Rural–Urban Fertility Convergence,Differential Stopping Behavior,and Contraceptive Method Mix in West Bengal,India: A Spatiotemporal Analysis

Abstract

The present study attempts to understand how the interplay between differential stopping behavior and contraceptive use dynamics may have resulted in the convergence of rural-urban fertility in West Bengal, India. Using data from the National Family Health Surveys and employing sequential logit regressions, we were able to confirm that differential stopping behavior, particularly among the rural women—by adopting contraception at first parity after having a boy, or choosing modern methods over natural methods at second parity after having two successive boys—may be one of the factors that led to the rural-urban convergence of fertility.

Keywords

rural-urban fertility convergence differential stopping behavior contraceptive method mix national family health surveys sequential logit regressions West Bengal

Introduction

Fertility transition in India, which was initiated during the late 1960s, has recently gained momentum. During the last decade, in particular, the transition has covered almost all parts of the country to varying degrees.¹ The western, north-western, and, especially the southern states of the country have achieved below replacement-level fertility (Total Fertility Rate (TFR) < 2.1), whereas the northern, central and some of the north-eastern states are yet to achieve replacement-level fertility.^2,3 Differences in the levels of fertility exist not only among states or between major and minor districts (in terms of development indicators) of a state but also between rural and urban areas.^4,5 The statistical reports of the Sample Registration System (SRS) of India reveal that TFR declined by nearly 28 percent between 1985–1986 and 2000–2001 and by 25 percent between 2001–2002 and 2015–2016. The decline was faster in urban areas compared to rural areas between 1985–1986 and 2000–2001 (30 percent in urban areas against an approximate 24 percent in rural areas); however, it was the other way round between 2001–2002 and 2015–2016 (18 percent in urban areas against about 27 percent in rural areas).⁶

Studies have pointed out several reasons for the decline in TFR, including increase in women’s education, in employment in various wage-earning sectors activities, and in mass media exposure.^7–

11 For example, in the state of Kerala, increase in female education was directly linked with fertility decline, while in other south Indian states, particularly exemplified by Tamil Nadu, mass media exposure and backward class movement were responsible for the same.^{12–

19} In the central and northern states, however, fertility continues to be high due to patriarchy, farm-based economy, and lack of women’s autonomy and empowerment.^20,21 Thus, apart from a few exceptions, the states of India can broadly be classified into two geographical regions in terms of fertility—the northern states having high fertility and preference for sons and the southern ones with relatively low fertility.²²

The eastern state of West Bengal—a middle-ranking state in terms of human development indicators, with more than 68 percent of rural population—does not fall into the “popular north-south demographic dichotomy.”^23,24 The performance of West Bengal in terms of fertility decline seems exceptional during the last four decades, given the slower pace of fertility decline in the other states of eastern India, namely, Bihar, Jharkhand, Assam, and Odisha.²⁵ In the rural areas, only Odisha, out of the four adjacent states of West Bengal, reached replacement-level fertility as recently as during 2014–2016. Urban TFRs are low in each of these states, except Bihar (Table 1). However, a recent study by Haque et al. (2019) depicts that there is no significant spatial dependence in the district-level TFR of West Bengal with the proximate districts of the adjacent states; rather it is surrounded by the high-TFR districts of east Bihar, Sikkim, and some districts of Jharkhand.²⁶

Table 1.

Differences in Rural–Urban Fertility Rates in West Bengal, India and Other Adjacent States; SRS, Various Years.

Year	Total fertility Rate (TFR)
	West Bengal		Odisha		Jharkhand		Bihar		Assam		India
	Rural	Urban	Rural	Urban	Rural	Urban	Rural	Urban	Rural	Urban	Rural	Urban
1990–1992	3.6	2.0	3.4	2.4	NA	NA	4.7	3.4	3.6	2.1	4.0	2.7
1993–1995	3.3	2.0	3.3	2.5	NA	NA	4.7	3.0	3.7	2.4	3.8	2.7
1996–1998	2.8	1.8	3.2	2.3	NA	NA	4.5	3.1	3.4	2.1	3.6	2.4
1999–2001	2.7	1.6	2.8	2.1	NA	NA	4.7	3.3	3.3	1.8	3.5	2.3
2002–2004	2.5	1.6	2.7	2.0	3.9	2.3	4.4	3.1	3.1	1.7	3.3	2.2
2005–2007	2.2	1.4	2.6	1.7	3.7	3.0	4.3	3.0	3.0	1.5	3.1	2.0
2008–2010	2.1	1.3	2.5	1.6	3.4	3.0	3.9	2.8	2.8	1.6	2.9	2.0
2011–2013	1.8	1.2	2.2	1.5	3.0	2.0	3.6	2.5	2.5	1.5	2.6	1.8
2014–2016	1.7	1.2	2.1	1.5	2.9	2.0	3.3	2.5	2.4	1.6	2.5	1.8

Source: Calculated from compendium of India’s fertility table and SRS statistical report, various years.

Note:

• Data on TFR was not available for Jharkhand until 2003.

• From 1999 onward, TFRs of Bihar have been calculated excluding Jharkhand.

• Excludes Jammu & Kashmir due to non-receipt of returns from 1991 to 1995.

• Estimates of vital rates at the national level up to 1995 do not include Mizoram as the SRS was not operational in Mizoram till 1995.

• Excludes Nagaland (Rural) due to part-receipt of returns from 1995 to 2003.

The demographic behavior of West Bengal has not received adequate attention among demographers, where fertility decline began well before independence.²⁷ Currently, West Bengal is one of the lowest fertility states in India. Available literature suggests that a secular decline in birth rate had started in this state in the mid-1950s.^28–
30 Notably, the decline in TFR in rural West Bengal was 25 percent between 1990–1992 and 1999–2001, which became remarkably faster (33 percent) between 1999–2001 and 2014–2016. Table 1 and Figure 1 show that although rural women had 1.6 children more on an average compared to urban women during 1990–1992, the gap reduced to 1.1 children during 1999–2001 and further to 0.5 during 2014–2016 in West Bengal.³¹

Figure 1.

Rural-urban convergence in fertility differences, West Bengal and India; SRS, various years.

Unlike West Bengal, the rural-urban gap in fertility in India as a whole was 1.3 children during 1990–1992, which reduced marginally to 1.2 children during 1999–2001 and further to 0.7 children by the end of 2014–2016.³² This implies that the reduction in the gap between rural and urban fertility in West Bengal has been much faster compared to the national level. This remarkable decline during the past two decades, apparently, has taken place without meeting standard pre-conditions of fertility transition; rather it has occurred in seemingly unfavorable conditions.³³ Table 2 presents some of the important socio-demographic and maternal health indicators by rural-urban residence for West Bengal and India.

Table 2.

Some Important Demographic and Health Indicators of West Bengal and India.

