearing ages was 89,777 in the NFHS-1, 90,303 in the NFHS-2, Muslims is at a level that could at most be called “moderate”, and 124,385 (93,089 of whom were married) in the NFHS-3 (IIPS c ertainly not “high”. In states that show low fertility for other and Macro International 2007). The surveys were conducted by r eligions, Muslim fertility has also fallen. The differences vary the International Institute for Population Sciences (IIPS), Mumbai across regions; while the gap is as wide as about one child in some and Macro International. They involved a number of research in-states, it is negligible in some others. The desired family size stitutions and professional survey o rganisations, and were sup-among Muslims is not very high and many have been regulating ported by the government of India and international agencies. fertility and many intend to do so in the future. According to The survey reports give details of the the NFHS-3, 74% of married Muslim

Figure 1: Trends in TFR by Religion, NFHS-1 to 3

methodology, including sampling de-5

Muslim Sikh Christian Hindu
w omen aged 15 to 49 with two children sign and sampling errors. The data are

want no more offspring, compared to in the public domain and are well doc-4

83% of women overall (IIPS and Macro umented, making it possible to subject

International 2007). Contrary to the these to r igorous statistical analysis. 3

view often heard in public debates, Extensive analyses have been car

there is indisputable evidence that a ried out on data from the NFHS-1 and 2

large proportion of Muslim couples NFHS-2 and the issue of religious fer-1.5

use contraception, and that this has


tility differentials investigated; a spe-increased in recent years, from 28% in

Figure 2: Trends in Mean Children Ever Born to Women of Age 40-49

cial issue of the EPW (No 5, Vol XL, 1992-93 to 46% in 2005-06 (IIPS 1995;

by Religion, NFHS-1 to 3

2005) was devoted to this. In particu-

Sikh Christian
IIPS and Macro International 2007).

6 Muslim

lar, questions on the nature and extent All evidence shows that a fertility



of differences and the factors respon

transition is in progress among Mussible were a ddressed. Now that we

lims as it is in other communities.


have access to data from the NFHS-3, it

There has been an impressive decline is possible to see whether the differen-3

in fertility among Muslims, and there


tials have been static or have been is fertility regulation using modern changing, and if so, in what direction. In the United States, dif-methods of contraception, including sterilisation. Yet differenferences in Catholic and Protestant fertility, which were once tials persist. Extensive and independent multivariate analyses of quite wide, narrowed considerably over time in spite of injunc-the NFHS-1 and the NFHS-2 data showed that the religious differtions on contraceptive use by the church (Jones and Nortman ences in fertility are not primarily attributable to differences in 1968; Westoff and Jones 1979; Mosher and Hendershot 1984; other socio-economic factors. Though the level of education is Goldscheider and Mosher 1988). Are the differentials in India lower among Muslims than other major religions, it does not also narrowing? The question is relevant since India has been e xplain the observed fertility differences or the “gross effects”. undergoing a fertility transition in the last three decades. Initial

Table 1: Trends in Total Fertility Rate and Completed Fertility by Religion,

analyses showed that some sections were lagging behind. How

NFHS-1 to NFHS-3

rapidly have they moved, and are they catching up? Specifically,

Religion Indicator at Absolute Per Cent Deviation from Mean

NFHS-1 NFHS-2 NFHS-3 Decline Decline from NFHS-1 NFHS-2 NFHS-3

from NFHS-1 NFHS-1

is a convergence taking place? The data from the three rounds of

the NFHS afford a unique opportunity to look at longitudinal

to NFHS-3 to NFHS-3

changes. Though the inter-survey period of 13 years is not long,

Total fertility rate it has witnessed fertility transition in all parts of the country, Hindu 3.30 2.78 2.65 0.65 20 -0.09 -0.07 -0.03 and a study of changes over this period is useful to address the

Muslim 4.41 3.59 3.09 1.32 30 1.02 0.74 0.41 Christian 2.87 2.44 2.35 0.52 18 -0.52 -0.41 -0.33

issues involved.

Sikh 2.43 2.26 1.96 0.47 19 -0.96 -0.59 -0.72 All* 3.39 2.85 2.68 0.71 21 ---

Present State of Knowledge

Completed fertility (mean children ever born to women of ages 40-49)

Analyses of the information from various sources, including the

Hindu 4.78 4.34 3.97 0.81 17 -0.06 -0.11 -0.03 NFHS, have put a number of contentions to rest. To take the most Muslim 5.83 5.72 4.60 1.23 21 0.99 1.27 0.60 obvious one, it is true that there are clear differences in fertility Christian 4.01 3.47 3.27 0.74 18 -0.83 -0.98 -0.73

Sikh 3.99 3.59 3.56 0.43 11 -0.85 -0.86 -0.44

by religion. Of the four largest religious groups, Muslims have

All* 4.84 4.45 4.00 0.84 17 --

higher and Christians and Sikhs lower fertility than Hindus; the

*: Includes all religions. NFHS-3 showed that the TFR was 3.09, 2.65, 2.35, and 1.96 for Source: IIPS (1995), IIPS and ORC Macro (2000), IIPS and Macro International (2007).

Economic & Political Weekly

There seems to be a “net” effect of religion after statistically controlling the influences of socio-economic factors such as education, standard of living, place of residence, and work status, which nullifies the characteristic hypothesis (Bhat and Zavier 2005; Bhagat and Praharaj 2005; Dharmalingam and Morgan 2003; Dharmalingam et al 2005; James and Nair 2005; Kulkarni and Alagarajan 2005; also see Shariff 1996; Moulasha and Rao 1999; Jeffery and Jeffery 2000 for earlier work). The minority status also does not explain the higher than average fertility of Muslims since the Christian and Sikh minorities do not have high fertility. Besides, there is evidence of higher than average fertility among Muslims even in some Muslim- majority areas (Chaudhury

Table 2: Trends in Contraceptive Prevalence Rate by Religion, NFHS-1 to NFHS-3

Religion Per Cent of Couples of Reproductive Increase from Deviation from Mean
Age Practising Contraception NFHS-1 to
Any method
Hindu 41.6 49.2 57.8 16.2 1.0 1.0 1.5
Muslim 27.7 37.0 45.7 18.0 -12.9 -11.2 -10.6
Christian 48.3 52.4 57.6 9.3 7.7 4.2 1.3
Sikh 57.6 65.2 66.5 8.9 17.0 17.0 10.2
All* 40.6 48.2 56.3 15.7 - - -
Any modern method
Hindu 37.7 44.3 50.2 12.5 1.3 1.5 1.7
Muslim 22.0 30.2 36.4 14.4 -14.3 -12.6 -12.1
Christian 40.3 44.9 48.9 8.6 4.0 2.1 0.4
Sikh 50.0 54.7 58.4 8.4 13.7 11.9 9.9
All* 36.3 42.8 48.5 12.2 - - -

*: Includes all religions. Source: IIPS (1995), IIPS and ORC Macro (2000), IIPS and Macro International (2007).

1984; Morgan et al 2002). The net effect of religion is seen within states as well, though the magnitude varies across them. We are now in a position to see whether the gross, and further the net, differences have remained constant or have been changing. In particular, are the gaps between different religious groups in I ndia narrowing, leading to a convergence of fertility?

