ArticlePDF Available

Education and Childlessness in India *

January 2019

Authors:

Thomas Baudin

Université Catholique de Louvain - UCLouvain

Koyel Sarkar

New York University Abu Dhabi

Content uploaded by Koyel Sarkar

Content may be subject to copyright.

January 2019

WORKING PAPER SERIES

2019-ECO-03

Education and Childlessness in India

Thomas Baudin*

IÉSEG School of Management and LEM-CNRS (UMR 9221)

and IRES, Université catholique de Louvain

Koyel Sarkar

Center for Demographic Research, Université Catholique de

Louvain

IÉSEG School of Management

Lille Catholic University

3, rue de la Digue

F-59000 Lille

www.ieseg.fr

Tel: 33(0)3 20 54 58 92

* Corresponding author: IÉSEG School of Management, Socle de la Grande Arche, 1

Parvis de La Defense - F-92044 Paris La Defense cedex, France,

Email address: t.baudin@ieseg.fr

Education and Childlessness in India∗

Thomas Baudin†and Koyel Sarkar‡

January 11, 2019

Abstract

In a developing setting like India, women have started their long way to eman-

cipation both at the family and societal levels. In this context, we study what

may be perceived as a key sign of emancipation regarding marriage and mother-

hood: childlessness. Using micro-level regressions, we show that the probability

of a woman ending her reproductive life without children exhibits a U-shaped

relationship with her educational attainment. This is indicative of the fact that

poverty and sterility are not the sole determinants of childlessness, but that bet-

ter economic opportunities and empowerment within couples also matter. This

result is robust to the introduction of important control variables such as the

development level of the state where women live, the husband’s education, age at

marriage, religion, and caste. India seems to be joining a list of countries where

adjustments to childlessness are much more than simple responses to boom-and-

bust poverty.

Keywords: Childlessness, education, poverty, sterility, development.

∗This research has received scientiﬁc and ﬁnancial support from the ARC Project 15/19-063 on

“Family Transformations - Incentives and Norms”. Thomas Baudin also beneﬁted from ﬁnancial sup-

port from the French National Research Agency through the project MALYNES (JCJC). We would

like to thank Dudley Poston, Ester Rizzi, David de la Croix, Philippe Bocquier, Mikko Myrskyla,

Akansha Singh, Li Ma, Malgorzata Mikckuka, and Robert Stelter for their helpful remarks. We have

also signiﬁcantly beneﬁted from presentations in seminars and conferences in Rostock, Louvain la

Neuve, Prague, Durbuy, and Cape Town. All errors and imprecisions remain solely ours.

†I´

ESEG School of Management (LEM UMR 9221), and IRES, Universit´e catholique de Louvain,

E-Mail: t.baudin@ieseg.fr. Phone number: +331.55.91.10.10. Corresponding author.

‡Center for Demographic Research, Universit´e Catholique de Louvain, E-Mail:

koyel.sarkar@uclouvain.be

IÉSEG Working paper series 2019-ECO-03

1 Introduction

We propose a new interpretation of the dynamics of childlessness in India over the

last decades. Based on the recent decomposition of childlessness by Baudin et al.

(2015), we use micro-level data to show that a signiﬁcant part of childlessness among

Indian couples may be explained by the emergence of better educational and economic

opportunities for women. This phenomenon exhibits a clear geographical heterogeneity,

without contradicting the fact that childlessness among many Indian women is due to

primary infertility related to sterility, venereal diseases, and poverty.

Talking about Indian demography is talking about big numbers. India is the second

most populous country in the world after China, with an estimated population of almost

1.3 billion in 2018 (US Census Bureau). By 2024, the United Nations expect the Indian

population to become the largest in the world, with one human out of six being Indian.

As shown by Bhatt (2010) and Kulkarni (2011), the twentieth century was highly

signiﬁcant in India’s demography as the country went through a signiﬁcant decline in

its mortality levels, followed by a fertility decline initiated around the last quarter of the

century. It took nearly four decades for the total fertility rate to fall from 5.1 in 1971-

72 to 2.4 in 2011. However, this general description hides huge spatial heterogeneities.

Overall, the total fertility rate is just above the replacement level, but on a regional

scale, it ranges from as high as 3.7 in Bihar to as low as 1.8 in Tamil Nadu (SRS, 2014).1

With such a huge population, the question of when and how the Indian fertility tran-

sition will end, as well as questions about family planning programs have monopolized

the attention of demographers. As a result, hardly any space is left to study the other

important dimension of fertility - childlessness. We start ﬁlling this gap in literature

by focusing on deﬁnitive childlessness among married Indian women. Childlessness is

deﬁned by the absence of any living children in a woman’s life; when a woman remains

childless after the age of 40, it is usually called deﬁnitive childlessness.2The latest cen-

sus, conducted in 2011, recorded the highest ever deﬁnitive childlessness rate in India:

7.89 percent among women above 40 years of age. Although the rate is only around

half of that measured in countries like the US for the year 2014,3it is nevertheless far

from natural sterility rates (Leridon, 2008). More importantly, given the size of the

Indian population in absolute numbers, deﬁnitive childlessness aﬀects more than twelve

million Indian women, almost the size of the populations of Belgium or Rwanda for

instance.4Childlessness in India is clearly not a marginal issue.

The deﬁnition of childlessness, whether voluntary (”childfree”) or involuntary (”child-

less”), has been vigorously debated for many decades among demographers and sociol-

ogists. As discussed by Baudin et al. (2015) and Toulemon (1996), deﬁning voluntary

and involuntary childlessness in a non-disputable way is almost impossible. It may lead

to potentially weak statistical analysis in contexts where the availability of attitudi-

nal data is limited. Instead, we use the concepts of poverty and opportunity-driven

childlessness, a phenomenon which has been studied by Baudin et al. (2015) as well as

Baudin et al. (2019). Before a complete deﬁnition of these concepts is provided in the

next sections, let us say that we decompose the group of deﬁnitively childless women

into three categories: women suﬀering from an innate inability to reproduce (natural

sterility), women who are childless because of factors related to poverty (poverty-driven

childlessness), and women who are childless because economic opportunities led them to

make decisions leading to childlessness (opportunity-driven childlessness). As poverty

and economic opportunities can be proxied by educational attainment, it gives these

concepts of childlessness an empirical counterpart.

One of the main predictions of our theory is that if opportunity-driven childlessness

exists in India, the relationship between the probability of a woman ending her repro-

ductive life childless and her level of education is U-shaped. Using micro-level regres-

sions, we validate the existence of such a U-shaped relation. This is indicative that

poverty (proxied by low educational attainment) and sterility are not the sole gradients

of childlessness; better economic opportunities and empowerment within couples (prox-

ied by high educational attainment) also determine the probability of being childless.

We show that this result is robust to the introduction of important control variables

and potential confounders, such as the development level of the state where women

live, the husband’s education, age at marriage, religion, and caste. We do not limit our

analysis to simple associations between women’s educational attainment and the prob-

ability of being childless; we identify causal relationships using some unique features

of the District-Level Household and Facility Survey. India seems to be joining a list of

countries where adjustments to childlessness are much more than simple responses to

boom-and-bust poverty.

The rest of the paper is divided into four sections. Section 2 builds a theory of child-

lessness in India. Section 3 presents our data and methodology while section 4 discusses

descriptive statistics for childlessness over time and across space, with special focus on

education and average fertility. Section 5 tests our theoretical predictions using micro-

level data, while Section 6 tests the causal validity of our results. Section 6 concludes.

2 Theoretical Hypothesis

We ﬁrst discuss why we rely on the decomposition of childlessness proposed by Baudin

et al. (2015) rather than on the more usual distinction between voluntary and invol-

untary childlessness. Then, we develop a theoretical framework adapted to the speciﬁc

context of India. From this will emerge a set of hypotheses which we will test in the

subsequent sections. These hypotheses appear in bold in Subsection 2.2.

2.1 Decomposing Childlessness

The literature about childlessness has extensively discussed the issue of voluntariness

and involuntariness. Exploring National Survey of Family Growth (NSFG) data on

American women, Poston and Cruz (2017) discuss the alternative methods to separate

childless women between those who are voluntarily and involuntarily childless. They

explain why none of these methods are perfect; they also point to the need for intensive

personal interviews to determine the underlying motivations for childlessness. Such

interviews should provide details about women’s reproductive health, their personal

aspirations, and their entire matrimonial history. This need has also been highlighted

by Veevers (1972) who, at the time, mentioned the absence of accurate, representative

datasets. To the best of our knowledge, such datasets still do not exist either in the

speciﬁc case of India or in any other country.5

The absence of accurate data leads to the impossibility of decomposing childlessness

into its voluntary and involuntary components in India and, more importantly, it does

not allow to evaluate the socio-economic gradient of the latter. To circumvent this

diﬃculty and to be able to understand the determinants of deﬁnitive childlessness in

India, we use the decomposition proposed by Baudin et al. (2015). They decompose

childlessness into three modalities: natural sterility, poverty-driven childlessness, and

opportunity-driven childlessness. Natural sterility refers to the innate inability to give

birth, and is uniformly distributed in the population, see for instance Leridon (2008).

Poverty-driven childlessness refers to women who were not a priori sterile, but failed to

have children because of their poverty. As discussed for instance by Romaniuk (1980),

McFalls (1979) and Frank (1983), one main cause of deﬁnitive childlessness among

the poor in less developed countries lies in their higher degree of exposure to vene-

real diseases and malnutrition. Similar regularities are reported by Retel-Laurentin

(1974), Poston et al. (1985), Ombelet et al. (2008), and Wolowyna (1977). Addition-

ally, in the present day, inegalitarian, poor societies do not oﬀer universal access to

assisted reproduction techniques (ART), which reinforces the socio-economic gradient

of poverty-driven childlessness: the rich can aﬀord expensive ART, while the poor are

excluded from this technology. In the case of India, the low access of poor people to

ART is documented by Rasool and Akhtar (2018), Allahbadia (2013), and Malpani and

Malpani (1992).

Opportunity-driven childlessness refers to women who have delayed motherhood to a

point where having children became either unfeasible or undesirable. One main reason

behind this postponement is the time cost of having children. In the vast majority

of societies, as is the case in India, having children requires, for women much more

than for men, investing time in child-rearing activities to the detriment of labor market

activities. The renunciation of labor market activities is part of the opportunity cost

of having children (Becker, 1981). Another dimension of this opportunity cost lies in

the need to abandon other personal aspirations, whose importance is assumed to be

positively associated with education. This fact is documented by Ghosh (2015) in the

case of Kolkata’s second demographic transition, while Surkyn and Lesthaeghe (2004)

provide a more general discussion.

Let us notice that the decomposition of Baudin et al. (2015) does not ascribe a reason

for being childless to each childless woman; from that point of view, it does not do

better than the classical decomposition into voluntary and involuntary childlessness.

Nevertheless, in contrast to the classical decomposition, Baudin et al. (2015) get rid of

attitudinal concepts and the psychological roots of childlessness6to rely on the measur-

able socio-economic and biological factors of childlessness. Thanks to this conceptual

framework, in the next section, we formulate a theory of childlessness in India. This

theory will lead to micro and macro predictions which can be tested on usual datasets,

like census or survey data. We will be able to determine to what extent childlessness

in India is mainly, if not completely, due to natural sterility and poverty, or whether a

new kind of childlessness is emerging, a childlessness due to improvements in the status

of women within society.

Before developing our theory, one can wonder whether, beyond numbers and identiﬁca-

tion strategies, (attitudinal) signs of modernization, which are compatible with modern

forms of childlessness, can be detected in Indian society, We claim that it is the case,

following the lines of developmental idealism theory. This theory states that individuals

who live in a particularly less developed setup will endorse certain behavioral ideals, as

characterized in developed, Western societies, spread gradually through media, aware-

ness, and education. Some recent works have pointed out the practice of modern forms

of marriage (similar to the British) among ethnic groups in the Darjeeling hills of India

(Allendorf, 2013; Allendorf and Pandian, 2016 and Allendorf and Thornton, 2015) as

examples of developmental idealism. Other works also show that single-child families

are an emerging fertility trend in the country (Basu and Desai, 2016). Also, recent on-

going qualitative projects have described how certain metropolitan (Mumbai, Chennai,

Vadodara, and Pune) Indian working women are giving their careers and personal aspi-

rations higher priority than merely being mothers (Bhambhani and Inbanathan, 2017).

These behaviors may be interpreted as the early signs of the second demographic tran-

sition in which childlessness due to women’s economic opportunities and aspirations

becomes more and more prevalent. Thus, childlessness by delay or choice is not such

an alien idea in the Indian context.