Demographic and health indicators	West Bengal			India
Demographic and health indicators	Total	Rural	Urban	Total	Rural	Urban
Population, 2011 (‘000s)^a	91.28	62.18	29.09	1,210.85	833.74	377.10
Population density, 2011 (persons/km²)^a	1,028	744	5,684	382	277	4,834
Sex ratio, 2011 (female/1000 male)^a	950	953	944	940	947	926
Percentage population aged 7+ that is literate, 2011^a	67.42	62.92	76.42	73.0	67.8	84.1
Percentage female population aged 7+ that is literate, 2011^a	62.33	57.24	73.28	31.4	28.3	38.2
Child (0–6 years) sex ratio, 2011^a	956	959	947	914	919	902
Life expectancy at birth (in years)^b	71.2	70.2	73.4	69.0	67.7	72.4
Crude birth rate^c	15.5	17.3	11.6	20.8	22.4	17.4
Infant mortality rate, 2016^c	25	25	22	34	38	23
Total fertility rate, 2015–2016^c	1.6	1.7	1.2	2.3	2.5	1.8
Maternal mortality ratio^d	113	—	—	167	—	—
Percentage women who began child bearing between age 15–19^e	18.3	20.6	12.4	7.9	9.2	5.0
Percentage mothers who had at least 4 ANC for last birth^e	76.5	75.8	78.1	51.2	44.8	66.4
Percentage skilled attendance at delivery^e	81.6	79.0	88.5	81.4	78.0	90.0
Percentage institutional delivery^f	75.2	72.0	83.6	78.9	75.1	88.7
Percentage mothers who received skilled postnatal care within two days of delivery^f	61.1	58.1	68.5	62.4	58.5	71.7

Source: Calculated from compendium of India's fertility table and SRS statistical report, various years.

^a Census of India and West Bengal, 2011.

^bSRS-based Abridged Life Table, 2013–2017.

^c Sample Registration System Bulletin 2016.

^d SRS; MMR Bulletin,2011–2013.

^e4th Round of National Family Health Survey Report, 2015–2016.

^f4th Round of National Family Health Survey; Factsheet, 2015–2016.

Works Which Have Led up to the Study

Studies on rural-urban differentials in fertility patterns in India began long ago.^34,35 Kleinman (1973) noted that differences in institutional factors, such as age at marriage, cultural patterns in conjugal relations, attitudes toward divorce and widow remarriage, economic constraints, and other socio-cultural factors, may have significantly influenced the observed fertility rates.³⁶ Contemporary works have found that fertility transition in India started in urban areas and then “diffused” to rural areas through social imitation and adaptation.^37–

40 However, the relationship between urbanization and fertility decline in India has been found to be weak.⁴¹ One can argue that starting from similar fertility levels, rural areas experienced a sharp increase in excess fertility over urban areas in the first phase because of earlier and faster decline in cities, but reached fertility levels comparable to urban areas in the later phase of the demographic transition.⁴²

Reliable estimates of the birth rates in West Bengal were exceptionally deficient before about 1979.⁴³ Some earlier studies on the fertility dynamics of West Bengal observed that in Kolkata (erstwhile Calcutta), among the upper class and educated people, the use of contraception was reasonably high even for women aged 40–44 at the time of survey during 1947–48.⁴⁴ MariBhat (1996) showed that Kolkata had the lowest total fertility rate (TFR), that is 2.0, in the country in the late 1970s, while the rate was between 4 and 5 in the predominantly rural districts of West Bengal. The study found that even during 1984–1990, the fertility rates in the rural districts remained almost at the same level or had declined only marginally.⁴⁵ Thus, it seems that the decline in fertility was essentially concentrated among urban “elites” and had taken place a long time ago, that is during the 1970s, while rural areas continued to have high fertility even in the late 1980s.

Drawing upon the experience of the European fertility transition⁴⁶ and the “innovation-diffusion” explanations,^47–
49 Basu and Amin (2000) argued that a unique combination of historical, cultural, and political factors in the Bengal region has resulted in a unique combination of Bengali “modernism” and Bengali nationalism, which have facilitated the development of a secular society, open to change and innovation whenever the opportunities for such change or innovation were available. The study argued that this willingness to change had begun with the urban “elites” but that it spread onto the rural masses due to a great deal of rural-urban interactions through which information, attitudes, and ideology were disseminated. The Bengali language played an important role in this entire dynamics.⁵⁰ Using period parity progression ratios (PPPRs), Paul and Kulkarni (2006) concluded that the age at first birth had not increased among women even after a marginal rise in the age at marriage. However, fertility decline in the state took place particularly during the 1990s when most women started to have at least two children but increasingly avoided subsequent pregnancies over the period.⁵¹ Thus, it seems that the hypothesis of the “elite-led” diffusion process in changing fertility norms does not hold good for West Bengal; on the contrary, the rural-urban interaction has been weak in this state.⁵² It, therefore, seems that the urban-rural time lag in the onset of the transition and its differential pace cannot be explained entirely by the social diffusion theory.

Studies have further argued that the family planning program that was expected to expedite the popularization of the small family norm has not been very effective whether in rural or in urban areas possibly due to the indifference of the Government of West Bengal in population matters.^53–
55 It is important to note that though the acceptance and usage of contraception has been higher in West Bengal compared to the other Indian states⁵⁶ and that the use of modern family planning methods has increased substantially over the years in rural areas, the use of the natural method of family planning persists at a high rate in the rural areas of the state (13 percent).⁵⁷

In a recent study, Chatterjee (2019) found that, in West Bengal, the glaring differences in rural-urban fertility rates, observed in earlier decades, have started to converge recently. The study also found that the decline may be attributed to the fact that more than four out of ten women in rural areas have stopped childbearing before attaining the “desired” number of children. By employing an untenable ordinary least square (OLS) regression model at the district-level for 19 districts of West Bengal, the study concluded that contraceptive methods, female education, and age at marriage have had a greater influence on rural fertility rates compared to the urban ones.⁵⁸ Another study conducted by Haque and Patel (2016) revealed that the practice of contraception was higher among “ruralites” and among couples who had a strong preference for sons.⁵⁹ However, these studies have failed to explore whether differential stopping behavior (DSB),⁶⁰ manifested through the preference for sons, could be one of the reasons for the remarkable mismatch between the desired number and the achieved number of children in rural West Bengal. In the demographic literature, differential stopping behavior (DSB) is a kind of male-preferring stopping rule which reflects a strong son preference and son-targeting fertility behavior.

Taking a cue from these studies, after describing the trends in district-level fertility from 1991 onward, the primary objective of the present study is to understand whether the preference for sons (or DSB), as manifested through contraceptive adoption and choice, has resulted in the contemporary fertility decline in rural West Bengal. It is postulated that DSB in rural West Bengal, as manifested by the choice of modern contraceptive methods over natural methods once the desired number of sons is achieved, may be one of the important factors behind the contemporary rural fertility decline in West Bengal.