Differentials at National Level

We first examine trends in the differentials in the most commonly used measure of fertility, the TFR. The TFR is the average number of births a woman will have in a lifetime if she passes through the reproductive ages following a given fertility schedule. As the fertility schedule (the set of age-specific fertility rates) refers to a period, the TFR is a period measure; in the NFHS this was the three years before the survey. So, the TFR estimated from the NFHS-1 roughly refers to 1990-92, from the NFHS-2 to 1996-98 and from the NFHS-3 to 2003-05. Evidence from the three rounds of the NFHS shows that the TFR declined by 0.71, from 3.39 to 2.68, that is, by 21% (Table 1, p 45). The largest decline occurred among Muslims, in absolute (1.32 points) as well as relative terms (30%), from 4.41 to 3.09. As a result, though fertility among Muslims remains above average, the gap (shown as deviation from the mean in the table) has narrowed over the inter-survey period, from 1.02 in the NFHS-1 to only 0.41 in the NFHS-3 (see Figure 1, p 45). At the other end, fertility among Christians and Sikhs has been below a verage, but to a lesser extent now than in the past. Among H indus, fertility levels and trends are naturally close to the average; thus comparison with the average practically amounts to c omparison with Hindus.

The lower panel of Table 1 gives trends in completed family size, measured as the mean children ever born to women aged 40 to 49 at the time of the survey (assuming that childbearing after the age of 40 is not common, this mean is taken to represent the completed family size of the women of this age group); this is a cohort measure (in contrast to the TFR which is a period measure) as it is based on the experience of the cohort of women born 40 to 50 years before the survey. This indicator also shows higher (than average) fertility among Muslims and lower fertility among Christians and Sikhs. Cumulative fertility has also declined and the fall has been greater among Muslims. But in relative terms, the fall is not as steep as in the case of the TFR. In particular, there is a 19% decline in TFR but only 11% decline in cumulative fertility among Sikhs, indicating that there has been a sharp decline recently. The same is true of Muslims. In other words, while declines have occurred for all the major religions, the timings have not been identical – relatively steep declines among Muslims and Sikhs have taken place recently (Figure 2, p 45).

While a number of intermediate factors influence fertility and four proximate determinants are well recognised (Bongaarts and Potter 1983), research has shown that the principal factor determining fertility differentials in recent decades is contraceptive use. Therefore, we look at differentials in the contraceptive prevalence rate, or the percentage of couples of reproductive age u sing contraception. The comparative picture shows substantially (by over 10 percentage points) higher than average use among Sikhs and lower use among Muslims in all the three rounds of the NFHS (Table 2). Overall, there has been a large rise in the prevalence rate, from 41% to 56%. The increase has been larger than average among Muslims, with the result the gaps have narrowed. The level among Christians, well above the average initially, came close to the average by the NFHS-3. Prevalence among Muslims remains below the average, by about 10 points, in spite of a large rise, from 28% to 46%, because the average increased as well. But some narrowing has occurred, by about two points (Figure 3, p 47). A similar picture is seen if the prevalence of only modern methods is examined (lower panel of Table 2); there is a slightly wider gap between Muslims and the others.

Table 3: Trends in Desire for Additional Children by Religion, NFHS-1 to NFHS-3

Religion Per Cent of Married Women of Increase Deviation from Mean
Reproductive Age Who Want from
No Additional Children NFHS-1 to
For women with two living children
Hindu 60.9 73.8 84.0 23.1 1.2 1.5 0.8
Muslim 38.6 51.3 74.0 35.4 -21.1 -21.0 -9.2
Christian 74.2 80.3 87.1 12.9 14.5 8.0 3.9
Sikh 74.4 88.7 92.1 7.7 14.7 16.4 8.9
All* 59.7 72.3 83.2 23.5 -
For women with three living children
Hindu 78.3 85.2 90.8 12.5 1.3 0.9 -0.9
Muslim 60.1 72.8 84.8 24.7 -16.9 -11.5 -6.9
Christian 79.8 83.2 91.0 11.2 2.8 - 1.1 -0.7
Sikh 89.0 93.6 96.1 7.1 12.0 9.3 4.4
All* 77.0 84.3 91.7 14.7 - - -

*: Includes all religions. Source: IIPS (1995), IIPS and ORC Macro (2000), IIPS and Macro International (2007).

november 29, 2008

The trend of Sikhs and Christians showing a slower rise after the prevalence reaches a high level suggests that a further rise is likely to be slow among Hindus as well, since the level in the NFHS-3 was fairly high. But there is scope for the rapid rise continuing among Muslims, and this would result in narrowing the gap in contraceptive use and, by implication, in fertility.

In a regime of voluntary fertility regulation, differentials in fertility and contraceptive use would arise out of differentials in the desired family size. In India, contraceptive use is primarily for limiting fertility rather than for spacing. According to the NFHS-3, 56.3% of couples were using some contraception. The bulk of this, 48.3%, used modern methods while a substantial 38.3% depended on sterilisation (mostly female). So, it is the d esire for additional children that has a strong bearing on contraceptive use. The NFHS asked married women (and also husbands in NFHS-3) whether they wanted to have an additional child. A lmost all couples (more than 90% in the NFHS-3) with four or more children do not desire additional children. Even at two or three children, the desire to limit family size is strong. Over time, a greater percentage of women have wanted to limit family size to just two or three children. At two children, the percentage d esiring no more has risen from 60 to 83, and at three children, from 77 to 90, from NFHS-1 to NFHS-3

Figure 3: Trends in Contraceptive Prevalence Rate (Table 3, p 46). As expected, religious by Religion, NFHS-1 to 3 religions vary. We have noted that

(% of couples of reproductive age using contraception)

differentials are seen in this, with a earlier analyses, especially of the lesser proportion of Muslims and


Muslim Christian Sikh Hindu
NFHS-1 and NFHS-2 data, did not supgreater proportion of Sikhs and

port this hypothesis and that the


higher than average fertility among than the average. But over time, the

Christians desiring no more children Muslims cannot be explained by this


alone. We now see if this continues to of Muslim women with two children 40

gaps have narrowed. While only 39%

hold good in the NFHS-3, and if so, desired no more in the NFHS-1, 21

whether the net influence of religion points below the average of 60%, by


has changed. To this end, a multiple the NFHS-3 the figure nearly doubled

classification analysis (MCA) has been


carried out with children ever born to 74% and the gap was less than 10


points. At three children, a majority as the dependent variable and four (60%) of Muslims did not want more Figure 4: Trends in Adjusted Mean Children Ever Born key factors that possibly have a bear

by Religion, NFHS-1 to 3

children even at the time of the 4
NFHS-1, but by the time of NFHS-3,
this had gone up to 85%.
Clearly, the desire for a small 3
f amily has become strong among
couples of all religions and the gaps
have narrowed in a very short time. If one looks at the trends, the percent 2 NFHS-1

ages for Muslims in the NFHS-3 are close to those for Sikhs and Christians in the NFHS-1 and Hindus in the NFHS-2. So, the desire to limit family size is common to all the communities, but the timing has varied, with Sikhs and Christians ahead, followed by Hindus with a lag and then by Muslims with an additional lag. The observed gaps at a point in time, specifically the time of the survey in the present exercise, are primarily products of the lag than of absolute differences.