2.2 A Theory of Childlessness for India

In line with Baudin et al. (2015), we argue that beyond natural sterility, childlessness,

at the individual level, is the result of an opposition between poverty and economic

opportunities. This central opposition may lead to a U-shape relationship between

the probability of a woman ending her reproductive life childless and her educational

attainment. In the following subsections, we study this theoretical argument and discuss

how it may be at work in the speciﬁc context of India, taking into account additional

determinants of childlessness.

2.2.1 The Role of Female Education

How do we measure poverty and economic opportunities at the individual level? We

proxy both by educational attainment. Tilak (2002) shows how educational poverty,

as measured by low levels of educational attainment, is one of the main, if not the

sole cause of income poverty. Even if income poverty can in turn amplify educational

poverty, the positive association between them is indisputable. Educational poverty

is also a main factor leading to capability poverty in the sense of Sen and Nussbaum

(1993). In the speciﬁc case of India, Duraisamy (2002) shows that the educational

premium is rather weak for low levels of education (primary), while it is signiﬁcantly

strong for higher levels (secondary and tertiary). They document a decrease in the wage

premium for primary education and an increase for higher levels during the period 1983-

1994. This movement ampliﬁed after the economic reform of 1991, and income/wage

inequality has rocketed in India since then. Kijima (2006) shows that this movement

is mainly due to the increase in the returns to skills and the associated increase in the

demand for skilled labor. These results are also conﬁrmed by Chakraborty and Bakshi

(2016) who show how learning English leads to higher wages. They estimate that,

on average, not learning English during primary grades reduces weekly wages by 68

percent. Using diﬀerent datasets and alternative measurements, Tilak (2007) reaches

the same results. Based on this rich literature, we assume, in the Indian context,

that lacking education leads to poverty but that higher education opens the

set of economic opportunities.

Education is the main engine of poverty and economic opportunities, but the intensities

of these two phenomena oppositely evolve when education attainment increases. At

low levels of educational attainment, poverty is severe, so that an increase in education

strongly reduces the burden of poverty. On the contrary, as explained in the previous

paragraph, it does not increase economic opportunities that much. To increase economic

opportunities signiﬁcantly, an increase in education has to occur in a context where the

person is educated enough. For such individuals, the burden of poverty is weak.

As a consequence, among women who have low levels of education, an increase in

educational attainment tends to reduce the probability of being childless because of

poverty, without signiﬁcantly increasing the probability of being childless because of

better economic opportunities. On the contrary, among women who already have a

high level of education, the risk of deprivations leading to infertility is already minimal,

so that an increase in educational attainment only translates into better economic

opportunities, and thus a higher probability of being childless. Finally, women who have

an intermediate level of education are those who are likely to face the lowest probability

of ending their reproductive lives childless: they are protected against extreme poverty

but do not enjoy the largest sets of economic opportunities. From this part of our

theory and echoing Baudin et al. (2015, 2019), we assume that the education of

women has a U-shape incidence on the probability of being childless.

Let us also point out that higher education levels are linked to higher degrees of per-

sonal aspirations, as explained by Surkyn and Lesthaeghe (2004). As women get more

education and economic opportunities, their status within their couple improves: they

enjoy stronger negotiation power. Following Chiappori (1988), this means that the fam-

ily’s objectives are more aligned with the wife’s aspirations. Related to childlessness, it

means that the wife’s fertility preferences will become more important when deciding to

have children or not. Nevertheless, the direction of the eﬀect remains ambiguous: it is

not given that an Indian woman systematically wants fewer children than her husband,

or that she more often wants to postpone her ﬁrst birth. Nevertheless, we also know

that late marriage is a way to avoid motherhood, or at least motherhood at young

ages. We then assume that because women who have a low level of education

have a weak voice on the marriage market, they cannot marry late in order

to escape motherhood, a strategy highly educated women may decide to

adopt. As becoming a married mother is a way to avoid extreme poverty,

poor women have no incentive to adopt such a strategy.

2.2.2 Complementary Mechanisms

What are the main factors which could mitigate the U-shape relationship between

female education and the probability of being childless? We identify at least four

factors: male education, the Indian caste system, the geographical and institutional

diversity of India, and religion.

The husband’s education does not only reduce poverty, it also shapes the way a couple

values the economic opportunities oﬀered to the wife. Indeed, the husband’s income

allows to reduce the relative opportunity cost of child-rearing activities for women, as

the couple has relatively less to lose when the husband also enjoys a high salary. 7

We therefore assume that for highly educated women, the husband’s education reduces

the incentive to postpone births, which in turn reduces the probability of remaining

childless. On the whole, we then reach the global assumption that male education

reduces a woman’s probability of being childless. Again, negotiation power

inside the family may mitigate this eﬀect. If a man is less inclined toward Westernized

ways of life than his wife, getting more education reinforces his negotiation power,

which reduces the probability of the wife not having children, while the reverse is also

true. We know that an eﬀect transiting through negotiation power exists, but cannot

formally identify its sign. Nevertheless, the economic literature on negotiation power

inside couples (Chiappori and Donni, 2011, Baudin et al., 2015, etc.) always identiﬁes

these eﬀects as second-order eﬀects.

India is a large, culturally and institutionally heterogenous country. India’s diversity lies

in its 29 oﬃcial languages,8its caste system, and its being home to all the major religions

in the world. Studies have documented a north-south divide in the country with respect

to (i) openness to fertility change (Dommaraju, 2009), (ii) how education level and

religious aﬃliation lead to diﬀerent fertility outcomes (Kulkarni and Alagarajan, 2005),

and (iii) how caste diﬀerences lead to a diﬀerential utilization of maternal health care

(Kumar and Gupta, 2015). Even the implementation of national policies diﬀers by state;

this is the case for instance for education policies. In a recent paper, Chakraborty and

Bakshi (2016) document how West Bengal has forbidden English classes in primary

schools and how it aﬀects the well-being of children. This state diversity is also driven

by bio-geographic factors like climate, the intensity of pollution, the types of natural

resources, etc.

Based on these documented empirical regularities, we identify three ways in which

state-level diversities may inﬂuence the probability of being childless. First, state

speciﬁcities contribute to the formation of reproductive norms regarding

the ideal size of families and about childlessness and the status of childless

women.9Second, state speciﬁcities and institutions directly inﬂuence the

reproductive conditions which women face, like malnutrition, sanitation,

protection against venereal diseases, delivery conditions, access to modern

ART, etc. Third, states have the power to change some aspects of educational

policies.

Indian economic development and the improvement of economic opportunities oﬀered

to women cannot hide the fact that India is a patrilineal country where women still

have tremendous pressure to bear a child soon after marriage. In some states, cultural

barriers may annihilate the positive impact of better economic opportunities on the

probability of remaining childless at the end of reproductive life.

Religious aﬃliation is another potential factor inﬂuencing the probability of remaining

childless at the end of reproductive life. Among others, Koropeckyj-Cox and Pendell

(2007) document an eﬀect of religious aﬃliation on attitudes and intentions toward

childlessness. Ram (2005) documents an impact of some religious aﬃliations on deﬁni-

tive childlessness rates in India. These eﬀects are discussed in alternative contexts like

Europe (Sobotka, 2017) and the United States (Abma and Martinez, 2006). We then

hypothesize that religious groups may have speciﬁc attitudes toward child-

lessness. Using the theoretical framework of Goldscheider and Uhlenberg (1969), we

assume that the minority status and pro-natalist values of religious groups

like Catholics and Muslims may drive religious diﬀerentials regarding fertil-

ity and childlessness.

Another important dimension linking culture and social structures in India pertains to

the division of society into castes. Historically, Indian society was divided into four

Varnas or groups - Brahmins (highest in the hierarchy), followed by Kshatriyas and

Vaishyas, with the lowest being Sudras. In post-independence India, the constitution

annulled these previous divisions and established scheduled castes (SC), scheduled tribes

(ST), other backward castes (OBC), and Others, with the aim of acknowledging and

uplifting the marginalized sections of society. According to the Indian Socio-Economic

Caste Census 2011, the SCs comprise 22.7 percent of the population, the STs 8.5 per-

cent, the OBCs 51.1 percent and the Others 17.8 percent. The SCs and the STs (lowest

in the caste hierarchy) are caste groups who were historically the most deprived of

certain basic human rights, lived in extreme poverty, malnutrition and were socially

excluded. The OBCs are also historically backward castes while the general castes

comprise of all the other upper castes. Several studies have shown that caste remains a

strong factor in Indian society and women from the SC and ST groups often experience

the highest burden of social exclusion including educational exclusion, poverty, lowest

maternal health care utilization (Kumar and Gupta, 2015), lack of occupational mo-

bility across generations (Banerji, 2012), high fertility outcomes (Ramesh, 2014), etc.

We then make the assumption that belonging to a caste has a direct impact on

the educational attainment of men and women, as well as on their poverty

status since for a low educational level, the burden of poverty is stronger on

the lowest castes, a phenomenon which disappears for high educational attainment.

From the theory developed above, we can build the causal diagram in Figure 1. It

provides a complete picture of our reasoning.

[Figure 1 about here.]

3 Data and Methodology

We use secondary data from all three rounds of the District-Level Household and Facility

Survey (DLHS, 1998-99, 2004-05, and 2007-08). The DLHS provides cross-sectional,

micro-level data that covers all districts and states in the country, respectively. The

survey was conducted by the Indian Institute for Population Studies (IIPS) Mumbai,

funded by the Ministry of Health and Family Welfare (MOHFW). DLHS data was

chosen for the study because it is one of the most recent sample surveys and the largest

ever conducted in the country, covering almost every district in all the country’s states.

The uniqueness of the dataset also lies in the fact that it is the only one which gives

information about fecundity at the individual level. In addition, the dataset also gives

suﬃcient information about childlessness, marriage characteristics, fertility, and other

socio-demographic characteristics for all categories of respondents.10

In the section presenting our descriptive results, two graphs and maps (Figure 3, left and

Figure 6) were constructed using the Indian census, which is a complete enumeration.

The objective was to complement our description of childlessness in India with the

census at the district level.

We then go one step further by conducting an individual-level analysis using DLHS

data only.11 After deleting missing values, input errors, grouping women into birth

cohorts, selecting only age groups 40 to 49 years old, and successive ﬁltering, we have

a ﬁnal sample size of 158,112 ever-married women born between 1953 and 1968, among

whom 4,725 are childless.12 In our regression models, we study the determinants of

the probability of a woman ending her reproductive life childless, for which we use

information about completed fertility to build the dichotomous variable ’childlessness’.

It takes value 1 if the respondent has no children and 0 otherwise.

We consider two kinds of ﬁxed eﬀects, the ﬁrst is a cohort ﬁxed eﬀect and the second

is a state ﬁxed eﬀect. Eight cohorts were compiled for the study, the oldest being

women born in 1953-54 and the youngest, women born in 1967-68. All 35 states and

union territories in India were considered under the state ﬁxed eﬀects in the models.13

Overall, we run 3 diﬀerent kinds of models; in the ﬁrst and main model (Section 5),

we apply both the cohort and state ﬁxed eﬀects. It means that all the interpretations

we propose are valid in a speciﬁc state and a speciﬁc cohort. In the second and third

models, we apply the two ﬁxed eﬀects separately. It allows understanding the temporal

and geographical aspects of the relationship between childlessness and education (see

Appendix B).

From our causal diagram, it appears that the relationship between education and the

causes of childlessness may be confounded by at least the caste system and spatial

diversity. Following the methodology of Wunsch (2007), we have to control for these

two elements when estimating the relationship between education and childlessness.

In order to take into account potential confounding factors and control variables in a

comprehensive way, we have introduced independent variables stepwise. In the ﬁrst

step, we consider the respondent’s education level grouped in four categories; no ed-

ucation, primary, secondary, and higher. In the second step, we add the husband’s

education level (no education, primary, secondary, and higher), teenage marriage (if

the respondent got married before the age of 18, which is the legal age for marriage

for women in India), and place of residence (rural or urban). In the third step, we

add cultural variables like religious denominations (Hindu, Muslim, Christian, Sikh,

Buddhist, and other ) and caste categories (scheduled caste, scheduled tribe, other

backward caste, general, and other ). In the fourth step, we add the variable ’develop-

ment level of the state’, i.e. states are categorized into most developed, least developed,

and intermediately developed, based on their average years of schooling among women,

using the India census of 2011 estimates. States with an average of more than 7.4 years

of education among women are categorized as developed, states with an average of less

than 5.1 years of education among women are categorized as least developed, and other

states are categorized as intermediate states.