Materials and Methods

Data and Variables

Data for the present research were obtained from all the four rounds of the National Family Health Survey (NFHS), the Indian variant of the Demographic and Health Surveys (DHS), conducted so far. NFHS is a large-scale, nationally representative, multi-round survey conducted to provide information on fertility, maternal and child health, reproductive choices, family planning services, etc. for India. The main purpose of each successive round of the NFHS is to highlight the emerging socio-demographic issues to inform policymakers. The four rounds of the survey were carried out between 1992–1993 and 2015–2016 by IIPS, Mumbai (1992–1993, round 1),⁶¹ IIPS and ORC Macro (1998–1999, round 2),⁶² IIPS and Macro International Inc. (2005–2006, round 3),⁶³ and IIPS and ICF (2015–2016, round 4).⁶⁴ Data for West Bengal in all the rounds were filtered to carry out the analyses. The present analysis is restricted to the currently married women aged fifteen to forty-nine years who had at least one living child and were non-menopausal at the time of the survey. The NFHS-1 (1992–1993) interviewed currently married women in the age bracket of thirteen to forty-nine years. Therefore, a subsample of 3,998 (weighted 37,260) individual women aged fifteen to forty-nine years was extracted from the first round. From the subsequent rounds conducted in 1998–1999, 2005–2006, and 2015–2016, information of 4,408 (weighted 41,932), 4,973 (weighted 49,453), and 2,208 (state-level data, weighted 68,025) women respectively (as covered in the state-level sample), aged fifteen to forty-nine years, were collected. State-level samples were used for the fourth round of the survey because of the following two reasons. First, it was appropriate to use state-level samples for comparisons with the earlier rounds. Secondly, information on women’s work and husband’s age and occupation were available only in the state-level samples. Thus, a total of 15,587 (weighted 196,661) currently married women from West Bengal, aged fifteen to forty-nine years, having at least one living child and non-menopausal at the time of the survey, were selected from the four successive rounds.

Data concerning the district-level TFR from 1991 to 2011 were drawn from secondary sources—the IIPS (2005)⁶⁵ and Guilmoto and Rajan (2013).⁶⁶

The bivariate analysis and t-ratios for rural and urban areas were carried out separately from the individual-level data by clubbing the first two rounds and the last two rounds, considering the state weight, for the ease of interpretation. This was done to highlight the discernible changes in fertility preferences—as indicated by the difference in the actual and the ideal number of boys and girls, contraceptive adoption, and choice according to the gender-parity composition—in rural and urban areas between 1992–1999 and 2005–2016. However, while carrying out the multivariate analyses, all the four rounds were pooled together and each round identified separately so that the independent effect of time (as represented by the year of the survey) could also be distinguished.

As our main objective was to find out whether DSB (or son-preferring fertility behavior) has any bearing on the contraceptive method mix in the contemporary fertility decline in the rural areas, the dependent variable was current contraceptive use. The contraceptive use was divided into the following three categories: couples not using any method; couples using natural/behavioral methods (folkloric methods, periodic abstinence/rhythm/safe period, and withdrawal), and couples using modern methods (IUD, pill, injectables, implants, Norplant, condom (including female condom), and female and male sterilization). A similar categorization was employed by Ghosh and Begum (2015),⁶⁷ Ghosh and Chattopadhyay (2017),⁶⁸ and Ghosh et al. (2020)⁶⁹ in their studies. For identifying the changes in the preference for contraceptive methods (Table 3), female and male sterilization were categorized separately as permanent methods (sterilization). For the multivariate analyses, like earlier studies, we made no distinction between modern temporary and permanent methods of contraception. This is so because our hypothesis was that couples prefer to use behavioral contraceptive methods transitorily; once they achieve the desired number of sons, they shift to the modern methods (temporary or permanent). Furthermore, the distinction between modern temporary and permanent methods at the third transition of multivariate analyses seemed implausible and non-concave because of substantial differences regarding the choice between these methods, particularly in the earlier rounds of NFHS (Appendix Table A1).

Table 3.

Contraceptive Preferences, Number of Living Children, Number of Living Sons, and Want of Children among Respondents Aged Fifteen to Forty-nine by Place of Residence in Various NFHS Rounds, West Bengal.

NFHS Rounds	Current use of types of contraceptive method (in percent)				Mean number of living children	Mean number of living sons	Not wanting future children (in percent)
NFHS Rounds	Not using any method	Natural/Behavioral	Modern Reversible method	Sterilization	Mean number of living children	Mean number of living sons	Not wanting future children (in percent)
NFHS1
Rural	39.3	18.1	5.5	37.1	3.1	1.6	48.8
Urban	33.6	26.0	11.9	28.6	2.7	1.4	65.5
NFHS2
Rural	31.0	15.8	13.2	40.0	2.5	1.3	73.3
Urban	21.4	27.4	19.4	31.7	2.1	1.1	80.1
NFHS3
Rural	25.6	19.2	17.0	38.3	2.4	1.2	74.9
Urban	19.4	26.2	21.3	33.1	1.9	1.0	79.3
NFHS4
Rural	20.6	15.6	29.9	33.9	2.0	0.9	77.5
Urban	27.9	18.4	29.7	24.1	1.7	0.7	78.9

Source: Calculated from unit level data of NFHS (Women’s file), four rounds (1992–1993 to 2015–2016).

For the present study, we considered two primary predictor variables: the sex composition of children in a specified parity and the four time periods (1992–1993, 1998–1999, 2005–2006, and 2015–2016) for rural and urban areas separately. For gender-parity composition, women were categorized by the number and sex composition of living children at each parity as follows: parity 1 (0 son, 1 son), parity 2 (0 son, 1 son, 2 sons), parity 3+ (0 son, 1 son, 2 sons, 3 or more sons).

The control variables were age and age-square of respondents (continuous); education of respondents (continuous); wealth index⁷⁰ (poorest, poorer, middle, richer, and richest as calculated by NFHS); work-status of respondent’s partner⁷¹ (not-working, primary, secondary, tertiary sector activities); exposure to family planning messages in any mass media during the month preceding the survey⁷² (no/yes); religious affiliation of respondents (Hindu, Muslim/others); caste of the respondents (SC, ST, and others); and degree of participation in household decision-making of respondents⁷³ (continuous; computed except in the first round of NFHS). The intention to have another child (no/yes) was controlled in the multivariate model as a covariate. For details of the computation procedure of the variables, please refer to the notes (70–73) below. Variables such as marital duration, partner’s age and education, and respondents’ work status were considered at the initial stage; however, they were dropped later to avoid multicollinearity.

Analytical Model

The study used descriptive statistics, t-ratios, Pearson correlations, and multivariate models. Descriptive statistics were produced by place of residence to understand the variations in sample characteristics. Pearson correlations were used to understand the association between the relative changes of TFR and the proportion of the urban population at the district-level of West Bengal. Parity-wise contraceptive usage by place of residence was computed to understand gross-differentials in son preference and contraceptive method mix according to the place of residence. t-ratios were used to understand whether gross differentials were statistically significant or not. t-ratio is the ratio of the mean of the difference to the standard error of the difference (t-statistics). Two-sample t-tests with equal variances were performed, considering the state weight, using the svy: mean command.