The translation of fertility desires into actual regulation is not always possible because this requires access to contraception, particularly a method that is acceptable and affordable. Given

Economic & Political Weekly EPW november 29, 2008

that the Indian population programme provides many free contraceptive services, the issue of affordability is not crucial. H owever, access to a contraceptive method of choice is important because some methods may not be desired even when fertility regulation is. So, in spite of the cafeteria approach of the Indian programme, there is a substantial unmet need for contraception, though this has fallen from about 20% to 13% from the NFHS-1 to the NFHS-3 (Table 4). This has remained relatively high for Muslims, suggesting that the higher than average fertility is at least partly due to the unavailability of a contraceptive method of choice. An issue that has been highlighted in the past is the relatively lower preference of Muslims for sterilisation, the most commonly used method of birth control in India. The NFHS-3 shows that while 38% of couples were sterilised overall, the e xtent was only 22% among Muslim couples (IIPS and Macro I nternational 2007). On the other hand, the use of reversible contraception is higher among Muslims (15%) than the average (10%).

Assessing the Net Effect of Religion

One of the most common explanations of religious fertility differentials is provided by the characteristics hypothesis, that is, the observed differences are because the other socio-economic conditions of populations of different

Muslim Sikh Christian Hindu

ing on fertility as explanatory variables, along with religion. The four factors are place of residence (rural or urban), woman’s education (illiterate, primary, middle, and high school or higher), woman’s work status (working or not), and standard of


living (low, medium, or high, ascertained in the NFHS on the basis of an index constructed from

Table 4: Unmet Need for Contraception by Religion, NFHS-1 to NFHS-3

Religion Per Cent of Couples with Unmet Decrease from Deviation from Mean Need for Contraception NFHS-1 to NFHS-1 NFHS-2 NFHS-3 NFHS-3 NFHS-1 NFHS-2 NFHS-3

Hindu 18.9 15.1 11.9 7.0 -0.6 -0.7 -0.9

Muslim 25.8 22.0 18.8 7.0 6.3 6.2 6.0

Christian 15.3 14.8 12.5 2.8 -4.2 -1.0 -0.3

*: Includes all religions. Source: IIPS (1995), IIPS and ORC Macro (2000), IIPS and Macro International (2007).

Table 5: Multiple Classification Analysis of Children Ever Born on Socio-economic and and Sikhs is lower than average (the average essentially means Religion Variables, NFHS-1, NFHS-2, NFHS-3

Hindus because of their large population share) primarily because

Variable/Category Unadjusted Means Adjusted Means NFHS NFHS the socio-economic conditions of these two communities are 1 2 3 1 2 3

more favourable to low fertility than the average. The character-

Grand mean 3.21 3.12 2.93 3.21 3.12 2.93

istics hypothesis is clearly supported here. However, the adjusted

Religion (eta/beta) 0.09 0.11 0.12 0.09** 0.11** 0.11**

means for Muslims do not markedly differ from the unadjusted

Hindu 3.16 3.05 2.86 3.14 3.02 2.85

means and remain above average. Moreover, the gap remains

Muslim 3.72 3.70 3.49 3.74 3.72 3.47 Christian 2.83 2.66 2.47 3.10 2.96 2.79 nearly the same in all the three survey analyses; the adjusted Sikh 2.96 2.67 2.47 3.27 3.10 2.81 mean for Muslims is higher than the average by 0.53 (3.74 - 3.21), Place of residence (eta/beta) 0.07 0.09 0.12 0.01** 0.03** 0.03** 0.60 (3.72 - 3.12), and 0.54 (3.47 - 2.93) in the NFHS-1, NFHS-2 and Urban 2.93 2.78 2.56 3.16 3.00 2.83

NFHS-3, respectively (Table 5). Thus, there is a clear net differ-

Rural 3.32 3.23 3.09 3.24 3.15 2.97

ence between Muslims and the others and it has remained fairly

Work status (eta/beta) 0.05 0.09 0.13 0.04** 0.02** 0.01**

constant over the period of study in spite of an overall decline

Working 3.37 3.36 3.23 3.08 3.06 2.91

in fertility.

Non-working 3.14 2.96 2.69 3.28 3.14 2.94

Education (eta/beta) 0.25 0.31 0.39 0.09** 0.13** 0.17** The persistence of a gap between Muslims and the populations Illiterate 3.58 3.61 3.69 3.36 3.34 3.28 of other religions does not necessarily mean that such a differLiterate/primary 3.01 2.88 2.78 3.05 2.89 2.73

ence exists at all socio-economic levels. It is possible that the gap

Middle complete 2.34 2.21 2.10 2.94 2.80 2.58

is narrower or non-existent for some sections. This can be exam-

High school + 1.97 1.89 1.53 2.82 2.64 2.39

ined by looking at the interactions between religion and other

Standard of living (eta/beta) 0.12 0.14 0.19 0.06** 0.06** 0.09**

factors. We have not looked at interactions in this paper (the

Low 3.42 3.37 3.40 3.32 3.22 3.17

o ther factors are merely controlled) but a detailed investigation

Medium 3.09 3.17 3.10 3.14 3.14 3.00 High 2.61 2.57 2.50 2.91 2.89 2.72 for Kerala showed that the gap was quite wide at a low level of R2 0.447 0.465 0.439 education but narrowed at a higher level (Alagarajan 2003). The No of observations 80,076 74,993 82,852

Table 6: Logistic Regression of Contraceptive Use by Religion,

* and **: Significant at 5% and 1% levels respectively. Source: Obtained from NFHS-1, NFHS-2, and NFHS-3 data on Children Ever Born (for usual resident, currently married for Couples of Reproductive Age, NFHS-1, NFHS-2, NFHS-3

women, married only once) with marital duration as covariate. Explanatory Variable With Two Living Children With Three Living Children NFHS-1 NFHS-2 NFHS-3 NFHS-1 NFHS-2 NFHS-3

i nformation on housing conditions and assets; details of the con

Odds Ratios

struction of the index are available in the NFHS reports). Demo-


graphic literature has extensively discussed the roles of these fac-

Hindu (Ref) tors in fertility behaviour. The analysis is for currently married Muslim 0.564** 0.505** 0.674** 0.410** 0.581** 0.622** women of reproductive ages who have been married only once – Christian 1.297** 1.155 1.265** 1.225* 0.935 1.311 this is to avoid complications introduced by marital dissolution. Sikh 1.277** 1.692** 1.027 1.631** 1.644** 1.082 Marital duration is the covariate since the dependent variable, Place of residence

Rural (Ref)

children ever born, increases with marital duration.

Urban 1.052* 1.215** 1.168** 1.095** 1.286** 1.234**

Table 5 presents the results of the MCA from each of the three

Education of the respondent

rounds of the NFHS so as to facilitate an examination of change.