It is important to notice that in our main regressions, observations are not weighted

using the sample weights oﬀered by DLHS waves. This decision comes from the diﬃculty

we faced when trying to gather information about the way weights were computed in

the ﬁrst two waves of the DLHS. For this reason, we suspect that the comparability

of data between waves is not guaranteed when using weights. Nevertheless, the results

of all our regression models considering weighted data instead of unweighted data are

provided in Appendix B.3. The only signiﬁcant change occurs when looking at the

impact of being a Catholic compared to being a Hindu. It has no signiﬁcance when

using weighted data, while it does when using unweighted data. For this reason, it is

reasonable to assert that weighting issues are minor in our study.

Still following Wunsch (2007) but also the econometric literature on causality (see for

instance Heckman, 2008), we know that measuring educational attainment and caste,

and taking into account state dummies will allow us to detect the potential existence

of opportunity-driven and poverty-driven childlessness. Nevertheless, the signs of the

eﬀect of education on the probability of remaining childless do not constitute an ir-

refutable proof that poverty explains the decreasing part of the relationship between

childlessness and education, or that better economic opportunities explain the increas-

ing part. Indeed, some underlying factors may inﬂuence both education and child-

lessness in a way not taken into account in our theory. To explore this possibility and

progress toward a causality analysis, we propose some identiﬁcation checks in Section 6.

Using the information we have, we test two of our main hypotheses: (i) female education

is associated negatively with poverty and positively with economic opportunities, and

(ii) both poverty and economic opportunities are positively related to the probability

of being childless.

4 Descriptive Statistics

In this section, we expose empirical regularities to show that Indian childlessness may

be more than a simple response to infertility and/or sterility issues.

4.1 Patterns of Fertility and Childlessness in India

As explained in the introduction, India has entered its demographic transition. The

trend of deﬁnitive childlessness among women in India over the last three decades ex-

hibits an N-shape as shown in Figure 2.14 Analyzing this in line with childlessness

theories, the rise and the subsequent fall in childlessness could be due to mortality and

poverty reasons. Datt (1998) documents an increase in poverty between 1950 and 1970,

a decrease in the 70s and a resurgence from the 80s onwards. Thus, women who ﬁnished

their reproductive lives between the 80s and 90s directly suﬀered from the stronger bur-

den of Indian poverty during their reproductive lives, hence the increase in childlessness.

On the contrary, women who ﬁnished their reproductive lives after the 90s experienced a

continuous decrease in poverty, leading to a drop in childlessness. Interestingly enough,

since poverty has continuously reduced in the country since the 90s, the second rise in

childlessness since 2001 cannot solely be due to poverty reasons. Something else must

be at play. This phenomenon, as studied in various childlessness theories - for instance

Poston and Trent (1982) - comes with further development, with increasing education

leading to rising aspirations, labor-market entry, and the postponement of marriage

among women, also explained as “opportunity-driven childlessness” by Baudin et al.

(2015).

[Figure 2 about here.]

The literature has widely documented a negative relationship between the fertility of

mothers and income, and thus a positive relationship between the degree of poverty

and fertility (Birdsall et al., 2001). Thus, if childlessness was mainly due to poverty

or infertility, one should ﬁnd either an absence of correlation between childlessness

rates and the average fertility of women in Indian states (only sterility matters in that

case), or a positive correlation between the two (poorer states should have both higher

childlessness and higher fertility). Using state-level data (Figure 2 right), we show

that the correlation between childlessness rates and the average fertility of women is

strongly negative. This goes against the idea that childlessness in India is only due to

infertility or poverty reasons, and is indicative of the possible existence of opportunity-

driven childlessness. However, this does not mean that childlessness is only opportunity

driven in India; outliers like Haryana and Nagaland indicate that both childlessness

and the average fertility of women can be high, which is indicative of a potentially high

prevalence of poverty-driven childlessness as well.

4.2 The Desire for No Children

Even if we dispense with attitudinal concepts in our theory and analysis, some atti-

tudinal data exist in the DLHS. We explore them to check whether women were able

to express reasons for their childlessness which were not related to reproductive issues,

poverty, or violence. We conduct this analysis to again highlight how causes for child-

lessness in India cannot be limited to sterility and poverty. These data echo the facts

we discuss at the end of Subsection 2.1.

We focus here on the variable ’desire for no children’ among women with zero fertility

in India. Figure 3 (left) shows that even if the percentage is low (2.34 percent, 1,157

women), the desire for no children exists in India. At the state level, this desire for no

children is negatively correlated with a fertility problem (center panel) and positively

correlated with deﬁnitive childlessness (right panel). This is indicative of the fact that

the whole of Indian childlessness, as has been conceptualized until now (Ram, 2005

and Sujata Ganguly, 2010), may not be driven solely by infertility issues.15 To put

it more simply, the states with a high desire for no children are also the states where

infertility is low; also, deﬁnitive childlessness is high for most states where the desire

for no children is also high.

[Figure 3 about here.]

4.3 The Education Gradient

In this section, we focus on the educational gradient of deﬁnitive childlessness. In our

sub-sample, 29 percent of women never went to school, while 24.8 percent received pri-

mary education, 38.6 received secondary education, and less than 10 percent of women

have a university degree.16 As indicated in Figure 4, at the country level, childlessness

exhibits a J-shaped relationship with years of schooling. This shape indicates that in

India, as in many other countries (Baudin et al., 2019), above an education threshold

(9 or 11 years of schooling), with an increase in years of education, childlessness among

women tends to increase. This fact is more salient when focusing on the youngest

cohorts as shown in the right panel of Figure 4.

[Figure 4 about here.]

In a cross-state perspective (Figure 5), childlessness rates exhibit only a weak correlation

with average education when all states are considered. Nevertheless, when we consider

only the major states in India, this correlation becomes clearly positive. This suggests

that the states with high levels of education (supposedly the most developed ones) are

also those with higher childlessness, while the states with low education levels are those

with the lower childlessness rates. This supports our hypothesis that opportunity-driven

childlessness does exist in India. Nevertheless, our correlation charts are populated with

outliers like Jharkhand (where average education is low and childlessness is high) and

Haryana (where childlessness is low is spite of high education).

[Figure 5 about here.]

As shown in Figure 6, a similar pattern can be noticed in the district-level maps, which

show that childlessness is higher among women who are highly educated (graduate

and above) than among women with less education in all districts of India. Though the

existence of opportunity-driven childlessness can be expected from the above descriptive

ﬁndings, it is still not clear whether the eﬀect of education is robust to other socio-

economic eﬀects among childless women. This doubt will be ruled out in the next

section using multivariate regression models.

[Figure 6 about here.]

5 Regressions

In this section, we go beyond descriptive statistics, and identify the main determinants

of the probability of a woman ending her reproductive life childless at the individual

level. We use a logistic regression speciﬁcation incorporating state and cohort ﬁxed

eﬀects such that, denoting P(χi= 1|Xi, si, ci) the probability that a woman ifrom

state s, born in cohort c, and characterized by the set/vector of social, economic and

cultural characteristics Xi, ﬁnishes her reproductive life childless (χi= 1), we get:

P(χi= 1|Xi, si, ci) = eXiβ+siγ+ciδ+εi

1 + eXiβ+siγ+ciδ+εi

where εiis a random error term and each element of the triplet {β, γ, δ} ∈ R3. In

Model 1, we regress childlessness with the education level of women. As hypothesized,

we ﬁnd a U-shaped curve after controlling for cohort and state ﬁxed eﬀects. This result

still holds after controlling for background and marital characteristics, and cultural

and ecological variables in subsequent models. These successive results conﬁrm the

main prediction of our theory, stating that the probability of women ﬁnishing their

reproductive lives childless has a U-shape relationship with their level of education.17,18

Let us point out that, thanks to cohort ﬁxed eﬀects, we identify an increasing trend of

childlessness in India over the last decades; see Table 4 in Appendix B.1 for detailed

results.

[Table 1 about here.]

In the second model, we add background variables like the husband’s education level,

place of residence, and teenage marriage. The higher the husband’s education level, the

lower the probability of the woman remaining childless. The gradient is signiﬁcantly

and strongly negative. This means that the husband’s education clearly plays the role of

insurance against poverty-driven childlessness among low educated women, while it re-

duces the opportunity cost of having children for highly educated women. Interestingly

enough, the non-linear shape of the probability estimations coming from our logistic

underlying distribution delivers the following result: among low educated women, in-

creasing education reduces the probability of being childless, and this marginal eﬀect is

reinforced by the husband’s education. On the contrary, for highly educated women, in-

creasing their education increases their probability of being childless, but this marginal

eﬀect is smaller when the husband’s education is high.19

Place of residence does not seem to have any eﬀect on childlessness. Teenage marriage

takes the value one if the respondent got married before the age of 18, which is indicative

of an arranged marriage. In the case of a marriage during teenage years, the chance not

to be childless is twice as high as that for a woman who married later. This reﬂects an

eﬀect of exposure time to the risk of conception. Being involved in a traditional marriage

may also increase the incentive or family pressure to conform to the traditional family

system, in which having children is compulsory. Interestingly enough, we ﬁnd that the

correlation between the respondent’s education and age at marriage is positive, but

this does not prevent education from exerting a U-shape inﬂuence on the probability

of remaining childless.

One may wonder why we have not divided age at marriage into more groups to capture

late entry into marriage. This may be quite important knowing that women who do not

want to have children may choose to delay marriage as much as possible. We test this

alternative in Appendix B.4, and show how our results improve in terms of goodness

of ﬁt. Indeed, women who marry later are likelier to remain childless. Let us point

out that including this ﬁner categorization for the age of entry into marriage makes

the U-shape relationship between education and childlessness disappear. It is replaced

by a decreasing relationship; see Appendix B.4. This comes from the fact that those

women who want to avoid having children by marrying late are the more educated

ones. Nevertheless, all the coeﬃcients and measures of ﬁt of this alternative model are

subject to endogeneity issues and are strongly suspected of being spurious. Indeed,

not wanting children (like not wanting to have a husband) determines both education

decisions (and thus the history on the labor market) and the probability of remaining

childless. This double causality issue may lead the eﬀect of age at marriage to confound

that of education.

This last criticism does not apply when only considering teenage marriage, as that kind

of marriage has been decided by the respondent’s family and not the respondent herself.

The possibility remains that reverse causality between education level and childlessness

exists, as women who did not want children may have focused on studying - we are

aware of this. Nevertheless, even if the causality is reversed, both mechanisms refer

to opportunity-driven childlessness: if a woman did not want children, she was able to

avoid having them by focusing on education and job opportunities.

In the third model, we see that caste has only a limited eﬀect on the probability of

remaining childless in our main regression, as only the ST have a higher probability

of remaining childless compared to the General Caste. Said diﬀerently, everything else

being equal, the SC, OBC, and Other castes seem not to suﬀer from any kind of excess

childlessness compared to the General Caste. If this result is surprising at ﬁrst sight,

it is not once we recall that we control for state ﬁxed eﬀects. Indeed, discrimination

against the lower castes has decreased over time and diﬀers in space, which is captured

by our ﬁxed eﬀects. Said diﬀerently, the ST seem to suﬀer from an extra risk of deﬁnitive

childlessness at any time in any state. Interestingly enough, in Appendix B.2, when we

suppress state and cohort ﬁxed eﬀects, we ﬁnd that women who belong to the General

Caste are signiﬁcantly less childless than women who belong to any of the other castes.

The geographical and temporal aspects of the discrimination against lower castes is

now included in the caste variable fully.

Turning to religion, we ﬁnd that both Catholics and Muslims are less likely to remain

childless than Hindus. As explained in our theoretical section, this could be due to the

pro-natalist aspects of these religions, as well as to their minority status in the country.

In the fourth model, we add the development level of the state, which is the state-speciﬁc

average years of schooling among women. In order to not introduce multicollinearity

issues, we have suppressed the state ﬁxed eﬀect from that model. The ﬁnding is in line

with the macro model of Poston and Trent (1982), as well as with those of Baudin et al.

(2019). It shows that education at the micro level has an impact in itself on the prob-

ability of being childless, and this eﬀect is reinforced by a macro eﬀect of development,

proxied by average education at the state level, on the individual probability of being

childless.

From Model 3, we learn that compared to a woman who has completed secondary

education, a woman who never went to school is 1.21 times likelier to ﬁnish her fertile

life childless. A woman who has some years of college is 1.38 times likelier to be

childless than a woman who has a secondary education level. As an alternative to state

ﬁxed eﬀects, we have tested models with state-level ecological variables like the average

childlessness rate and average fertility. We ﬁnd (see Appendix B.5) that the higher the

childlessness rate in a state, the higher the individual probability of remaining childless.