The present study applied the sequential logit (SL) model (Buis 2007)⁷⁴ while studying the net differentials in the contraceptive method mix with respect to son preference in the rural and urban areas of West Bengal. The model estimates the effect of the explanatory variables on the probabilities of passing through a set of transitions. In the present analysis, women’s decision regarding whether to use a method or not was referred to as the first transition, which indicates contraceptive adoption. This choice was modeled using a conventional logit model. Only those women who had decided to use a method were taken into consideration in the second transition. Their choice of method (indicating contraceptive choice) was modeled using another conventional logit model.^75–
77

Suppose that there are J alternatives, which are divided into H sub-choice sets, A₁ , A₂ ,…., A_H . The choice process of an individual could be divided into two transitions such that, in Transition 1, an individual chooses one of the H sub-choice sets, or A_h for some h, while in Transition 2, the individual chooses the alternative j∈A_h , that is,

P r (y \in A_{k}) = \frac{e x p (\dot{X} δ_{h})}{\sum_{u = 1}^{K} e x p (\dot{X} δ_{u})}, for h = 1,…H

P r (y = j | A_{h}) = \frac{e x p (\dot{X} α_{h})}{\sum_{k \in A_{h}}^{H} e x p (\dot{X} α_{k})},

where, $δ_{h} \in ℝ^{k}$ , h = 1, 2,…, H and $α_{j} \in ℝ^{k}$ , j = 1, 2,…, J.

It is possible to identify the model with normalization $δ_{h}$ = 0 and $α_{j h}$ = 0 ∀ h, where j_h is the first element in A_h . This model can be derived in the context of “utility maximizing behavior” by defining the utilities for each choice separately, at the first stage set A_h and at the second stage for each alternative j∈A_h .^{78, 79} It is worth noting that testing the property of Independence of Irrelevant Alternatives (IIA) was not required as in the case of the SL model. Estimates of the adjusted odds ratios (AOR) and the average marginal effects (AME) were used in the interpretation. Data were analyzed using Stata SE 15. In all the analyses, the sample weights, as calculated by the surveys and provided in the datasets, were used.

Results

Trends and Differentials of District-level Fertility Rates and Percentage of Urban Population in West Bengal

Table 4 shows the estimated relative change in TFR⁸⁰ and the percentage of the urban population from 1991 to 2011 by districts of West Bengal. Although we considered absolute change, with 14 of 19 districts having achieved below replacement-level fertility in 2011, in relative terms, the decline was greater during 1991–2001 in comparison to 2001–2011. During 1991–2001, the highest proportionate decline in fertility (measured in terms of TFR) was observed in Kolkata district (−54.8 percent), while during 2001–2011, it was observed in Nadia district (−29 percent). During 1991–2001, the lowest relative fertility decline was found in Birbhum (−21.1 percent), while during 2001–2011, it was observed in Purulia (−12.9 percent). It may be worth to point out that a recent study by Chatterjee (2019) found that, in 2011, out of 19 districts, rural areas of 10 districts reached below replacement-level fertility, while urban areas of 15 districts achieved it. The study also showed that the rural-urban fertility differentials were higher (0.4) among those districts which were on the verge of attaining replacement-level fertility during the last decade.⁸¹

Table 4.

Relative Change in Total Fertility Rate (TFR) and Percentage of Urban Population in West Bengal, District Level, 1991 to 2011.

Districts	Relative Change (In percentage)^a
	TFR¹		Percentage of urban population²
	1991–2001	2001–2011	1991–2001	2001–2011
Darjiling	−40.0	−23.8	6.1	21.9
Jalpaiguri	−28.2	−25.0	9.0	53.5
Koch Bihar	−26.8	−23.3	16.6	12.9
West Dinajpur^b	25.7	−25.6	−5.7	4.0
Maldah	−20.0	−27.5	3.5	85.5
Murshidabad	−28.6	−22.9	19.8	57.9
Birbhum	−21.1	−23.3	−4.6	49.7
Barddhaman	−36.1	−21.7	5.3	8.0
Nadia	−35.1	−29.2	−6.0	30.9
North 24Parganas	−36.4	−23.8	6.0	5.5
Hugli	−31.0	−20.0	7.3	15.2
Bankura	−25.7	−19.2	−11.1	13.0
Puruliya	−24.4	−12.9	6.7	26.5
Haora	−41.7	−14.3	1.6	25.9
Kolkata	−54.8	−14.3	100	100
South24 Parganas	−38.8	−26.7	18.2	62.6
Medinipur^c	−29.7	−23.1	3.9	16.5
West Bengal	−29.0	−22.2	4.8	30.6

Source: ¹Calculated from IIPS (2005) and Guilmoto and Rajan (2011).

²Census of India, PCA, 1991, 2001 and 2011.

Note: ^aRelative change is the proportionate change between two successive (given) data points (Years).

^b West Dinajpur consists of Uttar Dinajpur and Dakshin Dinajpur.

^c Medinipur consists of Purba Medinipur and Paschim Medinipur.

Average relative change of percentage of urban population was calculated by excluding Kolkata.

Table 4 also reveals an inconsistent and cascading pattern of urbanization and fertility decline in West Bengal. Negative, but statistically insignificant correlations, were found between relative change (in percent) in TFR and percentage of urban population (between 1991 and 2001, Pearson’s r was −0.35, while it was −0.34 between 2001 and 2011). One can observe that during 1991–2001, districts with higher proportionate increase in urban population (in relative terms)—such as Murshidabad (19.8 percent), South 24 Parganas (18.2 percent), and Koch Bihar (16.6 percent)—exhibited somewhat the same proportionate decline in their TFRs as those which experienced a negative urban population growth (Bankura and West Dinajpur) during the same period (see Table 4). In line with the trend in 2001–2011, Maldah, with the highest increase in its urban population (85.5 percent), witnessed nearly the same decline in its fertility rates as West Dinajpur, which had the lowest increase (4.0 percent) in its urban population.

Changes of Sample Characteristics between Rounds

Table 5 depicts sample characteristics of women in the four successive rounds. A careful look suggests some changes in the sample characteristics over the years. For instance, the proportion of urban women and their partner’s mean years of schooling increased over the years irrespective of place of residence. Over the years, the proportion of rural Bengali women belonging to the poor (including both poorest and poorer women; henceforth poor) households remained almost the same, while the proportion of the affluent sections declined and that of the middle-class increased.

Table 5.

Background Characteristics of Currently Married Women (Aged fifteen to forty-nine) according to Their Place of Residence, Various NFHS Rounds, West Bengal.