Illiterate (Ref)

The MCA gives unadjusted and adjusted means for each category;

Liter-primary 1.030 1.170** 1.091* 1.271** 1.210** 1.210**

the unadjusted mean is the raw value and shows gross differenc-

Middle complete 1.418** 1.225** 1.165** 1.306** 1.186** 1.243** es while the adjusted means are computed after controlling for High school + 1.521** 1.389** 1.371** 1.208** 1.169** 0.823

the effects of the other factors and of the covariate. For each Work status of the respondent

f actor, the eta values give the gross effect of the category and the Non-working (Ref) Working 1.286** 1.342** 1.247** 1.425** 1.338** 1.252**

beta the net effect, akin to partial correlation. Education of the

Standard of living index

woman shows the largest gross differentials as seen from the

Low (Ref)

large eta values; after controlling for other factors, the effect of

Medium 1.063* 1.003 0.948* 1.181** 1.014 1.006

education diminishes considerably yet it remains the most influ-

High 1.341** 1.238** 1.362** 1.357** 1.271** 1.331** ential factor. Standard of living has a notable gross effect but a No of living sons small though significant net effect. Place of residence and work 3 (Ref) (Ref) (Ref) status show very small net influences. Religion has a moderate

2 (Ref) (Ref) (Ref) 1.965** 1.803 1.798

1 0.539** 1.137** 1.121** 0.857** 0.920 0.917

gross effect that persists even after controlling for the influences

0 1.095** 0.517** 0.537** 0.340** 0.346 0.410

of other factors; note that the beta values are nearly identical to

Experienced child loss

the eta values. The results are consistent in the analyses of all the

No child loss (Ref)

three rounds of the NFHS.

Child loss 1.026 0.966 0.949** 0.873** 0.858 0.825 Looking at differences in pairs, the Hindu-Christian and Hindu-Chi square 2454.1 2405.1 2155.6 2836.3 1905.2 1566.5

Sikh gaps vanish after other socio-economic factors are control-Number of women 16,479 18,924 23,698 16,387 17,171 17,653

led (Figure 4, p 47) as the adjusted means for the three communi- * and **: Significant at 5% and 1 % levels respectively.

Source: Obtained from series of logistic regression analyses of NFHS-1, NFHS-2 and NFHS-3 data on contraceptive use ties are quite close. This implies that the fertility among Christians of couples (for usual resident, currently married women, married only once).

48 november 29, 2008

observed gap is thus an “average gap” between Muslims and the others, not a uniform gap.

What about contraceptive prevalence? To see to what extent the observed gross differences (shown in Table 2) are pure effects of religion, a regression analysis has been carried out with r eligion and the four socio-economic factors noted above as explanatory variables. Since the dependent variable, contraceptive use, is dichotomous (user or non- user), the technique of logistic regression has been invoked. In this formulation, the probability of contraceptive use (actually the log of the odds, log[p/(1-p)]) is postulated to depend on a set of variables. Since contraceptive use heavily depends on the number of living children, and it is at two and three children that the largest differentials occur (at one child the use is very low for all groups, and at four or more children, it is very high for all), two analyses have been carried out, one for women with two living children and the other for women with three living children. Besides, as the sex of living children matters in adoption of contraception, given the strong preference for male children in India, the number of sons has also been i ncluded as a variable. An additional variable is experience of child loss (dichotomous, experienced or not) because child loss would naturally lower the tendency to use contraception. Most of the explanatory variables are categorised, and the logistic regression procedure allows for this. This necessitated designating one category as reference and the regression coefficients can be used to give the odds ratio, that is, ratio of odds for a specific category to the reference category.

The results of the logistic regression analysis are given in Table 6 (p 48). There are six regressions, three each for the NFHS-1, NFHS-2 and NFHS-3 for women with two and three living children. We concentrate on the effect of religion. In this analysis, “Hindu” is treated as the reference category since it is the largest religious group and the odds ratios for other religions are the ratios of odds for the specific religion to the odds for Hindus. For Christians and Sikhs, the odds ratio is greater than one, implying a higher (than the reference group Hindu) tendency to use contraception at the same socio-economic levels though by the NFHS-3, this effect was small and generally not significant. But the odds ratio for Muslims is close to 0.5 in all the regressions. So, the probability of u sing contraception is lower among Muslims than among Hindus (and by implication also lower than among Christians and Sikhs) with identical other conditions. Some catching up seems to have occurred since the odds ratio in the NFHS-3 is closer to one than in the NFHS-1 (at two living children, the ratio rose from 0.564 to 0.674 and at three living children from 0.410 to 0.622); an odds ratio of one implies equality and values close to one imply low inequality.

It must be clarified that the odds ratio of 0.5 does not imply that the probability of contraceptive use among Muslims is half that of Hindus; the odds ratio is not a ratio of probabilities. If the prevalence level among Hindus is 0.6 (60%), at an odds ratio of 0.5, the level among Muslims would not be 0.3 (30%) but 0.43 (43%). So, though the logistic regression coefficients or the odds ratios are useful to assess the effects of a factor, it is useful to look at predicted probabilities for a clearer understanding. In this case, the predicted probability of contraceptive use for each

Economic & Political Weekly

november 29, 2008

r eligious group has been computed holding other factors at the average, thereby making it possible to assess the net effects in terms of differences in probabilities. The predicted probabilities are like the adjusted means in the MCA and have been presented in Table 7 in terms of percentages, along with the actual percentages, which are similar to the unadjusted means in the MCA. It can be seen that the predicted percentages for Muslims are higher than the actual percentages and thus, controlled for other factors, the gap is narrower. For Christians and Sikhs, the predicted percentages are lower than actual or about the same. However, even after other factors are controlled, a substantial difference between Muslims and the others persists. At the same time, over the survey period, some decline has occurred. For women with two children, the gap between Muslims and Hindus was 13 points after adjustment in the NFHS-1 and it fell to nine points by the NFHS-3. For women with three children, the difference in adjusted percentages fell from 22 points in the NFHS-1 to 10 points in the NFHS-3. Clearly, the gap seems to have narrowed over time. B esides, the contraceptive prevalence among Muslim couples with two or three children is fairly high, indicating a strong desire to limit family size.

Thus, at the national level, we see some convergence occurring, with the gap between Muslims and the others narrowing in TFR and in contraceptive use. However, in cumulative fertility, the difference between Muslims and the others persists over the survey period. The persistence of gaps in cumulative fertility (assessed from children ever born) but the narrowing in current fertility (assessed from TFR and from contraceptive use) shows that the gaps that persisted for some time in the recent past have narrowed over time. There is a movement towards convergence.