The opposite is true for average fertility. This may reﬂect the existence of norms about

family size and childlessness, but also factors related to state speciﬁcities in terms of

fertility. Said diﬀerently, state ﬁxed eﬀects may control eﬀectively for cultural norms

about reproduction and other kinds of ecological diﬀerences, like the prevalence of

venereal diseases or other factors leading to sub-fecundity.

6 Identiﬁcation Checks

In our theoretical framework, we argue that since education makes poverty recede and

economic opportunities oﬀered to women increase, it should have a U-shaped eﬀect

on the probability of remaining childless at the end of a woman’s reproductive life.

In Section 5, we have evidenced such a U-shaped relationship between educational

attainment and the probability of being childless. Nevertheless, we have not evidenced

that this U-shape is really driven by the opposition between poverty and economic

opportunities when education increases. In this section, we identify these mechanisms

using some unique features of DLHS data.

The ideal dataset would oﬀer a precise measure of both poverty and economic oppor-

tunities oﬀered to women for each wave of observations. This is not the case of the

DLHS as all the waves do not oﬀer these two measures, but Wave 3 does. In Wave 2,

ﬁne measures of poverty are also proposed. Regarding Wave 1, only education allows

us to proxy poverty and economic opportunities.

In the ﬁrst step, we use Waves 2 and 3 of the DLHS in order to enrich our main regression

model, namely Model 3 of Section 5, by introducing a measure of wealth/poverty. The

absence of this variable in Wave 1 leads to a reduction of our sample size to 86,711. In

the second wave, the DLHS measures Wealth as ’Household Standard of Living Index,’

which is labelled in three tiers as: Low, Medium, and High. In the third wave, the

DLHS measures Wealth as ’Wealth Index,’ which is labelled in ﬁve tiers as: Poorest,

Second, Middle, Fourth, and Richest. These index scores are measures given by the

DLHS itself and therefore can be considered as household wealth/poverty measures.

For ease of generalization, we combine these two variables from the ﬁrst and second

waves and use three labels: poor, middle, and rich.20

We then introduce the Wealth Index without changing anything else in our model.

What we obtain is that poor women have a much higher probability of being childless

than other women. We thus identify a direct eﬀect of poverty on the probability of

being childless. In addition, the eﬀect of women’s educational attainment has changed

profoundly. Now education has a purely positive eﬀect on the probability of being

childless. Said diﬀerently, when we control for the intensity of the respondent’s poverty,

education does not have a negative eﬀect on the probability of being childless among less

educated women. This negative eﬀect in Model 3 stemmed from the positive association

between education and poverty and, as such, we have identiﬁed the existence of poverty-

driven childlessness in India.21

[Table 2 about here.]

To now introduce a measure of the economic opportunities oﬀered to women, we have

to limit our analysis to Wave 3, making our number of observations shrink to 41,525.

While this number remains large enough, we limit the validity of our results to the

cohorts born between 1959 and 1967. Wave 3 oﬀers a unique variable, ’Occupation,’

which accounts for about 97 occupation categories. However, we divide this variable

into four categories: laborers, low skilled, medium skilled, and high skilled. This gives

us the possibility of running a model in which, instead of proxying poverty and economic

opportunities by education, we can directly confront the eﬀect of poverty and economic

opportunities on childlessness. In Table 3, we provide the complete classiﬁcation used.

Before commenting our results, let us notice that occupation is measured at the time of

the survey, it is then only a proxy for the economic opportunities which have been oﬀered

to women all along their life. It is then true that birth history may have inﬂuenced

the professional history of these women. Nevertheless, this does not contradict the fact

that present economic occupation is strongly linked to economic opportunities enjoyed

in the past.

[Table 3 about here.]

We perform this analysis in Model B in which we do not include educational measures.

We obtain that poverty keeps exerting a negative eﬀect on the probability of being

childless, while highly and medium-skilled women have a higher probability of remaining

childless than their less skilled counterparts. This is salient evidence in favor of our

theory, stating that childlessness in India results from the confrontation of poverty and

economic opportunities oﬀered to women.

In Model C, we reintroduce education under the form of the number of years of school-

ing. As explained in Section 3, the answer rate to questions related to education is

rather low in the DLHS. For this reason, our model with school attainment has only

18,777 observations. This being said, once we reintroduce education in the model

(Model C), variables describing economic opportunities stop having signiﬁcant eﬀects

on the probability of being childless, while educational attainment has a positive and

signiﬁcant eﬀect.22 Knowing that the eﬀect of poverty is left unchanged, this means

that education controls for the same eﬀect as economic opportunities, explaining why

the latter no longer has any eﬀect. Said diﬀerently, we have identiﬁed here that the

positive eﬀect of education on childlessness for high levels of education is due to the

better economic opportunities oﬀered to women. Indeed, if not, education would not

prevent economic opportunities from having an eﬀect on childlessness. This result also

reassures us about the possibility that the positive relationship between the current

economic status and the probability to be childless would suﬀer reverse causality as

education has been acquired before entering marriage in most of the cases.

Finally, let us point out that we validate the opposite eﬀect of economic opportunities

and poverty (Model B) on a sample of women which is larger than the one also oﬀering

data on years of schooling (Model C).23 This indicates that our main results are not

due to a selection bias implying that women answering education questions are selected

in a way which favors the validation of our theory.

7 Conclusion

At about 8 percent, the Indian childlessness rate is not the highest on Earth; this being

said, childlessness concerns more than 12 million women above 40 years of age. This is

clearly an issue in India, and yet it remains underrated and not explored enough. We

have extended the theoretical framework developed by Baudin et al. (2015) to the Indian

context. We show that our main hypothesis holds true: once controlling for micro and

macro factors, the relationship between the probability of remaining childless and a

woman’s educational attainment is U-shaped. This U comes from the opposite eﬀect

that education has on poverty faced by women and the economic opportunities they

may enjoy.

One could argue that very few women are highly educated in India, but this is not

accurate. In highly developed states like Kerala, more than 60% of women aged between

40 and 50 have at least some years of high school, and 7.5% spent some years at

university.24 In the state of Goa, we ﬁnd that more than 75% of women between 40 and

50 have at least completed high school, while more than 15% have a university degree

or some years at university. These states preﬁgure the future of education for Indian

women. The democratization of education can already be diagnosed comparing the

educational attainment of alternative age groups in our sample. For India as a whole,

6.35% of women aged between 40 and 50 went to university, while they are respectively

7.83% and 8.97% in the 30-40 and 25-30 age groups. This movement can be observed

despite some variance in all the states of India. Our results preﬁgure the future of

childlessness in India; it will be less and less poverty and sterility related and more and

more opportunity related.

While this paper is part of a recent literature showing how Indian families are changing

rapidly, it leaves many questions unanswered. We believe that two of them are key: is

celibacy a way to avoid childbirth in modern India? What are the economic, social,

and psychological consequences of remaining childless in India?

References

Abma, J. and Martinez, G. (2006). Childlessness among older women in the united

states: Trends and proﬁles. Journal of Marriage and Family,68, 1045–1056.

Allahbadia, G. N. (2013). Ivf in developing economies and low resource countries:

An overviewl. The Journal of Obstetrics and Gynecology of India,63, 291–294.

Allendorf, K. (2013). Going nuclear? family structure and youg women’s health in

india, 1992-2006. Demography,50 (3), 853–880.

—and Pandian, R. K. (2016). The decline of arranged marriage? marital change

and continuity in india. Population and Development Review,42 (3), 435–464.

—and Thornton, A. (2015). Caste and choice: the inﬂuence of developmental ide-

alism on marriage behavior. American Journal of Sociology,121 (1), 243–287.

Banerji, M. (2012). Fertility as mobility in india: Salience of caste, education and

employment opportunities, department of Sociology, University of Maryland.

Basu, A. M. and Desai, S. (2016). Hopes, dreams and anxieties: India’s one-child

families. Asian Population Studies,12 (1), 4–27.

Baudin, T.,de la Croix, D. and Gobbi, P. E. (2015). Fertility and childlessness

in the us. American Economic Review,105 (6), 1852–1882.

—,—and —(2019). Endogenous childlessness and the stages of development. Journal

of the European Economic Association,Forthcoming.

Becker, G. S. . (1981). A treatise on the family. Cambridge, MA, 30: Harvard

University Press.

Bhambhani, C. and Inbanathan, A. (2017). Womanhood beyond motherhood: ex-

ploring experiences of voluntary childless women, eSocial Sciences Working Papers

n12077.

Bhatt, M. (2010). India’s changing dates with replacement fertility: A review of recent

fertility trends and future prospects. Lecture Series Institute of Economic Growth.

Birdsall, N.,Kelley, A.,Sinding, S. and Sandy Sinding, S. (2001). Population

matters: demographic change, economic growth, and poverty in the developing world.

Oxford University Press.

Chakraborty, T. and Bakshi, S.-K. (2016). English language premium: Evidence

from a policy experiment in india. Economics of Education Review,50, 1–16.

Chiappori, P. A. (1988). Rational household labor supply. Econometrica,56 (1),

63–90.

Chiappori, P.-A. and Donni, O. (2011). Non-unitary models of household behav-

ior: A survey of the literature. In A. Molina (ed.), Household Economic Behaviors,

Springer, New York, pp. 1–40.

Datt, G. (1998). Poverty in india and indian states: an update. FCND Discussion

Paper, International Food Policy Research Institute, (47).

Dommaraju, P. (2009). India’s north-south divide and theories of fertility change.

Journal of Population Research,26.

Duraisamy, P. (2002). Changes in returns to education in india, 1983–94: by gender,

age-cohort and location. Economics of Education Review,21 (6), 609–622.

Firth, D. (1993). Bias reduction of maximum likelihood estimates. Biometrika,80 (1),

27–38.

Frank, O. (1983). Infertility in sub-saharan africa: estimates and implications. Pop-

ulation and Development Review,9, 137–144.

Ghosh, S. (2015). Second demographic transition or aspirations in transition: An

exploratory analysis of lowest low fertility in kolkata, india. Asia Research Centre

working paper 68, pp. 1–42.

Goldscheider, C. and Uhlenberg, P. R. (1969). Minority group status and fertil-

ity. American Journal of Sociology,74 (4), 361–372.

Heckman, J. (2008). Econometric causality. International statistical review,76 (1),

1–27.

Kijima, Y. (2006). Why did wage inequality increase? evidence from urban india

1983-99. Journal of Development Economics,81 (1), 97–117.

King, G. and Zeng, L. (2001). Logistic regression in rare events data. Political Anal-

ysis,9, 137–163.

Koropeckyj-Cox, T. and Pendell, G. (2007). Attitudes about childlessness in

the united states: Correlates of positive, neutral, and negative responses. Journal of

Family Issues,28 (8), 1054–1082.

Kulkarni, P. (2011). Towards an explanation of india’s fertility transition. Lecture

Series George Simon Memorial Lecture.

—and Alagarajan, M. (2005). Population growth, fertility, and religion in india.

Economic and Political Weekly,40.

Kumar, P. and Gupta, A. (2015). Determinants of inter and intra caste diﬀerences

in utilization of maternal health care services in india: Evidence from dlhs-3 survey.

International Research Journal of Social Sciences,4.

Leridon, H. (2008). A new estimate of permanent sterility by age: sterility deﬁned as

the inability to conceive. Population Studies,62 (1), 15–24.

Long, J. S. and Freese, J. (2006). Regression models for categorical dependent vari-

ables using Stata. Stata Press.

Malpani, A. and Malpani, A. (1992). Simplifying assisted conception techniques to

make them universally available-a view from india. Human Reproduction,7, 49–50.

McFalls, J. A. (1979). Frustrated fertility: A population paradox. Population Bul-

letin,34 (2), 3–43.

Mehta, C. R. and Patel, N. R. (1995). Exact logistic regression: theory and exam-

ples. Statistics in medicine,14 (19), 2143–2160.

Ombelet, W.,Cooke, I.,Dyer, S.,Serour, G. and Devroey, P. (2008). Infer-

tility and the provision of infertility medical services in developing countries. Human

Reproduction Update,14 (6), 605–621.

Pew-Research-Center (2015). Childlessness falls, family size grows among highly

educated women.

Poston, D. L.,Briody, J. E.,Trent, K. and Browning, H. L. (1985). Modern-

ization and childlessness in the states of mexico. Economic Development and Cultural

Change,34 (3), 503–519.