	NFHS1 (1992–1993)		NFHS2 (1998–1999)		NFHS3 (2005–2006)		NFHS4 (2015–2016)
Place of residence
Rural	71.1		76.9		68.7		67.7
Urban	28.9		23.1		31.3		32.3
	Rural	Urban	Rural	Urban	Rural	Urban	Rural	Urban
Wealth quintiles
Poorest	26.5	4.2	27.6	4.7	32.6	1.8	26.5	8.3
Poorer	24.4	6.0	20.4	6.8	33.1	6.2	37.4	14.1
Middle	22.5	7.8	23.0	7.8	19.9	17.2	25.2	18.4
Richer	18.0	19.9	20.2	16.9	11.4	33.2	9.7	30.6
Richest	8.7	62.2	8.8	63.8	3.0	41.5	2.2	28.5
Mean years of schooling of respondent	2.7	5.9	2.7	6.6	3.6	7.2	5.4	7.8
Mean years of schooling of respondents’ partners	4.4	7.6	4.4	8.2	4.3	8.5	5.7	8.4
Religious affiliation
Hinduism	73.6	83.7	69.0	92.8	66.8	85.4	69.9	73.6
Islam	23.9	15.2	27.9	6.2	32.4	13.4	26.3	24.5
Respondent’s occupation
Not working	77	86.7	71.1	86.4	62.8	69.9	80.6	83.0
Primary activities	19.3	5.0	26.5	7.4	31.4	10.4	16.8	10.6
Secondary activities	3.0	5.2	1.4	2.7	1.5	3.6	1.1	1.8
Tertiary activities	0.8	3.1	1.0	3.5	4.3	16.1	1.6	4.6
Partner’s occupation
Not working	2.5	3.0	4.5	8.7	1.7	2.7	3.2	4.4
Primary activities	72.6	60.6	76.8	40.9	77.7	44.6	73.9	40.2
Secondary activities	20.7	54.9	13.6	32.4	12.7	33.6	19.2	43.6
Tertiary activities	4.2	11.5	5.1	18.0	7.9	19.1	3.6	11.8
Median age of respondents	28	31	29	33	28	30	29	30
Median age of respondents’ partners	35	39	35	40	36	40	38	40
Percent of respondents exposed to family planning messages in any mass media	31.7	44.1	49.6	77.4	48.9	79.7	45.8	70.3
^aMean number of children ever born	3.1	2.7	2.8	2.3	2.7	2.1	2.1	1.8
^aMean number of living children	3.0	2.7	2.5	2.1	2.4	1.9	2.0	1.7
^aMean number of living sons	1.5	1.4	1.3	1.1	1.2	1.0	1.0	0.9
^aMean number of living daughters	1.4	1.3	1.2	1.0	1.1	0.9	0.9	0.8
^aMean number of child losses	1.6	1.4	0.8	0.5	0.3	0.2	0.2	0.1
Degree of participation in household decision making by the respondents	__	__	1.4	1.5	1.9	2.3	3.0	3.3
Total cases (Un-weighted)	3,165	833	2,461	1,947	2,511	2,462	1,626	582
Total cases (weighted)	26,478	10,782	32,221	9,702	35,154	14,299	47,733	20,292

Source: Calculated from unit level data of NFHS (Women’s file), four rounds (1992–1993 to 2015–2016).

Note: ^aHave been calculated from women’s file who are currently married and aged between fifteen and forty-nine and other values are calculated from all women’s file.

In the urban sample, the representation of affluent women declined substantially, while the proportion of the poor and middle-class women increased in a sizeable proportion. The proportion of Hindus in urban areas declined in the last decade, while the proportion of Muslims in the rural areas increased first and then declined in the recent years. The proportion of non-working women in the sample has increased in recent years in both rural and urban areas. Occupation of respondent’s partner has been fluctuating over the years. In both rural and urban areas, the percentage of those engaged in primary sector activities has declined in recent years, while that of those engaged in secondary sector activities has increased. Rural respondents, as well as their partners, were observed to be younger compared to their urban counterparts. As expected, the proportion of women exposed to any family planning messages was higher among urban respondents compared their rural counterparts. In the sampled women, the mean number of children ever-born declined by one child in both rural and urban areas over the last two decades. The mean number of child losses was somewhat higher in the rural sample; however, it declined substantially over the years. Degree of participation in household matters increased among the sampled women over the years irrespective of place of residence.

Changes in the Preference for Contraceptive Methods

Table 3 depicts the difference in the preference for contraceptive methods according to the rural-urban residence over various NFHS rounds. Irrespective of the place of residence, by and large, the percentage of women using natural methods declined marginally across the four rounds. It declined from 18.1 percent in 1992–1993 to 15.6 percent in 2015–2016 in rural settings and from 26 percent to 18.4 percent in urban settings. Although the proportion of urban women who were not using any method at the time of the survey declined between the first and the third rounds, it increased marginally between the third and the fourth rounds (by 8 percentage points). On the other hand, the use of modern temporary methods increased by nearly 13 percentage points in the rural areas and only by 8 percentage points in the urban areas between 2005–2006 and 2015–2016. The proportion of women/couples using permanent sterilization declined from 38.3 percent to 33.9 percent in rural settings and from 33.1 percent to 24.1 percent in urban settings between NFHS-3 and NFHS-4. Irrespective of place of residence, want of future children, the mean number of living children as well as the mean number of living sons decreased over the survey rounds.

Changes in Fertility Preferences

Table 6 reveals changes in fertility preferences among respondents between 1992–1999 and 2005–2016 according to their place of residence. It was found that users of modern methods were of higher ages compared to those who were using natural methods, particularly in the rural areas, irrespective of survey periods. It is worth noting that between 1992–1999 and 2005–2016, the use of modern methods increased more than four-fold in rural areas, while the use of natural methods declined by 6 percentage points among those who wanted to have another child at the time of the survey. Increase in the use of modern methods, accompanied with a decline in natural methods, was also observed in the urban areas. Among users of modern methods, although the actual number of sons was significantly higher than the ideal number of sons in both the rural and urban areas before the current millennium, such differences became insignificant over the period. In other words, between 1992 and 1999, although a greater number of sons than desired were born in both the rural and urban areas, such differences reduced substantially between 2005 and 2016. On the other hand, statistically significant differences between the actual and the ideal number of daughters persist among users of modern methods in the urban areas, though they reduced marginally during the last decade. However, from these bivariate observations, it is hard to draw any inference regarding son preference and contraceptive method mix in rural and urban West Bengal.

Table 6.

Fertility Preferences and Gap between the Actual and Ideal Numbers of Boys and Girls by Contraceptive Choice among Various NFHS Rounds Corresponding with Their Place of Residence, West Bengal.

	Before 2000 AD (includes NFHS 1 & 2)						After 2000 AD (includes NFHS 3 & 4)
	Rural			Urban			Rural			Urban
	Natural	Modern	t-ratios	Natural	Modern	t-ratios	Natural	Modern	t-ratios	Natural	Modern	t-ratios
Median age of respondents	27	32	15.7***	32	33	2.1**	30	32	7.9***	33	33	11.8***
Percentage of respondents wishing to have another child	23.6	8.2	—	26.6	16.4	—	17.4	36.3	—	17.3	40.2	—
(a) Mean difference between actual number of sons and ideal number of sons	0.06	0.38	6.3***	0.10	0.43	6.4***	−0.08	0.19	−2.2	0.05	0.13	2.7
(b) Mean difference between actual number of daughters and ideal number of Daughters	0.31	0.45	2.1	0.24	0.46	3.5***	0.09	0.17	1.5	0.06	0.12	3.5**

Source: Calculated from unit level data of NFHS (Women’s file), four rounds (1992–1993 to 2015–2016).