Analysis at State Level

Given the large and well-known regional variations in fertility transition in India, and that the populations of different religions are not uniformly spread over various regions, it is possible that the observed differences are because of the regional variations. For example, a large share of the Christian population is in Kerala, which has very low fertility. Similarly, a large share of the Sikh

Table 7: Actual and Adjusted Percentages of Couples of Reproductive Age Using Contraception by Religion, NFHS-1, NFHS-2 and NFHS-3

Survey NFHS-1 NFHS-2 NFHS-3
Actual Adjusted Actual Adjusted Actual Adjusted
For women with two living children
Hindu 45 45 61 60 73 72
Muslim 28 32 41 44 57 63
Christian 57 52 70 64 77 76
Sikh 60 51 76 72 76 72
All 44 59 71
For women with three living children
Hindu 58 58 69 69 77 76
Muslim 34 36 55 56 61 66
Christian 64 63 70 68 75 78
Sikh 71 69 79 79 76 77
All 57 68 75

The percentages are for women currently married and married only once. The adjusted percentages are obtained from predicted probabilities of contraceptive use, in a manner like the MCA, from logistic regression coefficients with place of residence, education of respondent, work status, standard of living, number of living sons, and experience of child loss as other factors (see Table 6).

population lives in Punjab, which also has low fertility. On the appropriate weights to obtain estimates for pre-division Bihar. other hand, two states with relatively high fertility, Uttar Pradesh Similar pooling has been done for Madhya Pradesh and Chhatand Bihar, have large shares of the Muslim population. In order tisgarh and for Uttar Pradesh and Uttarakhand. In many states, to avoid a confounding of the region and religion factors, it is the populations of only two religions could be considered necessary to control the effect of one to examine that of the other. b ecause the sample sizes of other religious groups were quite small. The Hindu-Muslim differences could be examined in 13

Table 8: Unadjusted and Adjusted Differences in Mean Children Ever Born by Religion, Large States, NFHS-1, NFHS-2 and NFHS-3 states, the Hindu-Christian differences in three states, and the

Survey NFHS-1 NFHS-2 NFHS-3

Hindu-Sikh difference in two states. Instead of giving all the

Actual Adjusted Actual Adjusted Actual Adjusted

MCA results for each state and each survey, the results as they

Muslim-Hindu Andhra Pradesh 0.42 0.58 0.69 0.88 0.28 0.37 relate to religion are shown in Table 8. Assam 0.56 0.37 0.30 0.38 0.99 0.65 The mean number of children ever born is higher for Muslims

Bihar1 0.51 0.51 0.65 0.53 0.94 0.80 than Hindus in all the states, but the magnitude varies. Most Gujarat 0.05 0.18 0.20 0.36 0.15 0.20 states show gaps around 0.5; some show wider gaps in one or Jammu and Kashmir --0.29 0.15 0.72 0.39

two rounds of the survey. At the same time, differences are quiet

Karnataka 0.59 0.68 0.84 0.90 0.57 0.72

small in a few states, notably Gujarat and Tamil Nadu. In gener-

Kerala 0.74 0.62 0.72 0.62 0.61 0.53

al, the adjusted means do not differ much from the unadjusted

Madhya Pradesh2 0.65 0.71 0.41 0.70 0.09 0.37

means, implying that the gaps persist even after controlling for

Maharashtra 0.77 0.84 0.55 0.82 0.59 0.75

other socio-economic factors, as was observed at the national

Rajasthan 0.36 0.31 0.82 0.76 0.41 0.38

level (notable exceptions are West Bengal in the NFHS-2 and

Tamil Nadu 0.47 0.51 0.28 0.42 0.02 0.16 Uttar Pradesh3 0.26 0.40 0.71 0.83 0.56 0.60 A ssam and Jammu and Kashmir in the NFHS-3, where the gap West Bengal 0.60 0.67 1.21 0.90 0.98 0.82 narrows, and Madhya Pradesh and Maharashtra in the NFHS-2, India 0.56 0.60 0.65 0.70 0.63 0.62 where the gap widens after adjustment). The Hindu-Christian Christian-Hindu gaps and Hindu-Sikh gaps are relatively small and become even Andhra Pradesh 0.19 0.13 0.03 -0.12 -0.13 -0.10

smaller (sometimes with a reversal of sign) after adjustment.

Kerala 0.17 0.15 0.05 0.15 0.03 0.11

Contraceptive prevalence presents a similar picture. Hindu-

Tamil Nadu -0.03 0.12 -0.05 0.25 0.08 0.16

Christian and Hindu-Sikh differences are small in the few states

India -0.33 -0.04 -0.39 -0.06 -0.39 -0.06

where it is possible to make comparisons (Table 9). However,

Sikh-Hindu Haryana -0.17 -0.05 -0.20 0.11 -0.15 -0.21 Hindu-Muslim differences are notable in many states, generally Punjab -0.08 -0.09 -0.21 -0.18 -0.17 -0.15 more than 10 percentage points, though smaller gaps are seen in India -0.20 0.13 -0.38 0.08 -0.39 -0.04 Tamil Nadu, Gujarat, Madhya Pradesh and Andhra Pradesh.

(1) Includes Bihar and Jharkhand;

States that have an overall high prevalence also show high

  • (2) includes Madhya Pradesh and Chhattisgarh;
  • (3) includes Uttar Pradesh and Uttarakhand. Source: Computed from MCA for children ever born for each state using data from the NFHS-1, NFHS-2 and NFHS-3,
  • Table 9: Per Cent of Couples of Reproductive Ages Practising Contraception by

    for women currently married and married only once with place of residence, education of respondent, work status,

    Religion, All India and Large States, NFHS-1, NFHS-2 and NFHS-3

    and standard of living as other factors and marital duration as the covariate.

    Survey NFHS-1 NFHS-2 NFHS-3 This could either be done by inserting a variable for region in the Religion Hindu Muslim Other@ Hindu Muslim Other@ Hindu Muslim Other@

    statistical analysis or carrying out separate analyses for states. State Andhra Pradesh 45.3 42.6 50.0 63.3 47.2 55.4 70.0 63.9 70.9

    We prefer the latter route since it allows one to see if the effect of

    Assam 44.2 29.5 -50.9 34.0 -61.5 46.7

    religion varies across regions without getting into a large number

    Bihar1 25.8 7.5 -27.8 8.9 -39.9 21.6

    of interaction terms.

    Gujarat 49.9 32.2 -61.0 58.9 -69.4 63.1 -

    Ideally, one must first look at trends in TFRs by religion in

    Haryana 51.5 -47.2 64.8 -70.9 68.2 -69.2

    various states. The NFHS-3 report gives TFRs for various reli-

    Jammu and Kashmir ---55.2 50.3 -57.5 51.7

    gions at the national level but not for different states; state-level

    Karnataka 48.4 35.6 -63.0 46.5 -66.5 57.0 differentials would be available in the state reports but only Kerala 68.2 35.7 69.6 73.9 49.9 73.0 76.1 57.6 76.6

    some of the NFHS-3 state reports have been released. So, we Madhya Pradesh2 35.7 38.8 -45.4 45.7 -56.5 56.9 e xamine children ever born, that is, cumulative fertility, com-Maharashtra 54.1 35.4 -63.5 49.0 -69.5 59.6 puted from the NFHS-3 data, along with similar values from the Punjab 59.0 -58.3 70.3 -67.9 63.5 -65.6 NFHS-1 and NFHS-2. For each state, an MCA has been done for

    Rajasthan 32.5 18.1 -43.8 28.0 -50.4 39.0 -

    Tamil Nadu 47.3 42.1 46.9 55.4 53.9 55.0 62.3 56.4 64.5

    children ever born with religion and other socio-economic vari-

    Uttar Pradesh3 22.1 10.9 -32.4 22.1 -49.4 31.1

    ables (place of residence, woman’s education, woman’s work

    West Bengal 57.9 40.3 -72.6 58.5 -75.9 62.3

    status, and standard of living), and marital duration as the

    India 40.9 26.8 51.2 38.1 59.8 47.2

    c ovariate. This allows assessments of net effects of religion,

    Christian 45.1 53.7 58.6

    seen from the adjusted means, after controlling for other socio-

    Sikh 56.9 66.2 67.4 economic factors and marital duration. The analysis has been @ : Other religion is Christian in Andhra Pradesh, Kerala, and Tamil Nadu and Sikh in Haryana and Punjab.