—and Cruz, C. E. (2017). Voluntary, involuntary and temporary childlessness in the

united states. Revue Quetelet/Quetelet Journal,4.1.

—and Trent, K. (1982). International variability in childlessness: A descriptive and

analytical study. Journal of Family Issues,3(4), 473–491.

Ram, U. (2005). Childlessness in time, space and social groups and its linkages with

fertility in india, paper presented at the EPC Conference 2006.

Ramesh, P. (2014). An analysis of fertility diﬀerentials among caste groups in andhra

pradesh, working paper number - 1350, Gokhale Institute of Population and Eco-

nomics.

Rasool, S. and Akhtar, O. S. (2018). The huge burden of infertility in india: Are

we crumbling underneath? Global Journal of Reproductive Medicine,5.

Retel-Laurentin, A. (1974). Inf´econdit´e en Afrique noire : maladies et cons´equences

sociales. Masson.

Romaniuk, A. (1980). Increase in natural fertility during the early stages of mod-

ernization: Evidence from an african case study, zaire. Population Studies,34 (2),

293–310.

Sen, A. K. and Nussbaum, M. (1993). The Quality of Life, Clarendon Press, chap.

Capability and Well-Being, pp. 30–53.

Sobotka, T. (2017). Childlessness in Europe: Reconstructing long-term trends among

women born in 1900-1972. Springer.

SRS, R. (2014). Estimates of fertility indicators. Oﬃce of the Registrar Genral and

Census Commissioner, India,3, 30–66.

Sujata Ganguly, S. U. (2010). Trends of infertility and childlessness in india: Find-

ings from nfhs data. Facts, Views and Vision in ObGyn,2, 131–138.

Surkyn, J. and Lesthaeghe, R. (2004). Value orientations and the second demo-

graphic transition (sdt) in northern, western and southern europe: An update. De-

mographic research,3, 45–86.

Tilak, J. (2002). Education and poverty. Journal of Human Development,3(2), 191–

207.

—(2007). Post-elementary education, poverty and development in india. International

Journal of Educational Development,27 (4), 435–445.

Toulemon, L. (1996). Very few couples remain voluntarily childless. Population,1-27.

Veevers, J. E. (1972). Factors in the incidence of childlessness in canada: An analysis

of census data. Social Biology,19, 266–274.

Wolowyna, J. E. (1977). Income and childlessness in canada: A further examination.

Social Biology,24 (4), 326–331.

Wunsch, G. (2007). Confounding and control. Demographic research,16, 97–120.

Notes

1This spatial heterogeneity is presented in detail in Section 3.

2The 40-year-old threshold may be disputable, especially when considering rich countries where the

postponement of the ﬁrst birth is constantly increasing. In the case of a developing country, it is less

disputable as explained in Baudin et al. (2019). Also, see Poston and Cruz (2017) for alternative types

of childlessness at lower ages.

3According to the estimation proposed by Pew-Research-Center (2015).

4Estimation by Eurostat in the case of Belgium and by the CIA in the case of Rwanda.

5In addition to the absence of accurate data, we do not believe that deﬁning voluntary and involun-

tary childlessness in a non-disputable way is possible. For instance, in the sense of Toulemon (1996),

if a woman desired to have children but ”did not meet the right person,” she should be considered

as involuntarily childless. Focusing on attitudinal data, Toulemon (1996) explains that only a limited

number of women remain childless for purely voluntary reasons. In his set-up, a woman is voluntarily

childless if she reports no desire to have children during her whole reproductive life. Is such a deﬁnition

acceptable? In the terms of Becker (1981), we could argue that this woman received some marriage

oﬀers that she rejected voluntarily. If she had accepted one of these marriage oﬀers, she might have

had children. Numerous arguments in favor of one deﬁnition of involuntary childlessness rather than

another may be used without reaching any kind of consensus.

6Like for instance wanting or never wanting to have children.

7Said diﬀerently, let us assume that to raise a child, a woman needs to spend one year out of the

labor force. Let us also assume that a woman earns 100,000 rupees per year. Having a child would

cost 100,000 rupees, whatever the wage of the husband, but it would represent 10% of the household

income if the husband earns 900,000 rupees per year, while it would represent 90% of the household

income if the husband earns 11,111 rupees per year.

8Our dataset does not oﬀer variables about language spoken at home.

9In fact, this is also true for districts and villages/cities, nevertheless, we do not have access to

these geographical scales.

10 Let us notice that single women were not included in all the rounds, and when included, they

were not asked about their number of children. This leaves at least two questions unanswered: (i)

how do single women contribute to childlessness rates in India? and (ii) do the determinants of the

probability of remaining childless at the end of women’s reproductive lives diﬀer between married and

single women?

11We do not use census data to conduct our regression analysis at the micro level because of the

unavailability of individual census data in India.

12A total of 529,817 households and 474,463 ever-married women were covered by the DLHS in the

ﬁrst round, 620,107 households and 507,622 ever-married women in the second round, and 720,320

households and 643,944 ever-married women in the third round of the survey. The age group selection

allows preventing selection bias due to cohort-based mortality after 50.

13The terminology ﬁxed eﬀect has to be understood here as the use of dummy variables controlling

for the cohort of birth and the state of residence of the respondent.

14The advised reader will see that childlessness rates computed by the Indian census are signiﬁcantly

higher than those computed using our dataset. One reason for this diﬀerence comes from a selection

issue: the Indian census computes childlessness rates over the population of women aged over 40,

while we restrict our attention to women between 40 and 44. Our analysis of census data reveals

that there exists a mortality diﬀerential between childless and non-childless women: childless women

survive more than mothers. This may be due to either social and economic reasons or medical reasons,

like the absence of health troubles after delivery.

15 In the literature, Ram, 2005 and Sujata Ganguly, 2010 are among the very few who study infertility

and childlessness and give a comprehensive idea of the Indian context in this regard. However, an

exploration of whether childlessness is solely driven by infertility, sterility, poverty, or opportunity

seems to have been overlooked by Indian demographers.

16Having primary education or secondary education does not mean here that a woman completed

primary or secondary education, but that she had at least some years of this educational cycle.

17Goodness of ﬁt measured by adjusted count R2 is low at ﬁrst sight. It comes from the fact that

being childless is a very rare event in India, which puts the logistic model in a bad position. As

methods like penalized likelihood cannot be used with the high number of observations we have, we

propose an alternative exercise in Appendix B.6. In this exercise, we draw a limited number of non-

childless women randomly in order to diminish the size of their group and thus increase the prevalence

of childlessness. We show that all our results hold in that situation, while adjusted count R2 increases

signiﬁcantly. This kind of issue with very rare events, as well as the impossibility of using penalized

likelihood in our case, is documented for instance by King and Zeng (2001).

18Notice that if we suppress a state or cohort ﬁxed eﬀect, the U-shape relationship between child-

lessness and education remains; see Appendix B.2. We also run logistic regressions with clustered

standard errors; the results remain unchanged.

19Mathematically speaking, denoting female and male education efand emrespectively, one can

verify that for low values of ef,∂2P(χi=1|Xi,si,ci)

∂ef∂ em>0, while for high values of ef,∂2P(χi=1|Xi,si,ci)

∂ef∂ em<0.

20Technically speaking, we have recoded the Wealth Index into three categories: Poor which gathers

Poor and Second, Middle which corresponds to Middle, and Rich which gathers Fourth and Richest.

21Interestingly enough, the husband’s educational level still prevents childlessness, suggesting, as

explained in the theoretical part, that the eﬀect of male education does not transit through poverty

only. Our result regarding female education, poverty, and childlessness (Model A) remains valid even

when not including the husband’s education, which suggests that our results are not confounded by

male education.

22 Also, we did not include the husband’s education in Model B in order to prevent male education

from capturing the eﬀect of female education on both childlessness and economic opportunities for

women. Indeed, like in all countries on the planet, we observe rather strong educational homogamy

in India. In our sample, the Pearson’s χ2between the variables Education and Husband’s Education

equals 1.1∗105, which is indicative of strong educational homogamy. When we test a model including

male education, our main results remain, with the exception that the coeﬃcient attached to the

’laborer’ category is no longer signiﬁcant.

23Indeed, the percentage of women not answering the question related to years of schooling is higher

among those who exert low-skilled activities: 8.37% among the highly skilled, 29.17% among the

medium skilled, 65.35% among the low skilled, and 61,82% among laborers.

24 The data come from our sample.

List of Tables

1 Determinants of childlessness - Unweighted sample . . . . . . . . . . . . 42

2 Determinants of childlessness - Unweighted sample . . . . . . . . . . . . 43

3 List of occupations and their classiﬁcation. . . . . . . . . . . . . . . . . 46

4 Determinants of childlessness - Unweighted sample - Using all the Fixed

eﬀects .................................... 48

5 Determinants of childlessness - Unweighted sample - Only Cohort Fixed

Eﬀects .................................... 50

6 Determinants of childlessness - Unweighted sample - Only State Fixed

Eﬀects .................................... 53

7 Determinants of childlessness - Unweighted sample - No Fixed Eﬀects . 54

8 Determinants of childlessness - Weighted sample - Using Both Fixed Eﬀects 55

9 Determinants of childlessness - Using age of entry into marriage . . . . 56

10 Determinants of childlessness - Using age of entry into marriage . . . . 57

A Data

[Figure 7 about here.]

B Regressions

B.1 With ﬁxed eﬀects, Unweighted sample

Table 4 here shows the details for cohort and state ﬁxed eﬀects from our main regres-

sions. Table 5 delivers our results without State Fixed Eﬀects. Table 6 shows our

results without Cohort Fixed Eﬀects.

[Table 4 about here.]

[Table 5 about here.]

[Table 6 about here.]

B.2 Without ﬁxed eﬀects

Table 7 shows our main results without any kind of ﬁxed eﬀects.

[Table 7 about here.]

B.3 Weighted sample

This table 8 uses individual weights provided by DLHS and both kinds of ﬁxed eﬀects.

[Table 8 about here.]

B.4 Age of entry into marriage

In Table 10, we use the age of entry into marriage instead of only teen-marriage.

[Table 9 about here.]

B.5 Alternatives to state dummies

In this section, we evidence that state-ﬁxed eﬀects may control for cultural or social

norms regarding fertility and childlessness. To do so, we propose two models in which

we replace our state dummies by either the average childlessness or average number of

children of mothers prevailing in the state according to our sample. To alleviate tables,

we only report coeﬃcients for female and male education in addition to the coeﬃcients

attached to our new state variables.

[Table 10 about here.]

Interestingly enough, we can observe that a woman living in a state where average

childlessness is higher has more chances, everything else equal, to be childless than a

woman living in a state where childlessness is small. This result may capture cultural

and social norms regarding childlessness: it is easier to stay childless in environments

where childlessness is more current. Inversely, a woman living in a state where the

average fertility of mothers is high, and so fertility norms are potentially high, has less

chances to be childless.

B.6 Artiﬁcial samples and goodness of ﬁt

As explained in the core of the paper, the low childlessness rates prevailing in our sub-

sample explain the low values of our measures of goodness-of-ﬁt. This problem is known

as the rare event phenomenon in case of small samples, see King and Zeng (2001) and

the blog of Paul Allison for a discussion. In our case, the total sample size is not small

but as we use both state and cohort ﬁxed eﬀects, our sample is supposed to explain the

probability of being childlessness for a given cohort in a given state, what reduces the

size of each population tremendously and makes the number of events (women being

childless) much smaller. Alternative methods of estimation exist, they are supposed to

ﬁx this issue but they have their own problems. For instance, the penalized likelihood

estimation of a logistic model proposed by Firth (1993) suﬀers over-estimation bias of

the coeﬃcient of the regressions. In a paper like ours, it means that we could attribute

meaning to meaningless variables. The exact-logistic regression model proposed by

Mehta and Patel (1995) works only when the number of observations is below 200

because it is too demanding in terms of computing power.

As an alternative, we propose here to build artiﬁcial datasets in which we systematically

keep all childless women while we select the non-childless persons randomly. We then

show that the quality of our ﬁt becomes much more satisfying. To build our three

alternative datasets, we have generated a random variable following a standard normal

distribution. In sample 1, we have kept all the persons who have drawn a value between

-0.5 and +0.5 reducing the sample size to 58603 observations without reducing the

number of childless women. In sample 2 and 4 respectively, we have kept women who

have drawn a value between -0.1 and +0.1 and -0.05 and +0.05; sample sizes then

become 9024 and 6141.

[Table 11 about here.]