***p\0.001; **p\0.05.

Changes in Contraceptive Adoption and Choice by Gender-parity Composition

Tables 7 and 8 unveil contraceptive adoption and choice according to the gender-parity composition between 1992–1999 and 2005–2016. Although contraceptive adoption increased irrespective of place of residence, it increased more substantially in rural areas compared to urban areas. It can be seen that in rural areas, adoption of natural methods increased by nearly 10 percentage points among first parity women without any living son, while it improved by only about 2 percentage points among first parity women who had one living son. At the same time, use of modern methods for first parity women with and without any living son increased by nearly 10 and 9 percentage points respectively. Adoption of modern contraceptive methods increased by nearly 9 percentage points and 12 percentage points among second parity women having two living sons and third or more parity women having three or more living sons respectively. However, use of natural methods remained either almost at the same level or declined somewhat among the said parities.

Table 7.

Contraceptive Choice among Currently Married Women Aged Fifteen to Forty-nine According to Gender and Parity Composition of Living Children, West Bengal, Rural.

Gender– parity composition	Rural
	Before 2000 AD (Includes NFHS 1 & 2)				After 2000 AD (Includes NFHS 3 & 4)
	Not using	Using Natural/Behavioral methods	Using modern methods	Number of cases (weighted)	Not using	Using Natural/Behavioral methods	Using modern methods	Number of cases (weighted)
Parity 1
No son	57.8	23.3	18.9	4,670	39.5	33.1	27.3	2,408
One son	50.0	28.7	21.3	5,275	37.2	31.0	31.9	3,164
Parity 2
No son	39.0	24.6	36.4	2,324	34.9	29.0	36.2	2,128
One son	28.1	19.0	52.9	7,608	21.0	18.8	60.2	5,208
Two sons	24.6	13.6	61.8	3,805	16.3	13.2	70.5	2,660
Parity 3
No son	46.6	18.9	34.5	1,507	30.5	22.9	46.7	1,470
One son	28.6	13.1	58.3	6,817	19.9	14.6	65.5	5,278
Two sons	26.3	9.6	64.1	9,558	20.5	13.0	66.5	5,390
Three or more sons	34.6	13.8	51.6	7,804	22.5	14.1	63.4	2,674
Total	34.8	16.9	48.3	49,368	25.0	19.5	55.5	30,380

Source: Calculated from unit level data of NFHS (Women’s file), two rounds (1992–1993 to 1998–1999).

Table 8.

Contraceptive Choice among Currently Married Women Aged Fifteen to Forty-nine According to Gender and Parity Composition of Living Children, West Bengal, Urban.

Gender–parity composition	Urban
	Before 2000 AD (Includes NFHS 1 & 2)				After 2000 AD (Includes NFHS 3 & 4)
	Not using	Using Natural/Behavioral methods	Using modern methods	Number of cases (weighted)	Not using	Using Natural/Behavioral methods	Using modern methods	Number of cases (weighted)
Parity 1
No son	34.9	43.0	22.1	2,444	21.1	43.1	35.8	1,839
One son	27.1	45.2	27.7	2,734	25.6	38.1	36.3	2,567
Parity 2
No son	34.3	35.3	30.4	940	18.8	21.4	59.8	892
One son	26.1	29.6	44.3	2,936	17.5	27.3	55.1	2,180
Two sons	21.8	19.6	58.6	1,580	9.0	24.0	67.0	1,019
Parity 3
No son	29.0	23.9	47.1	518	21.5	15.0	63.5	433
One son	24.1	14.9	61.1	1,710	13.3	14.9	71.8	1,422
Two sons	21.5	10.0	68.5	2,452	16.0	10.7	73.3	1,423
Three or more sons	35.8	10.9	53.3	2,168	22.4	9.7	67.9	557
Total	23.0	26.6	45.5	17,482	18.8	26.7	54.5	12,332

Source: Calculated from unit level data of NFHS (Women’s file), two rounds (2005–2006 to 2015–2016).

In urban areas, over the years, use of behavioral methods remained almost at the same level among first parity women without any living son; however, it decreased by 7 percentage points among first parity women with a living son. At the same time, the adoption of modern methods increased by 13 and 9 percentage points among first parity women without and with a living son. Adoption of modern methods increased by nearly two-folds among second parity women even if they did not have any living son. While it increased by nearly 9 percentage-points among second parity women with two living sons.

By contrast, adoption of natural methods declined by 14 percentage points among second parity women without any living son, while it increased by 4 percentage points for two living sons. Although the adoption of behavioral methods remained almost at the same level for third or more parity women having three or more living sons, the use of modern methods rose by nearly 14 percentage points. It was also observed that the use of behavioral methods declined by nearly 9 percentage points for third or more parity women even if they did not have a living son.

The rural-urban difference in contraceptive method mix between 1992 and 1999 reveals that at any given parity, the use of natural methods was higher in urban areas compared to their rural counterparts. At second parity, women from urban areas with at least a son were more likely to switch to modern methods from the natural ones; but such a pattern was not that clear for the last two NFHS rounds (2005–2016).

From the above findings, it is difficult to ascertain whether the change in contraceptive method mix was due to women switching from behavioral to modern methods once the desired number of sons was achieved or due to them starting to use modern methods directly once they accomplished the desired number of sons. Findings from multivariate models discussed below would throw some light on this issue.

Econometric Analysis

Adjusted odds ratios (AOR) and average marginal effects (AME), obtained from the sequential logit regression models were estimated for parity by place of residence after pooling the data sets of all four survey rounds, after controlling for several potentially confounding covariates. The estimated AOR and AME were also reported for the year of the survey. The odds and transitional probabilities of the first transition implied adopting a contraceptive method as opposed to not using a method (Transition 1), while the odds and transitional probabilities of choosing a modern method over a natural method signified a second transition (Transition 2). AOR and AME estimates of the first transition (Table 9) indicate that the likelihood of adopting any contraceptive method among women from West Bengal increased significantly with an increased number of sons, particularly up to second parity irrespective of place of residence. For example, at parity one, the probability of contraceptive use increased by 6 and 3 percentage points among women with one living son in rural and urban areas respectively compared to their counterparts who did not have a living son. At parity two, the likelihood of adopting a method increased by 8 percentage points among women with two living sons compared to those who did not have a living son irrespective of place of residence. Except for women having three or more living sons, similar observations can be made for the remaining parities.

Table 9.

Adjusted Odds Ratios (AOR) and Average Marginal Effects (AME) of Contraceptive Method Adoption by Place of Residence, West Bengal.