    (1) Includes Bihar and Jharkhand;

    done only for states that have large samples of at least two reli

  • (2) includes Madhya Pradesh and Chhattisgarh;
  • (3) includes Uttar Pradesh and Uttarakhand.
  • gions. To allow comparisons across the three surveys, the

    Source: Computed from data files of NFHS-1, NFHS-2, NFHS-3 for women usual resident, currently married and NFHS-3 data for Bihar and Jharkhand have been pooled with married only once.

    50 november 29, 2008

    p revalence for Muslims. Low levels of contraceptive use are seen only in Bihar, Uttar Pradesh and Rajasthan, states where overall levels are low. By the NFHS-3, or 2005-06, the majority of Muslim couples of reproductive ages in more than half the large states were using contraception.

    The net influence of religion on contraceptive use in various states was examined using logistic regression, as was done at the national level, but for the sake of brevity, only the principal findings are noted here. In the case of couples with two living children, contraceptive use among Muslims is significantly lower than the average (or Hindus) in six of the 13 states considered. The odds ratios show fluctuations over the three surveys, suggesting that the relative gaps have been changing, not necessarily in the same direction. A similar picture is seen for couples with three living children. But the odds ratios for Muslims are generally lower here and the effect is significant in nine states, implying that differences at this stage of childbearing are sharper. G ujarat, Tamil Nadu and Madhya Pradesh do not show significant differences between contraceptive use among Hindus and Muslims.

    To sum up, the analysis for individual states does not present a picture much different from the national. The Hindu-Muslim gaps are not explained by differences in socio-economic background while the Hindu-Christian and Hindu-Sikh gaps are small and mostly accounted for by the characteristics explanation. But there is variation in the degree of the effect of religion, especially the Hindu-Muslim difference, across states, though a few states do not show significant differences. Moreover, the effect of r eligion has been changing over time, as can be seen from the three rounds of the NFHS.


    Past research on the fertility-religion link in India showed that fertility is higher and contraceptive use lower among Muslims than others, and that the difference cannot be explained by the characteristics hypothesis. At the same time, a notable decline in fertility occurred among Muslims. Though the broad picture was similar in large states, the degree of religious difference varied considerably. The analysis of the data from the NFHS-3 corroborates these findings. The principal question is: are the gaps narrowing?

    Differences have narrowed in the TFR and, to a smaller extent, trends. Since we do not yet have the state-level TFR values by religion, changes in gaps in adjusted mean children ever born ( obtained from Table 8) from the NFHS-1 to the NFHS-2 and from the NFHS-2 to NFHS-3 are cross-tabulated in Table 10. This has been done only for the Hindu-Muslim gaps since the other gaps are available for very few states and they are quite small anyway.

    A gap is labelled as having “narrowed” if the decline is at least 0.1, and “widened” if the rise is at least 0.1, otherwise it is called “steady”; minor changes are ignored as the results are based on sample surveys. Ideally, one should examine the pace of change in the inter-survey periods but since the intervals between the successive rounds have been nearly equal, six years between the NFHS-1 and the NFHS-2 and seven years between the NFHS-2 and the NFHS-3, a comparison of inter-survey changes serves the purpose. Between the NFHS-1 and the NFHS-2, the gap did not narrow in any of the states, but widened in six large states, and remained steady in five. Between the NFHS-2 and the NFHS-3, the gap narrowed in seven states, five of them where it had widened recently, Andhra Pradesh, Gujarat, Karnataka, Rajasthan, and Uttar Pradesh. In two states, the gap widened after the NFHS-2. These were Bihar and Assam, states at an early stage of transition. The gaps were steady in two states, Kerala and Maharashtra, in both the inter-survey periods. Overall, the situation is in a flux, with gaps narrowing or widening in different states.

    This was, of course, expected; the states are at different stages of transition and through the process, gaps open up and then close. In the US, the Catholic-Protestant gap was narrow in the 1950s, widened in the 1960s when many Catholics were reluctant to adopt contraception, and narrowed in the 1970s when most Catholics accepted contraception (Westoff and Jones 1979). So, the gaps should be narrow for states at an early transitional stage and also for states at an advanced stage, and wide for states in the middle. In many states, a widening has been followed by narrowing, as expected. The widening of gaps in Assam and Bihar also falls into a pattern. But, contrary to expectation, Kerala continues to show wide gaps even with overall low fertility, as do Maharashtra and West Bengal. At the other end, Madhya Pradesh, Rajasthan, and Uttar Pradesh show a narrowing of gaps even though fertility among Hindus has not reached a low level. Thus, no clear correspondence is seen between the level of fertility or the stage of transition and the gap.

    in contraceptive use but the changes are not so

    Table 10: Changes in Hindu-Muslim Gaps in Adjusted Mean Children Ever Born, from NFHS-1 to NFHS-2 and conspicuous in cumulative fertility. A problem from NFHS-2 to NFHS-3 in Large States

    From NFHS-2 to NFHS-3 From NFHS-1 to NFHS-2 Gap Narrowed Gap Remained Steady Gap Widened

    with national-level indicators is that these are ag

    gregates from states that are at vastly different

    Gap narrowed None Madhya Pradesh2 (0.71, 0.70, 0.37) Andhra Pradesh (0.58, 0.88, 0.37) stages of fertility transition. While many states Tamil Nadu (0.51, 0.42, 0.16) Gujarat (0.18, 0.36, 0.20) have below r eplacement level fertility, some have Karnataka (0.68, 0.90, 0.72)

    Rajasthan (0.31, 0.76, 0.38)

    a TFR well over three. The gaps between seg-

    Uttar Pradesh3 (0.40, 0.83, 0.60)

    ments of population are expected to change in

    Gap remained None Kerala (0.62, 0.62, 0.53) West Bengal (0.67, 0.90, 0.82) accordance with the stage of transition (a low steady Maharashtra (0.84. 0.82, 0.75)

    gap initially, high in the middle of transition, and Gap widened None Assam (0.37, 0.38, 0.65) None Bihar1 (0.51, 0.53, 0.65)

    again low towards the end of transition). So at no

    (1) Only states with large enough sample sizes of Hindus and Muslims are shown.

    point during the process will all states show simi- (2) The gap is ascertained from the adjusted mean children ever born (based on MCA results in Table 8) to currently married and only once married women in the age range 15-49.

    lar gaps as some high values will be combined

  • (3) The gap is said to have “narrowed” if it reduced by at least 0.10, “widened” if it increased by at least 0.10, otherwise it is called “steady”.
  • (4) The estimates of gap in the three surveys are given in the parentheses in the order of survey.
  • with some low ones. For this reason, trends in

    Source: Tabulated from Table 8. (1) Includes Bihar and Jharkhand; (2) includes Madhya Pradesh and Chhattisgarh; (3) includes Uttar Pradesh and states should be looked at rather than the national Uttarakhand.