If we see that the improvements in Mc Fadden R2are real but limited, the main change

comes from count-R2. The count-R2measures the number of well-predicted cases over

the total number of observations. It is reputed to be spuriously high when the event

to be predicted is very rare because a lack of variance. To ﬁx this bias, one can use

the number of events which are well predicted beyond the largest marginal (the most

common event which is not being childless). Long and Freese (2006) deﬁne this adjusted

measure of ﬁt as follows: “The adjusted count R2 is the proportion of correct guesses

beyond the number that would be correctly guessed by choosing the largest marginal.

One can notice that once we reduce our sample size, our model provide satisfying

performances.

List of Figures

1 Causaldiagram ............................... 34

2The evolution of Indian childlessness over a period of four decades (left, Source:

Census of India 1981, 1991, 2001, and 2011) and Correlation between childlessness

rates and average fertility of mothers at the state-level among birth cohorts of women

aged 40 - 49 years (right, Source: DLHS 1998-99, 2002-04 and 2007-08) ...... 35

3Fertility desire among women (left), Correlation between fertility problem and de-

sire for no children in all Indian states (center), and Correlation between deﬁnitive

childlessness and desire for no children in major Indian states (right). Source: DLHS

1998-99, 2002-04, and 2007-08 ......................... 36

4Childlessness rates by years of schooling among women aged 40-49. Source: DLHS

1998-99, 2002-04, and 2007-08 ......................... 37

5Correlation between childlessness rates and average education across states: including

all states (left) and including only the major Indian states (right) ......... 38

6Childlessness rates among highly educated women (left), overall population (center),

and not educated women (right) at the district level. Source: Census of India 2011 39

7Deﬁnitive Childlessness in rural areas (left) and urban areas (right) in Indian Districts. 40

Female education Female negotiation power

inside marriage

Intensity of poverty

Caste

Female economic

opportunities

Opportunity cost

of children

CHILDLESSNESS

Causes of sub-fecundity

- malnutrition

- venereal diseases

- poor access to MAP

- etc.

Spatial diversity

- State speciﬁc quality of institutions

- State speciﬁc policies

- Bio-geographic conditions

- Aspirations to modernity

- Neighborhood

- etc.

Fertility and childlessness norms

Religion (pro-natalism)

Male Education

+/-

Figure 1: Causal diagram

Figure 2: The evolution of Indian childlessness over a period of four decades (left, Source: Census

of India 1981, 1991, 2001, and 2011) and Correlation between childlessness rates and average fertility

of mothers at the state-level among birth cohorts of women aged 40 - 49 years (right, Source: DLHS

1998-99, 2002-04 and 2007-08)

Figure 3: Fertility desire among women (left), Correlation between fertility problem and desire for

no children in all Indian states (center), and Correlation between deﬁnitive childlessness and desire for

no children in major Indian states (right). Source: DLHS 1998-99, 2002-04, and 2007-08

Figure 4: Childlessness rates by years of schooling among women aged 40-49. Source: DLHS 1998-99,

2002-04, and 2007-08

Figure 5: Correlation between childlessness rates and average education across states: including all

states (left) and including only the major Indian states (right)

Figure 6: Childlessness rates among highly educated women (left), overall population (center), and

not educated women (right) at the district level. Source: Census of India 2011

Figure 7: Deﬁnitive Childlessness in rural areas (left) and urban areas (right) in Indian Districts.

List of Tables

Table 1: Determinants of childlessness - Unweighted sample

VARIABLES Model 1 Model 2 Model 3 Model 4

Education

No education 1.36*** 1.22*** 1.21*** 1.12*

Primary 0.937 0.97 0.97 0.94

Secondary Education Ref. Ref. Ref. Ref.

Higher 1.399*** 1.38*** 1.38*** 1.35***

Husband’s education

No education Ref. Ref. Ref.

Primary 0.73*** 0.74*** 0.78***

Secondary Education 0.66*** 0.66*** 0.66***

Higher 0.54*** 0.54*** 0.54***

Teen marriage

No Ref. Ref. Ref.

Yes 0.54*** 0.54*** 0.58***

Place of residence

Rural Ref. Ref. Ref.

Urban 1 1 1.04

Religion

Hindu Ref. Ref.

Muslim 0.89* 0.805***

Christian 0.74*** 0.821***

Sikh 1.02 0.68***

Buddhist 0.8 0.77*

Others 1.04 1.12

Caste

General Ref. Ref.

SC 1.05 1.08

ST 1.13* 0.98

OBC 0.98 0.89*

Others 1.11 1.14

State development level

Developed States Ref.

Least developed States 0.76***

Intermediate States 0.87***

Fixed eﬀects

Cohort FE YES YES YES YES

State FE YES YES YES NO

Pseudo R2 0.0123 0.0207 0.0213 0.0122

BIC 33585.197 33365.756 33454.349 33374.480

Number of obs. 158112 158112 158112 158112

Count (adj) 0 0 0 0

Notes: Odds-ratio reported. *** p<0.01, ** p<0.05, * p<0.1

Table 2: Determinants of childlessness - Unweighted sample

VARIABLES Model A Model B Model C

Education

No education 0.526* - -

Primary 0.980 - -

Secondary Education Ref. - -

Higher 1.367*** - -

Years of schooling - - 1.034***

Wealth Index

Poor 1.251*** 1.554*** 1.372*

Middle Ref. Ref. Ref.

Rich 0.843*** 1.032 0.965

Occupation

Labourer - 0.759** 1.041

Lowly skilled - 0.679*** 0.877

Medium skilled 0.882 1.051

Highly skilled - Ref. Ref.

Husband’s education

No education Ref. - -

Primary 0.272*** - -

Secondary Education 0.282*** - -

Higher 0.241*** - -

Teen marriage

No Ref. Ref. Ref.

Yes 0.471*** 0.491*** 0.406***

Religion YES YES YES

Caste YES YES YES

Fixed eﬀects

Cohort FE YES YES YES

State FE YES YES YES

Pseudo R2 0.025 0.034 0.042

Number of obs. 86,711 41,525 18,777

Notes: Odds-ratio reported. *** p<0.01, ** p<0.05, * p<0.1

Skill Job designation Population % Cum. %

physical scientists 45 0.02 0.02

HIGHLY SKILLED

physical science 553 0.19 0.21

architects, engineers, technologists an 696 0.24 0.45

engineering technicians 81 0.03 0.48

aircraft and ships oﬃcers 17 0.01 0.48

life scientists 32 0.01 0.49

life science technicians 13 0 0.5

physicians and surgeons 103 0.04 0.53

nursing and other medical and health te 2,613 0.9 1.44

scientiﬁc, medical and technical perso 48 0.02 1.45

mathematicians, statisticians and relat 15 0.01 1.46

economists, and related workers 23 0.01 1.47

accountants, auditors and related worke 145 0.05 1.52

social scientists and related workers 574 0.2 1.72

jurists 70 0.02 1.74

teachers 8,178 2.83 4.57

poets, authors, journalists and related 125 0.04 4.62

sculptors, painters, photographers, and 75 0.03 4.64

composers and performing artists 117 0.04 4.68

professional workers 135 0.05 4.73

elected and legislative oﬃcials 86 0.03 4.76

administrative and executive oﬃcials 903 0.31 5.07

working proprietors, directors and mana 105 0.04 5.11

directors and managers, ﬁnancial insti 54 0.02 5.13

working proprietors, directors and mana 39 0.01 5.14

working proprietors, directors managers 52 0.02 5.16

working proprietors, directors and mana 20 0.01 5.17

MEDIUM SKILLED

administrative, executive and manageria 30 0.01 5.18

clerical and other supervisors 210 0.07 5.25

village oﬃcials 979 0.34 5.59

stenographers, typist and card and tape 95 0.03 5.63

book keepers, cashiers and related work 114 0.04 5.66

computing machine operators 157 0.05 5.72

clerical and related workers 815 0.28 6

transport and communication supervisors 77 0.03 6.03

transport conductors and guards 21 0.01 6.04

mail distributors and related workers 43 0.01 6.05

telephone and telegraph operators 62 0.02 6.07

merchants and shopkeepers, wholesale an 3,372 1.17 7.24

manufacturers, agents 185 0.06 7.3

technical salesmen and commercial trave 25 0.01 7.31

salesmen, shop assistants and related w 3,201 1.11 8.42

insurance, real estate, securities and 556 0.19 8.61

money lenders and pawn brokers 106 0.04 8.65

sales workers 717 0.25 8.91

hotel and restaurant keepers 398 0.14 9.05

house keepers, matron and stewards dome 613 0.21 9.26

cooks, waiters, bartenders and related 1,751 0.61 9.87

maids and related house keeping service 3,527 1.22 11.09

building caretakers, sweepers, cleaners 746 0.26 11.35

launderers, dry-cleaners and pressers 531 0.18 11.53

LOW SKILLED

hair dresser, barbers, beauticians and 461 0.16 11.69

protective service workers 112 0.04 11.73

service workers 505 0.17 11.92

farm plantation, dairy and other manage 471 0.16 12.08

cultivators 73,941 25.61 37.69

farmers, other than cultivators 12,960 4.49 42.17

agricultural labourer 95,037 32.91 75.09

plantation labourers & related workers 1,573 0.54 75.63

other farm workers 1,288 0.45 76.08

forestry workers 856 0.3 76.38

hunters and related workers 125 0.04 76.42

ﬁshermen and related workers 364 0.13 76.55

LABOURERS

miners, quarrymen, well drillers & related 160 0.06 76.62

metal processors 90 0.03 76.65

wood preparation workers and paper make 217 0.08 76.73

chemical processors and related workers 57 0.02 76.75

spinners, weavers, knitters, dyers and 3,236 1.12 77.87

tanners, fellmongers and pelt dressers 35 0.01 77.88

food and beverage processors 590 0.2 78.08

tobacco preparers & tobacco product maker 4,652 1.61 79.7

tailors, dress makers, sewers, upholste 9,101 3.15 82.85

shoemakers & leather goods makers 92 0.03 82.88

carpenters, cabinet & related wood work 111 0.04 82.92

stone cutters & carvers 304 0.11 83.02

blacksmiths, tool makers and machine to 32 0.01 83.03

machinery ﬁtters, machine assemblers a 18 0.01 83.04

electrical ﬁtters & related electrical 54 0.02 83.06

broadcasting station and sound equipmen 8 0 83.06

plumbers, welders, sheet metal 20 0.01 83.07

jewellery & precious metal workers 129 0.04 83.11

glass formers, potters & related worker 273 0.09 83.21

rubber and plastic product makers worker 69 0.02 83.23

paper & paper board products makers 111 0.04 83.27

painters 1,237 0.43 83.7

stationery engines and related equipmen 252 0.09 83.79

transport equipment operators 85 0.03 83.82

labourers 40,012 13.86 97.67

other new workers seeking employment 2,981 1.03 98.71

MISSING

none workers not reporting any occupation 2,849 0.99 99.69

dont know 581 0.2 99.89

missing 309 0.11 100

Total 288,742 100 -

Table 3: List of occupations and their classiﬁcation.

Variable Model 1 Model 2 Model 3 Model 4

See main results in Table 2.

Cohort FE

1953-54 Ref. Ref. Ref. Ref.

1955-56 0.96 0.98 0.98 1.05

1957-58 1.02 1.04 1.04 1.12

1959-60 1.19 1.2 1.20* 1.30* *

1961-62 1.09 1.11 1.12 1.23*

1963-64 1.262** 1.27** 1.28** 1.39**

1965-66 1.07 1.07 1.07 1.15

1967-68 1.3 1.30** 1.31** 1.42***

State FE

Jammu and Kashmir Ref. Ref. Ref. .