Gender–parity composition	Transition 1: Adoption of Contraception (not using vs. using)
	Rural			Urban
	Odds ratio	Average marginal effect	Z-stat	Odds ratio	Average marginal effect	Z-stat
Parity 1
No son	1.00			1.00
One son	1.31	0.06	8.20***	1.21	0.03	3.83***
NFHS Rounds
1	1.00			1.00
2	1.38	0.07	7.70***	1.66	0.10	7.57***
3	2.05	0.15	17.76***	1.76	0.11	9.80***
4	2.22	0.17	22.62***	1.29	0.05	5.18***
N (Weighted)	16,497			9,908
Parity 2
No son	1.00			1.00
One son	1.28	0.04	6.64***	1.07	0.01	0.96
Two sons	1.63	0.08	10.75***	1.79	0.08	6.73***
NFHS Rounds
1	1.00			1.00
2	1.17	0.02	3.80***	1.49	0.06	5.44***
3	1.76	0.08	14.09***	2.65	0.14	15.50***
4	1.64	0.07	13.98***	1.53	0.07	8.27***
N (Weighted)	24,349			9,735
Parity 3
No son	1.00			1.00
One son	1.17	0.03	2.64**	1.34	0.05	2.60**
Two sons	1.19	0.03	3.00**	1.45	0.06	3.29***
Three or more sons	0.94	−0.02	−1.01	0.86	−0.02	−1.25
NFHS Rounds
1	1.00			1.00
2	1.26	0.04	7.95***	2.42	0.14	12.64***
3	1.77	0.10	19.44***	2.96	0.17	16.91***
4	1.99	0.12	24.45***	2.84	0.17	19.32***
N (Weighted)	39,966			10,527
Total weighted cases	80,812			30,170

Source: Calculated from unit level data of NFHS (Women’s file), four rounds (1992–1993 to 2015–2016).

*** p\0.001; **p\0.01; *p\0.05.

The effect of time dummy suggests that although contraceptive adoption increased phenomenally at every parity over time in both rural and urban areas, the increase was very sharp in the rural areas, particularly at first parity. At first parity, the likelihood of adoption of any contraceptive method increased by 7, 15 and 17 percentage points among rural women in the later three rounds of the survey compared to the first one.

In urban areas, the increase was to the extent of 10, 11 and 5 percentage points in the later rounds of the surveys compared to the first one. AOR values for rural and urban areas demonstrate that the likelihood of adoption of any method was substantially higher among the rural women compared to their urban counterparts in the last two rounds. At second parity, the probability of adoption of a method increased by 2, 8 and 7 percentage points in the rural areas, and by 6, 14 and 7 percentage points in the urban areas in the later three survey rounds compared to the first round of the survey. In this case also, the AOR values suggest that compared to urban women, the likelihood of adoption of any contraceptive method was higher among the rural women in the last round of the survey. At parity three and above, the adoption increased by 4, 10 and 12 percentage points among rural women and by 14, 17 and 17 percentage points among urban women compared to the same reference category.

AOR and AME estimates from the second transition (Table 10) reveal that, at parity one, although rural women with one living son were significantly less likely to switch from natural to modern methods of contraception compared to their counterparts with no living son, such a likelihood increased by 2 percentage points among their urban counterparts. It possibly reflects an unsatisfied desire for another child, possibly a daughter after having a son at the first parity.

Table 10.

Adjusted Odds Ratios (AOR) and Average Marginal Effects (AME) of Contraceptive Method Choice by Place of Residence, Various NFHS Rounds, West Bengal.

Gender–parity composition	Transition 2: Choice of contraceptive methods (Natural vs. Modern)
	Rural			Urban
	Odds ratio	Average marginal effect	Z-stat	Odds ratio	Average marginal effect	Z-stat
Parity 1
No son	1.00			1.00
One son	0.91	−0.02	−1.94*	1.12	0.02	2.18**
NFHS Rounds
1	1.00			1.00
2	2.78	0.21	15.42***	2.29	0.17	10.17***
3	2.40	0.18	14.18***	2.84	0.22	15.10***
4	8.36	0.44	36.97***	6.11	0.38	28.11***
N (Weighted)	16,497			9,908
Parity 2
No son	1.00			1.00
One son	1.29	0.04	5.92**	0.88	−0.03	−2.05***
Two sons	1.98	0.12	12.97***	1.57	0.09	6.29***
NFHS Rounds
1	1.00			1.00
2	1.76	0.07	11.53***	1.43	0.05	4.93***
3	1.65	0.07	10.94***	1.90	0.10	10.37***
4	3.22	0.15	27.40***	3.86	0.20	22.88***
N (Weighted)	24,349			9,735
Parity 3
No son	1.00			1.00
One son	1.56	0.06	6.29***	1.35	0.04	2.54**
Two sons	2.04	0.10	10.13***	2.01	0.09	5.88***
Three or more sons	1.45	0.05	5.22*	1.64	0.06	3.94**
NFHS Rounds
1	1.00			1.00
2	1.27	0.03	6.17***	1.09	0.01	1.07
3	1.17	0.02	4.39***	1.17	0.02	1.97**
4	2.04	0.08	18.61***	1.96	0.08	9.03***
N (Weighted)	39,966			10,527
Total weighted cases	80,812			30,170

Source: Calculated from unit level data of NFHS (Women’s file), four rounds (1992–1993 to 2015–2016).

*** p\0.001; **p\0.01; *p\0.05.

At parity two, among women with two living sons, the likelihood of choosing a modern method over a natural one increased by 12 and 9 percentage points among rural and urban women respectively compared to women without any living son. Interestingly, at parity two, although the probability of choosing a modern method over a natural one increased by 4 percentage points among rural women with one living son compared to those having no living son, it declined by 3 percentage points among urban women compared to the same reference category. For three or more parity, the likelihood of choosing modern methods over natural ones increased with increased number of sons, except for those who had three or more sons. The decreasing propensity to use any method or to choose modern methods could be due to the diminution of fertility desire and/ or coital frequency among women, which basically stems from the attainment of higher ages together with reduced fecundity.⁸²

The time dummy reveals that the likelihood of choosing modern methods over natural methods has significantly increased over time, at every parity, particularly among first and second parity women irrespective of place of residence. In comparison to 1992, the probability of choosing modern methods over natural methods in 2016 was the highest among rural women at parity one (44 percentage points) compared to urban women (38 percentage points in urban areas). Such higher AOR values for the rural compared to the urban women were observed in the second round. For parity two, the likelihood of choosing modern methods increased by 20 and 15 percentage points in urban and rural areas, respectively, in 2016 compared to 1992.

From these findings, the adoption of contraception as well as the switch from the behavioral methods of contraception to the modern ones were both evident once women achieved their desired number of sons. Among rural women, the adoption of any method intensified after achieving a son at first parity itself in the recent years, while switching from behavioral methods to modern methods increased after achieving two sons at second parity compared to the urban women.

Discussions

The findings of this study illustrate the role of son preference and dynamics of contraceptive use in the convergence of rural-urban fertility differentials during the last twenty-five years or so. At the same time, the study also demonstrates how time has an independent effect in the adoption and switching of contraceptive methods. In other words, while there is evidence of varying contraceptive behaviors after achieving the desired number of sons in the convergence of fertility norms between rural and urban areas, there is also a strong positive and significant independent effect of time in the adoption and choice of contraceptive methods, which impacts fertility, particularly rural fertility.