    Economic & Political Weekly

    Mention was earlier made about spatial variations in India’s fertility transition, and the lead taken by the southern states in this. Hence, it is worth exploring if the gaps have a regional p attern. As can be seen from Tables 8 and 10, the gaps are wide in Kerala, West Bengal, Maharashtra and Karnataka, and narrow in Tamil Nadu and Gujarat in all the three rounds. No north-south or east-west pattern is discernible.

    To look for reasons for the absence of a pattern, it is necessary to understand the manner in which religion influences fertility. The classical effect of religion, after ruling out characteristics and minority status explanations, is because of a particularised theology. Precepts and injunctions of a religion influence the d esire for family size, the inclination to regulate fertility, and the use and method of contraception. There is no doubt that the

    o pposition of the church had an influence in the fertility behaviour of Catholic couples. But today, in the West, predominantly Catholic countries, including Italy, have very low fertility. And Catholic populations in non-Catholic dominated countries such as the US have reached low fertility, with the result that Catholic-Protestant differences have nearly disappeared. Thus, even categorical opposition to contraception no longer seems to have an effect. Over time, researchers have been highlighting the issue of “religiosity” or “religiousness” rather than religious “affiliation”, and McQuillan’s formulation separates the various facets, religious values and norms, institutions, and identity (see McQuillan 2004; Westoff and Frejka 2006). The effect of religion will be seen only if people are religious, in the sense that they follow the precepts and observe the injunctions, or adhere to the values and norms that have a bearing on fertility. Religious affiliation matters little if people are not religious. Moreover, people may not necessarily interpret values and norms in the manner scriptures, institutions, or leaders do, but in some way that they feel is right. This may be the reason why many Catholics adopt contraception without feeling that they are violating any prescriptions or have become irreligious. For instance, research in rural Mexico suggests that Catholic women of the present generation feel that adopting contraception to limit the size of the family is not necessarily against doctrine (Hirsch 2008).

    While the injunctions are quite explicit in the case of the Catholic church, this is not true of other religions. There is no recognised central authority for Muslims (or for Hindus), not just globally but even in India, and there is no consensus on what could be called an “Islamic” position on fertility and contraception (for a discussion, see Omran 1992). As a result, different individuals have different perceptions of the position of the religion. While some argue that contraception is not acceptable to Islam, fertility regulation has been widely accepted by many Muslim populations around the world. A number of predominantly Muslim countries now have fertility below replacement level. Tunisia has a TFR of 1.87 and Iran 2.03, with Kuwait, T urkey, the United Arab Emirates, and Algeria close behind, i ndicating the widespread use of contraception (UNFPA 2007). This seems to be happening among Muslims in India as well. At the same time, some feel that contraception is not permitted by their religion, and the NFHS-2 results show that 12.5% of Muslim non-users of contraception mentioned this as a factor in contrast to 2% overall (IIPS and ORC Macro 2000). Thus, while perceived religious injunction matters more to Muslims than others, it does only to a small extent. Given that plenty of Muslims use contraceptives, many either do not feel that contraception violates r eligious prescriptions (or interpret the prescriptions as such) or prefer to use contraception anyway.

    The strong preference for sons among Hindus, something a bsent among Muslims, could conceivably operate in favour of higher fertility among them because it could lead to continued childbearing until the desired number of sons is born. But, in the present setting, such a preference has tended towards sex detection and selection, adversely affecting the sex ratio at birth rather than fertility. There is evidence that the sex ratio at birth has b ecome more unbalanced among Hindus than Muslims in recent years (Bhagat and Praharaj 2005).

    There is undoubtedly some gap in fertility and contraceptive adoption between Muslims and the others but this is not uniform over space or constant over time. In different parts of India, populations are adopting fertility regulation by accepting modern methods of contraception, but it has happened at different stages of the family building process and the timings have varied. Religious factors have probably delayed the transition but not prevented it. As the process of fertility transition has not moved at the same pace throughout the country, the delays have also varied. So, the gaps are not identical nor are they of the expected size. In some states, the gaps have persisted, and in some, n arrowed unexpectedly. In a matter as complex as fertility b ehaviour, a great deal of regularity cannot be expected. Interactions between religious groups can shape attitudes, weakening the adherence to norms and values of one’s own religion without necessarily leading to a loss of religious identity as such, thus hastening convergence (Chamie 1977; Chaudhury 1984; Alagarajan 2003). But that depends on how close such interactions are. One could argue that if the share of a minority community is small, it is likely to have more inter-community interactions, while with a large share, intra-community interactions tend to be dominant. In Tamil Nadu, Madhya Pradesh and Gujarat the share of the Muslim community is small, suggesting close interactions with others and small fertility differentials. At the other end, the share is large in West Bengal and Kerala, over 20%, and the gaps are wide. However, Assam with a large Muslim share did not show wide gaps in the NFHS-1 and NFHS-2, while Karnataka and Maharashtra with just about 10% Muslims show wide gaps. D irect data on inter-religious interactions are not available, preventing an assessment of the hypothesis. With increased channels of communication, closer interactions and speedier diffusion are expected in the future but the pace could vary over space and societies. This means that it is difficult to predict when the gap will close.

    Future Trend in Differentials

    To conclude, fertility transition is continuing among all major religious groups in India and sharp declines have occurred because of a steep rise in the use of contraceptives. But differentials exist and will probably continue to exist for some time, with changes in magnitude. The higher than average fertility among Muslims

    november 29, 2008

    is not explained by differences in other socio-economic charac-2.1, and fertility transition was said to be complete when the teristics. This implies that a convergence in socio-economic condi-TFR reached this level. If so, the decline in fertility among the tions by itself will not achieve a convergence in fertility, though a l eaders will stop at this level and the others could then catch up, substantial improvement in socio-economic conditions can expe-leading to a convergence. But given that the fertility in many dite the transition to low fertility. The minority factor also does countries, and almost all of Europe, has fallen well below the not seem to matter. But the gap is narrowing, as the evidence replacement level, with TFRs much lower than two, no low from the latest survey shows. In the absence of a clear trend so a symptote can be taken for granted. This means that fertility far, it is difficult to predict at this stage when the gap will close. decline among the leaders could possibly continue even beyond At the same time, the process of fertility decline is generally not (here it means below) the replacement level and the others will reversed, at least no reversal has occurred so far. So all communi-have to cover some more distance if a convergence is to occur. In ties are likely to move towards low fertility. such a situation, convergence per se could take a longer time Until recently, in most demographic discussions, “low fertili-though all groups would reach low and probably below replace

    ty” meant “replacement level low fertility”, or a TFR of close to ment level fertility.