Himachal Pradesh 1.29 1.54** 1.45*

Punjab 1.29 1.42** 1.32

Chandigarh 1.55 1.91 1.78

Uttarakhand 1.24 1.55* 1.46

Haryana 0.76 0.98 0.93

Delhi 1.44 1.83*** 1.72**

Rajasthan 1.63*** 2.19*** 2.07***

Uttar Pradesh 1.51*** 2.15*** 2.08***

Bihar 1.60*** 2.35*** 2.26***

Sikkim 1.35 1.42 1.42

Arunachal Pradesh 2.22*** 2.41*** 2.27***

Nagaland 3.12*** 3.49*** 3.79***

Manipur 2.38*** 2.66*** 2.81***

Mixoram 2.60*** 2.82*** 3.12***

Tripura 1.73** 1.82*** 1.88***

Meghalaya 1.34 1.54 1.52**

Assam 2.11*** 2.6*** 2.46***

West Bengal 2.56*** 3.43*** 3.24***

Jharkhand 2.65*** 3.67*** 3.47***

Odisha 1.77*** 2.26*** 2.13***

Chhattisgarh 2.14*** 3.12*** 2.93***

Madhya Pradesh 1.82*** 2.48*** 2.34***

Gujarat 1.84*** 2.23*** 2.09***

Daman and Diu 1.7 2.08** 1.94*

Dadra and Nagar Haveli 1.92*** 2.77*** 2.66***

Maharashtra 2.18*** 2.93*** 2.81***

Andhra Pradesh 2.45*** 3.35*** 3.20***

Karnataka 3.20*** 3.92*** 3.78***

Goa 2.56*** 2.96*** 2.92***

Lakshadweep 1.84*** 2.10*** 2.08***

Kerala 2.89*** 3.36*** 3,38***

Tamil Nadu 2.72*** 3.30*** 3.23***

Pondicherry 3.11*** 3.66*** 3.61***

Andaman and Nicobar Islands 0.64 0.63 0.7

Notes: Odds-ratio reported. *** p<0.01, ** p<0.05, * p<0.1

Table 4: Determinants of childlessness - Unweighted sam-

ple - Using all the Fixed eﬀects

Variables Model 1 Model 2 Model 3 Model 4

Education

Secondary Education Ref. Ref. Ref. Ref.

No education 1.23*** 1.12* 1.1 1.12

Primary 0.91** 0.92 0.92 0.94

Higher 1.33*** 1.35*** 1.37*** 1.35***

Husband’s Education

No education Ref Ref. Ref. Ref.

Primary 0.80*** 0.79*** 0.79***

Secondary Education 0.66*** 0.66*** 0.66***

Higher 0.54*** 0.54*** 0.54***

Teen Marriage

No Ref. Ref. Ref. Ref.

Yes 0.60*** 0.58*** 0.58*** 0.58***

Place of Residence

Rural Ref. Ref. Ref. Ref.

Urban 1.06 1.06 1.04

Religion

Hindu Ref. Ref. Ref. Ref.

Muslim 0.81*** 0.81***

Christian 0.81*** 0.82***

Sikh 0.62*** 0.68***

Buddhist 0.72** 0.77*

Others 1.09 1.12

Caste

General Ref. Ref. Ref. Ref.

SC 1.34** 1.11*

ST 1.17** 1.21***

OBC 1.15*** 1.10**

Others 1.3* 1.27

State Development Level

Developed States Ref. Ref. Ref. Ref.

Least developed States 0.87***

Intermediate States 0.76***

Fixed Eﬀects

State FE NO NO NO NO

Cohort FE YES YES YES YES

1953-54 Ref. Ref. Ref. Ref.

1955-56 1.01 1.08 1.06 1.05

1957-58 1.09 1.16 1.14 1.12

1959-60 1.29** 1.35*** 1.32** 1.30**

1961-62 1.24* 1.29** 1.27** 1.23*

1963-64 1.41*** 1.46*** 1.43*** 1.39***

1965-66 1.17 1.22 1.19 1.15

1967-68 1.47*** 1.53*** 1.48*** 1.42***

Pseudo R2 0.0022 0.0094 0.0111 0.0122

BIC 33518.786 33337.379 33385.512 33744.48

Number of observations 158112 158112 158112 158112

Count (adj) 0 0 0 0

Notes: Odds-ratio reported. *** p<0.01, ** p<0.05, * p<0.1

Table 5: Determinants of childlessness - Unweighted sam-

ple - Only Cohort Fixed Eﬀects

Variables Model 1 Model 2 Model 3

Education

Secondary Education Ref. Ref. Ref.

No education 1.23*** 1.12* 1.11

Primary 0.93 0.97 0.97

Higher 1.40*** 1.38*** 1.38***

Husband’s Education

No education Ref Ref. Ref.

Primary 0.75*** 0.75***

Secondary Education 0.67*** 0.67***

Higher 0.55*** 0.55***

Teen Marriage

No Ref. Ref. Ref.

Yes 0.54*** 0.53*** 0.53***

Place of Residence

Rural Ref. Ref. Ref.

Urban 1 1

Religion

Hindu Ref. Ref. Ref.

Muslim 0.9

Christian 0.74***

Sikh 1.02

Buddhist 0.8

Others 1.06

Caste

General Ref. Ref. Ref.

SC 1.06

ST 1.14**

OBC 0.99

Others 1.07

State Development Level

Developed States Ref. Ref. Ref.

Least developed States

Intermediate States

Jammu and Kashmir Ref. Ref. Ref.

Himachal Pradesh 1.3 1.56** 1.49**

Punjab 1.28 1.42** 1.33

Chandigarh 1.61 2 1.89

Uttarakhand 1.29 1.61** 1.54*

Haryana 0.75 0.98 0.93

Delhi 1.48* 1.89*** 1.80***

Rajasthan 1.62*** 2.19*** 2.09***

Uttar Pradesh 1.51*** 2.16*** 2.10***

Bihar 1.61*** 2.37*** 2.30***

Sikkim 1.36 1.43 1.44

Arunachal Pradesh 2.27*** 2.47*** 2.31***

Nagaland 3.07*** 3.46*** 3.74***

Manipur 2.40*** 2.69*** 2.84***

Mixoram 2.67*** 2.90*** 3.19***

Tripura 1.70** 1.80*** 1.86***

Meghalaya 1.31 1.52** 1.50**

Assam 2.12*** 2.63*** 2.51***

West Bengal 2.64*** 3.56*** 3.41***

Jharkhand 2.59*** 3.62*** 3.44***

Odisha 1.84*** 2.36*** 2.25***

Chhattisgarh 2.13*** 3.12*** 2.96***

Madhya Pradesh 1.81*** 2.48*** 2.36***

Gujarat 1.88*** 2.29*** 2.17***

Daman and Diu 1.75 2.13** 2.01*

Dadra and Nagar Haveli 1.82*** 2.65*** 2.56***

Maharashtra 2.20*** 2.98*** 2.88***

Andhra Pradesh 2.41*** 3.32*** 3.21***

Karnataka 3.30*** 4.05*** 3.94***

Goa 2.52*** 2.94*** 2.91***

Lakshadweep 1.73** 1.99*** 1.98***

Kerala 2.89*** 3.38*** 3.41***

Tamil Nadu 2.78*** 3.40*** 3.34***

Pondicherry 3.17*** 3.75*** 3.73***

Andaman and Nicobar Islands 0.62* 0.62* 0.68

Pseudo R2 0.0115 0.02 0.0205

BIC 33528.758 33306.037 33395.361

Number of observations 158112 158112 158112

Count (adj) 000

Notes: Odds-ratio reported. *** p<0.01, ** p<0.05, * p<0.1

Table 6: Determinants of childlessness - Unweighted sam-

ple - Only State Fixed Eﬀects

Table 7: Determinants of childlessness - Unweighted sample - No Fixed Eﬀects

Variable Model 1 Model 2 Model 3 Model 4

Education

Secondary Education Ref. Ref. Ref. Ref.

No education 1.07 0.98 0.97 1

Primary .90** 0.92* 0.92* 0.93

Higher 1.33*** 1.36*** 1.38*** 1.36***

Husband’s Education

No education Ref Ref. Ref. Ref.

Primary 0.82*** 0.81*** 0.80***

Secondary Education 0.67*** 0.67*** 0.67***

Higher 0.54*** 0.54*** 0.55***

Teen Marriage 0.60*** 0.57*** 0.58***

Religion

Hindu Ref. Ref. Ref. Ref.

Muslim 0.82*** 0.81***

Christian 0.80*** 0.82***

Sikh 0.62*** 0.67***

Buddhist 0.73** 0.78***

Others 1.11 1.2

Caste

General Ref. Ref. Ref. Ref.

SC 1.15** 1.12*

ST 1.19** 1.22***

OBC 1.17*** 1.11**

Others 1.25 1.2

State Develoment Level

Developed States Ref. Ref. Ref. Ref.

Least developed States 0.86***

Intermediate States 0.75***

Pseudo R2 0 0.0083 0.0102 0.0113

BIC 33473.264 33289.534 33334.48 33320.313

Number of observations 158112 158112 158112 158112

Count (adj) 0 0 0 0

Notes: Odds-ratio reported. *** p<0.01, ** p<0.05, * p<0.1

Table 8: Determinants of childlessness - Weighted sample - Using Both Fixed Eﬀects

Variable Model 1 Model 2 Model 3 Model 4

Education

Secondary Education Ref. Ref. Ref. Ref.

No education 1.40*** 1.39*** 1.40*** 1.40***

Primary 0.92 1 1 1

Higher 1.38*** 1.37*** 1.36*** 1.36***

Husband’s Education

No education Ref. Ref. Ref. Ref.

Primary 0.72*** 0.72*** 0.72***

Secondary Education 0.72*** 0.72*** 0.72***

Higher 0.57*** 0.57*** 0.56***

Teen Marriage

No Ref. Ref. Ref. Ref.

Yes 0.51*** 0.51*** 0.51***

Place of Residence

Rural Ref. Ref. Ref. Ref.

Urban 0.99 1 1

Religion

Hindu Ref. Ref. Ref. Ref.

Muslim 0.80*** 0.80***

Christian 0.91 0.91

Sikh 1 1.01

Buddhist 0.81 0.81

Others 1.12 1.12

Caste

General Ref. Ref. Ref. Ref.

SC 0.99 0.99

ST 1.1 1.1

OBC 1.01 1.01

Others 1.1 1.1

State Development Level

Developed States Ref. Ref. Ref. Ref.

Least developed States 0.71

Intermediate States 0.28***

Fixed Eﬀect

Cohort FE YES YES YES YES

State FE YES YES YES YES

Pseudo R2 0.0126 0.0215 0.0219 0.0219

BIC 45199.105 44859.052 44946.209 44357.917

Number of observations 158112 158112 158112 158112

Count (adj) 0.011 0.019 0.019 0.019

Notes: Odds-ratio reported. *** p<0.01, ** p<0.05, * p<0.1

Table 9: Determinants of childlessness - Using age of entry into marriage

Variable Model 1 Model 2 Model 3 Model 4

Education

Secondary Education Ref. Ref. Ref. Ref.

No education 1.40*** 1.41*** 1.42*** 1.42***

Primary 0.92 1.06 1.06 1.06

Higher 1.38*** 1.02 1.02 1.02

Husband’s Education

No education Ref. Ref. Ref. Ref.

Primary 0.74*** 0.74*** 0.74***

Secondary Education 0.72*** 0.72*** 0.72***

Higher 0.59*** 0.58*** 0.58***

Teen Marriage

Less than 18 years Ref. Ref. Ref. Ref.

18 to 29 years 1.68*** 1.68*** 1.68***

30 years and above 25.05*** 25.12*** 25.13***

Place of Residence

Rural Ref. Ref. Ref. Ref.

Urban 0.97 0.98 0.98

Religion

Hindu Ref. Ref. Ref. Ref.

Muslim 0.82** 0.82**

Christian 0.85 0.85

Sikh 1.04 1.04

Buddhist 0.81 0.81

Others 1.2 1.2

Caste

General Ref. Ref. Ref. Ref.

SC 1 1

ST 1.11 1.11

OBC 1.01 1.01

Others 1.13 1.13

State Development Level

Developed States Ref. Ref. Ref. Ref.

Least developed States 0.83

Intermediate States 0.33***

Fixed Eﬀects

Cohort FE YES YES YES YES

State FE YES YES YES YES

Pseudo R2 0.0126 0.0706 0.0711 0.0711

BIC 45199.105 42648.85 42735.172 42735.172

Number of observations 158112 158112 158112 158112

Count (adj) 0 0.068 0.068 0.068

Notes: Odds-ratio reported. *** p<0.01, ** p<0.05, * p<0.1

Table 10: Determinants of childlessness - Using age of entry into marriage

Variable Alternative 1 Alternative 2

Education

Secondary Education Ref. Ref.

No education 1.24*** 1.17**

Primary 0.98 0.94

Higher 1.37*** 1.36***

Husband’s Education

No education Ref. Ref.

Primary 0.75*** 0.78***

Secondary Education 0.67*** 0.67***

Higher 0.56*** 0.56***

Average childlessness 11.60***

Average fertility of mothers 0.820***

Teen Marriage YES YES

Place of Residence YES YES

Religion YES YES

Caste YES YES

Fixed Eﬀects

Cohort FE YES YES

State FE YES YES

Notes: Odds-ratio reported. *** p<0.01, ** p<0.05, * p<0.1

Benchmark regressions With age at marriage

Measure of Fit Sample 1 Sample 2 Sample 3 Sample 1 Sample 2 Sample 3

Age at marriage NO NO NO YES YES YES

McFadden R20.024 0.034 0.036 0.077 0.084 0.080

Count R20.944 0.645 0.590 0.946 0.692 0.616

Adj-Count R20.000 0.023 0.121 0.028 0.153 0.175

Note: Variables of regressions omitted for sake of space and clarity.