    Alagarajan, M (2003): “An Analysis of Fertility Differentials by Religion in Kerala State: A Test of the Interaction Hypothesis”, Population Policy and Research Review, 22, pp 557-74.

    Alagarajan, M and P M Kulkarni (1998): “Fertility Differentials by Religion in Kerala – A Period Parity Progression Ratio Analysis”, Demography India, 27 (1), pp 213-28.

    Balasubramaian, K (1984): “Hindu-Muslim Differentials in Fertility and Population Growth”, Artha Vijnana, 26 (3), pp 189-216.

    Bhagat, R B and Purujit Praharaj (2005): “Hindu- Muslim Fertility Differentials”, Economic & Political Weekly, XL (5), pp 411-18.

    Bhat, P N Mari and A J Frances Zavier (2005): “Role of Religion in Fertility Decline: The Case of Indian Muslims”, Economic & Political Weekly, XL (5), pp 385-402.

    Bongaarts, John and R G Potter (1983): Fertility, Biology, and Behaviour: An Analysis of the Proximate Determinants (New York: Academic Press).

    Chamie, J (1977): “Religious Differentials in Fertility in Lebanon 1971”, Population Studies, 31 (2), pp 365-82.

    Chaudhury, R H (1984): “Hindu-Muslim Differential Fertility: How Much Religion and How Much Socio?”, Social Action, 34 (3), pp 251-73.

    Das, N and D Pandey (1985): “Fertility Differentials by Religion in India: An Analysis of 1971 Census F ertility Data”, Canadian Studies in Population, 12 (2), pp 119-35.

    Day, L H (1984): “Minority-Group Status and Fertility: A More Detailed Test of the Hypothesis”, Sociological Quarterly, 25 (4), pp 456-72.

    Dharmalingam, A and S Philip Morgan (2003): “Pervasive Muslim-Hindu Fertility Differences in I ndia”, Demography, 41 (3), pp 529-45.

    Dharmalingam A, K Navaneetham and S Philip Morgan (2005): “Muslim-Hindu Fertility Differences: Evidence from National Family Health Survey-II”, Economic & Political Weekly, XL (5), pp 429-36.

    Freedman, R and P K Whelpton (1961): “Socioeconomic Factors in Religious Differentials in Fertility”, American Sociological Review, 26 (4), pp 608-14.

    Goldscheider, C and W D Mosher (1988): “Religious Affiliation and Contraceptive Usage: Changing American Patterns, 1955-82”, Studies in Family Planning, 19 (1), pp 48-111.

    Goldscheider, C and P R Uhlenberg (1969): “Minority Group Status and Fertility”, American Journal of Sociology, 74 (4), pp 361-73.

    Hirsch, Jennifer S (2008): “Catholics Using Contraceptives: Religion, Family Planning, and Interpretive Agency in Rural Mexico”, Studies in Family Planning, 39 (2), pp 93-104.

    India, Registrar General (2007): Sample Registration System Statistical Report 2006 (New Delhi: Controller of Publications).

    Economic & Political Weekly november 29, 2008


    International Institute of Population Sciences (1995): National Family Health Survey (MCH and Family Planning), India, 1992-93, IIPS, Mumbai.

    IIPS and ORC Macro (2000): National Family Health Survey (NFHS-2), India, 1998-99, IIPS, Mumbai.

    IIPS and Macro International (2007): National Family Health Survey (NFHS-3) 2005-06, IIPS, Mumbai.

    James, K S and Sajini B Nair (2005): “Accelerated D ecline in Fertility in India since the 1980s: Trends among Hindus and Muslims”, Economic & Political Weekly, XL (5), pp 375-84.

    Jeffery, R and P Jeffery (2000): “Religion and Fertility in India”, Economic & Political Weekly, 35, pp 3253-59.

    Jones, G and D Nortman (1968): “Roman Catholic Fertility and Family Planning: A Comparative Review of Literature”, Studies in Family Planning, 34, pp 1-27.

    Kulkarni, P M and Manoj Alagarajan (2005): “Population Growth, Fertility and Religion in India”, Economic & Political Weekly, XL (5), pp 403-11.

    McQuillan, Kevin (2004): “When Does Religion Influence Fertility?”, Population and Development R eview, 30 (1), pp 25-56.

    Morgan, S P, S Stash, H L Smith and K Oppenheim M ason (2002): “Muslim and Non-Muslim Differences in Female Autonomy and Fertility: Evidence from Four Asian Countries”, Population and D evelopment Review, 28 (3), pp 515-37.

    Mosher, W D, and G E Hendershot (1984): “Religion and Fer tility: A Replication”, Demography, 21 (2), pp 185-90.

    Moulasha, K and G Rama Rao (1999): “Religion- Specific Differentials in Fertility and Family Planning”, Economic & Political Weekly, 34 (42), pp 3047-51.

    Omran, A R (1992): Family Planning in the Legacy of Islam (London: Routledge).

    Shariff, A (1996): “Socio-economic and Demographic Differentials between Hindus and Muslims in I ndia”, Economic & Political Weekly, 30 (46), pp 2947-53.

    United Nations Population Fund (UNFPA) (2007): State of World Population 2007 (New York: United Nations Population Fund).

    Visaria, L (1974): “Religious Differentials in Fertility” in Ashish Bose et al eds. Population in India’s D evelopment: 1947-2000 (Delhi: Vikas Publishing House).

    Westoff, C and E P Jones (1979): “The End of ‘Catholic’ Fertility”, Demography, 16 (2), pp 209-17.

    Westoff, C F and Tomas Frejka (2006): “Religion, Religiousness and Fertility in the US and in Europe”, paper presented at the Annual Meetings of the Population Association of America, Los Angeles, 30 March-1 April 2006.

    CD-ROM 2006

    The digital version of Economic and Political Weekly is now available for 2006 on a single disk.

    This electronic edition contains the complete content of all the issues published in 2006. The CD-ROM 2006 comes equipped with a powerful search as well as utilities to make your browsing experience productive. The contents are indexed and organised as in the print edition, with articles laid out in individual sections in each issue. Users can browse through the sections or use the sophisticated search facility to locate articles and statistics of interest.

    Price for CD-ROM 2006 (in India) Individuals – Rs 285 (Rs 250 plus postage and handling charges of Rs 35) Institutions – Rs 535 (Rs 500 plus postage and handling charges of Rs 35)

    International – US$ 40 (including airmail postage)

    Also available 2003, 2004 and 2005 on three separate CDs, individual CD price as above

    Any queries please email:

    To order the CD-ROMs (please specify the year) send a bank draft payable at Mumbai in favour of Economic and Political Weekly. The CDs can also be purchased on-line using a credit card through a secure payment gateway at

    Circulation Manager,

    Economic and Political Weekly

    320, 321, A to Z Industrial Estate, Ganpatrao Kadam Marg, Lower Parel, Mumbai 400 013, India

    Dear Reader,

    To continue reading, become a subscriber.

    Explore our attractive subscription offers.

    Click here

    Back to Top