Rising trend of childlessness in China: analysis of social and regional disparities with 2010 and 2020 census data

Article

Full-text available

May 2023

Introduction In recent decades, there has been a significant increase in childlessness. This paper analysed childlessness in China, specifically examining its socio and regional disparities. Methods With data from China’s 2020 population census, supplemented with data from China’s 2010 population census and 2015 inter-censual 1% population sample survey, we used a basic indicator of age-specific childlessness proportion, a decomposition method, and probability distribution models to analyse, fit and project childlessness. Results We presented age-specific childlessness proportions for women as a whole and by socioeconomic features, decomposition and projection results. The childlessness proportion increased markedly from 2010 to 2020, reaching 5.16% for women aged 49. The proportion is highest for city women, followed by township women, and is lowest among village women, at 6.29%, 5.50% and 3.72 % for women aged 49, respectively. The proportion for women aged 49 with high college education or above was 7.98%, and only 4.42% for women with junior high school education. The proportion also exhibits marked provincial discrepancies, and the total fertility rate is negatively correlated with childlessness at the province level. The decomposition results distinguished the different contribution of change in educational structure and change in childlessness proportion for subgroups to the total childlessness proportion change. It is projected that city women, women with high education will have higher childlessness proportion, and the proportion will further increase with the rapid increase in education level and urbanisation. Conclusions Childlessness has risen to a relatively high level, and varies among women with different characteristics. This should be taken into consideration in China’s countermeasures to reduce childlessness and curtail further fertility decline accordingly.

Examining Spatial Heterogeneity and Potential Risk Factors of Childlessness Across 412 Districts of India: An Analysis of 4 Decades

Article

Full-text available

Oct 2023

Adrita Banerjee

There has been a rise in childlessness in India in the last couple of decades. The rate of increase differs remarkably across districts. However very little is known about the spatial variation of childlessness. Thus using Census of India data the study explores the district level variation in prevalence of childlessness and its determinants over time. Univariate Moran’s I and LISA statistics have been used to confirm the spatial clustering and dependence in childlessness. The Ordinary Least Square (OLS) and Spatial (lag/error) models were employed to examine the effects of correlates on district level childlessness. The results indicate spatial clustering across districts in all the census rounds. High–high cluster i.e. hotspots are observed in the southern states of Andhra Pradesh and Tamil Nadu and low–low clusters along the northern and north–western part of the country. The spatial OLS, SLM and SE model identified the major factors affecting childlessness as literacy rate, age at marriage, percentage of urban population and Scheduled Tribe population. Owing to the importance of childlessness further research should focus on exploring the association of childlessness and fertility in Indian context.

Childlessness and Economic Development: A Survey

Chapter

Full-text available

Nov 2019

This chapter shows why, beyond average fertility, childlessness matters in itself. Childlessness reacts to economic incentives in a peculiar way, which cannot be described by usual economic models of fertility. We make a distinction between childlessness due to poverty and childlessness due to economic opportunities. It allows to better understand the dynamics of childlessness along the history but also why the extensive margin of fertility (childlessness) does not adjust to economic shocks and development policies in the same way as the intensive margin (number of children of mothers). Introducing marriage into this framework provides new insights: higher educational homogamy per se decreases childlessness as it favors marriage. Nevertheless, because educational homogamy is most of the time accompanied by a rise in female education, it may also be associated with increases in childlessness in the data.

Lowest-Low Fertility in India

Chapter

Jan 2024

Koyel Sarkar

The chapter discusses the rising trend of ultra-low fertility, i.e., the prevalence of having either no children or a single child among women in India. It highlights the influence of education levels, sociocultural groups, and regional factors on childlessness and single-childness. The analysis shows that female education increases the likelihood of having no or one child, with childlessness following a J-shaped relationship and single-childness having a positive linear relationship. Economic opportunities and poverty play significant roles in determining fertility preferences, and religion, caste, and tribe indicators have varying effects. The study identifies two main factors behind childlessness and single-child motherhood: higher socioeconomic status and cultural environment specific to different regions in India. The findings suggest that the increasing proportion of women with one or no child has implications for India’s demographic growth and future fertility trends, which may be influenced by economic opportunities and educational attainment among women.

Examining a non-conformist choice: The decision-making process toward being childfree couples

Article

Jul 2020

The norm of parenthood in a heteronormative union expects individuals to become parents as a preordained milestone in their life-course. This norm emanates from the ideology of pronatalism that promotes procreation for the well-being of the individual, family and society. The present paper is set against varying yet converging manifestations of pronatalistism in India and Canada, as evident in the dominant discourses on family forms, and policies that reinforce the hegemony of the heteronormative nuclear family. Drawing on the conjoint narratives of thirty-six childfree couples based in cities of India and Canada, we explore the diverse processes that ultimately result in the decision to forego parenthood. The cross-cultural similarities and dissimilarities among various decision making pathways in turn reveal the complex interplay between individual agency and the overarching pronatalist habitus. This interpretation of the decision-making process to forego parenthood contributes to an understanding of non-conformist choices that are made.

The Huge Burden of Infertility in India- Are We Crumbling Underneath?

Article

Full-text available

Aug 2018

Sabahat Rasool

Childlessness in Europe: Reconstructing Long-Term Trends Among Women Born in 1900–1972

Chapter

Full-text available

Jan 2017

Tomas Sobotka

This study combines census, survey, and vital statistics data to reconstruct childlessness trends among women born between 1900 and 1972 in 30 European countries. After discussing the available sources, their strengths and weaknesses, and the selection of data used in this article, I present childlessness trends in individual countries and broader European regions, and analyse between-country variability. Europe has experienced a U-shaped pattern in permanent childlessness among women born between 1900 and 1972, with the lowest levels typically reached among the 1940s cohorts. A clear contrast exists between the generally low levels of childlessness in the former state-socialist countries of Europe and the higher levels of childlessness in other regions. However, as childlessness in central and eastern Europe has increased among women born in the late 1960s and early 1970s, this long-standing divide is now eroding. The trend towards increasing childlessness is levelling off in much of western Europe (e.g., in Switzerland and England and Wales). In conclusion, I suggest that women in southern Europe, especially in Italy and Spain, are on track to have the highest levels of childlessness in Europe. I also argue that childlessness trends might become more unstable and more difficult to predict as many women have put off parenthood until late reproductive ages, when they have relatively little time left to have a child.

Second demographic transition or aspirations in transition: an exploratory analysis of lowest-low fertility in Kolkata, India

Article

Full-text available

Aug 2016

Ghosh Saswata

Fertility transition occurred during the 1970s in Kolkata (erstwhile Calcutta), capital of the state of West Bengal, India, and has remained persistently at the low fertility level. Fertility rate in Kolkata attained its lowest-low level in the beginning of the present century. It currently has the lowest fertility rate (TFR 1.2) in India. This could be a case of second demographic transition (SDT) or pertains to changing dynamics in childbearing, childrearing, and/or aspirations for children. Using primary data of 600 couples (1200 individuals) and employing quantitative and qualitative methods, this study found that constraints in childbearing and childrearing; and aspirations for children have a strong negative and significant effect on second and higher order childbearing among couples, particularly among women. In our study there was no strong evidence of voluntary childlessness as well as decline in the importance of marriage, family and children as posited by SDT. Thus we argue that SDT might have to be redefined for a developing country context.

Modernization and Childlessness in the States of Mexico

Article

Full-text available

Apr 1985

Begins by hypothesizing that, among the 32 states of Mexico in 1970, levels of modernization should be negatively related to childlessness. Since childlessness is mainly involuntary in developing countries such as Mexico, increasing modernization should directly and indirectly affect subfecundity, which is the more proximate cause of involuntary childlessness because of its influence in reducing disease and nutritional deficiencies. Sets forth a number of modernization variables reflecting structural economic development and health conditions, and hypothesizes that the health variables should be more highly associated with childlessness in the expected direction at the bivariate level. Then factor analyzes the input variables, obtains four factors, one of which clearly reflects health conditions; the other three are identified as various dimensions of structural economic development. When childlessness is regressed on these four modernization factors, the health conditions dimension is shown to have the highest standardized regression coefficient. The analyses support both the general hypothesis that modernization is related inversely to childlessness among the states of Mexico and the more specific hypothesis that the health conditions dimension of modernization is more influential than factors dealing mainly with structural economic development. -from Authors

Fertility and Childlessness in the United States †

Article

Full-text available

Jun 2015

We develop a theory of fertility, distinguishing its intensive margin from its extensive margin. The deep parameters are identified using facts from the 1990 US Census: (i) fertility of mothers decreases with education; (ii) childlessness exhibits a U-shaped relationship with education; (iii) the relationship between marriage rates and education is hump-shaped for women and increasing for men. We estimate that 2.5 percent of women were childless because of poverty and 8.1 percent because of high opportunity cost of childrearing. Over time, historical trends in total factor productivity and in education led to a U-shaped response in childlessness rates while fertility of mothers decreased.

Trends of Infertility and Childlessness in India: Findings from NFHS Data

Article

Full-text available

Mar 2010

Etiology of infertility varies from region to region and from one population to another and even from one locality to another within the same population. Childlessness has serious demographic, social and health implications. Hence an attempt has been made to get some approximation about levels and patterns of infertility and childlessness in India by using National Family Health Survey-2 (1998-1999) and National Family Health Survey-3 (2005-2006) data. The study population consists of women aged 20-49 years married for ≥ 5 years. Age of women, age at first marriage, place of residence, standard of living, working status of women, and region are some of the variables related with the rate of infertility and childlessness. Infertility rate is high among women in urban areas. This may be due to lifestyle or a later age at first marriage. Considering religion, Muslims show the lowest infertility rate. Scheduled tribes have high infertility rate. With increasing levels of educational attainment among women, infertility rate increases. This can be related to the fact that with aspirations for attaining higher educational level, marriage is delayed as a result of which in confirmation with aforementioned causation factors (higher age at marriage, urban living style etc.), infertility rate is high among this sub group of population.

IVF in Developing Economies and Low Resource Countries: An Overview

Article

Full-text available

Oct 2013

Gautam Allahbadia

Hopes, Dreams and Anxieties: India's One-Child Families

Article

Jan 2016

While rapid fertility decline in India in the last two decades has received considerable attention, much of the discourse has focused on a decline in high parity births. However, this paper finds that, almost hidden from the public gaze, a small but significant segment of the Indian population has begun the transition to extremely low fertility. Among the urban, upper income, educated, middle class, it is no longer unusual to find families stopping at one child, even when this child is a girl. Using data from the India Human Development Survey of 2004–2005, we examine the factors that may lead some families to stop at a single child. We conclude that the motivations for this very low fertility are likely to be a more extreme form of those for low fertility rather than reflecting the qualitative change in ideologies and worldviews that is hypothesized to accompany very low fertility during the second demographic transition.

Infécondité en Afrique Noire. Maladies et conséquences sociales

Article

May 1975

A. Retel-Laurentin

Determinants of Inter and Intra caste Differences in Utilization of Maternal Health Care Services in India: Evidence from DLHS-3 Survey

Article

Jan 2015

This paper examines utilization of maternal health care services within and across social groups among married women in India using data from third round of District Level Household and Facility Survey conducted during 2007-08. Maternal health care utilization is measured through full antenatal care, safe delivery and postnatal care. Besides, selected socioeconomic and demographic factors have been included as predictor variables. Bi-variate, logistic regression and concentration curve have been employed to understand the inter-caste differences, net effect of the predictor variables on selected outcomes and intra-caste differentials respectively. Our findings show significant difference in the utilization of maternal health care services by caste, women’ age at first birth, educational attainment, place of residence, economic status and region. Besides, high inequality is found among poor and non-poor in ‘Other’ Caste followed by Other Backward Classes. This paper concludes that Scheduled Tribe and Scheduled Caste married women bear the multiple burdens of social exclusion, poverty and womanhood in patriarchal Indian society. As a result, their conditions are worse on utilizing maternal health care services which could lead to higher maternal and child mortality rate.

Education and Childlessness in India *

Recommended publications

Éducation au développement durable : accompagner l’émergence d’une attitude collaborative

Entrepreneurship and Innovation Education: Frameworks and Tools

Trends and Challenges in Management Education around the world

Educating transcultural managers for open environments and organizations