The district-level proportionate decline (overall relative change) in fertility rates in West Bengal confirms a faster decline during 1991–2001 (by −29.0 percent) than during 2001–2011 (−22.2 percent). However, the gap between rural-urban fertility in West Bengal started to wane only after the year 2000, primarily due to the remarkable decline in fertility in rural areas. As mentioned earlier, urban Bengal had already attained a low level of fertility before the year 2000. It is important to note that despite an insubstantial influence of urbanization on fertility decline (Table 4), low mean years of schooling, low exposure to family panning massages, and lack of participation in wage-earning sector activities (Table 5), rural women have startlingly managed to achieve below replacement-level fertility by having only two living children on an average in the recent past (Table 5).

In the beginning of this paper, it was postulated that the convergence between rural and urban fertility could be due to differential stopping behavior (DSB) after achieving the desired number of sons among rural women. Evidence to this effect was found, particularly during the last two rounds of the survey, and, among rural women, as manifested through the adoption of a contraceptive method even after achieving a son at first parity and the switch to modern methods from the natural ones after achieving two sons at second parity. In other words, DSB, as manifested by the adoption of contraception at first parity after having a boy or the choice of modern methods over natural ones at second parity after having two boys successively could be one of the factors that have led to the sharp decline in fertility in rural areas of West Bengal in recent times. In line with these findings, results show that in rural Bengal, fewer than half of the women chose permanent sterilization at second parity after having two sons (47.4 percent) and/or one girl and one son (46.2 percent) between 2005–2006 and 2015–2016. For urban Bengal (data not shown), the corresponding figures were 42.5 percent and 41.3 percent respectively. Notably, over the last twenty-five years or so, irrespective of place of residence, the desired as well as the actual number of children has reduced to around two births per women—with a convergence of mean difference between actual and desired number of sons—particularly in rural areas (Table 3). Hence, it can be inferred from the findings that couples have started to achieve their desired number as well as the sex composition of children in the recent past through the process of DSB.

It is worth noting that child sex ratio (sex ratio of children aged zero to six years) declined from 969 to 963 between 1991 and 2001 and further to 959 between 2001 and 2011 (10 points during 1991–2011) in rural West Bengal. In urban areas, the decline was from 955 to 947 (8 points) between 1991 and 2011.^83–
85 Das (2014) noted that the decline in child sex ratio in West Bengal was substantially higher in the predominantly rural districts, which had completed or were on the verge of completing the fertility transition. This decline was neither due to sex-selective abortion nor due to gender discrimination in early childhood.⁸⁶ The present study supports these arguments.

In earlier studies, Ghosh and Begum (2015),⁸⁷ Ghosh and Chattopadhyay (2017),⁸⁸ and Ghosh et al. (2020)⁸⁹ have argued that the preference for sons over daughters in the Bengali community remains restricted at the attitudinal level and can be termed as “latent” son preference. In support of this argument, it was observed that couples without the desired number of living sons either did not adopt any contraceptive method or preferred natural methods over modern ones because they were not overtly concerned with pregnancy or failure of contraceptive (natural) methods or costs (economic and/or social) of unwanted children. These studies have, furthermore, argued that son preference among the Bengali community is in contrast to son preference in north-western India, which is generally manifested through sex-selective abortions and poorer nutritional status and higher mortality among girl children compared to boys.

Perhaps, the influence of son preference on the contraceptive method mix in West Bengal is not analogous to the culturally similar settings of Bangladesh. Depending on the sex-composition of the previous children, the influence of son preference on contraceptive method mix is greater in West Bengal,^90,91 which allows for the desire for additional children as compared to the same in Bangladesh.⁹² Except a few small-scale studies^93–
95—which have reported a preference for boys over girls cutting across religion, economic standing, and educational attainment—no large-scale evidence has emerged suggesting if the undesired child is a girl, if she is tagged as “unwanted,” or if she is deprived of privileges such as immunization, nutrition, education, and well-being. In line with the findings of Roy and Chattopadhyay (2012),⁹⁶ the present study also argues that such a gender preference in West Bengal would have resulted in further decline in fertility, plausibly at “lowest-low” level before stabilization, and would have contributed to a negative population growth rate in the foreseeable future.

There are a few limitations of the study that need to be pointed out. First, the effect of time upon the contraceptive method mix could have been better understood with the help of a cohort study rather than a pooled cross-sectional analysis. Better insights regarding the change in contraceptive behavior could be gleaned if a common group of individuals were followed over a period of time. Secondly, due to the small number of observations, neither could male and female sterilization be considered as separate variables nor could the likelihood of choice of modern temporary methods over permanent methods in relation with DSB at the third transition be established in the analytical models. As such, the explicit effect of sterilization on DSB could not be ascertained. Thirdly, the data on son preference was based on self-reporting, which means that it may have been misreported to present a favorable impression or that there may have been post-facto rationalization. Furthermore, there was scant information regarding community-level characteristics like mean educational attainment and age at marriage at the primary sampling unit level in the dataset. Thus, the findings could not be analyzed and adjusted against community-level variables.

Nonetheless, the present study has the ability to flesh out one of the plausible reasons for the contemporary rural fertility transition in West Bengal, on which studies are limited. Previous studies carried out in this regard have mostly tried to elucidate factors influencing contemporary fertility decline in West Bengal, whereas the present study attempts to provide a plausible explanation for such decline. Apart from DSB and its manifestation through contraceptive adoption and choice, at least other two alternative hypotheses can be put forth for the contemporary fertility decline in rural areas of West Bengal, which should be tested empirically in future research. First, diffusion of aspirations for children and constraints in childbearing and childrearing from urban to rural areas of West Bengal could be one of the important reasons for the contemporary fertility decline in rural West Bengal. Ghosh (2016)⁹⁷ observed 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 urban West Bengal. Secondly, one can also postulate that economic hardship, in absolute as well as relative terms—as manifested through the decline in the size of agricultural landholding and transformation of the workforce from cultivators to agricultural laborers—induces poor, uneducated and socioeconomically marginalized women/couples of rural areas to limit their family size under the changing mortality regime through an increase in contraceptive use.

Footnotes

Appendix Table A1.

Percent Distribution of Currently Married Women According to Current Usage of Contraceptive Methods, West Bengal, Various NFHS rounds.

NFHS Rounds	Modern Methods						Behavioral Method			Other Methods	Not Currently Using any Method
	Reversible				Permanent		Natural		Periodic
	Oral Pill	IUD	Injectables	Condom	Female Sterilization	Male Sterilization	Any Traditional Method	Withdrawal	Periodic-Abstinence/Rhythm/safe period
NFHS1	5.2	1.6	—	4.3	23.3	2.1	25.3	11.8	12.8	0.7	38.2
NFHS2	9.2	1.4	—	2.9	32.0	1.8	18.5	9.8	8.7	0.9	33.4
NFHS3	11.7	0.6	0.3	4.3	32.2	0.7	21.3	8.4	12.3	0.5	28.8
NFHS4	20.0	1.2	0.2	5.9	29.2	0.1	13.9	6.8	7.1	0.0	29.1

Source: NFHS reports of West Bengal, Four rounds.

Note: Only in NFHS1 women aged between 13 to 19 were interviewed and for rest of the rounds age-ranged between 15 to 49 were considered; inconsistent methods were excluded from the tables.

Authors’ Note

A statement or an argument has been proven in different ways in different papers for different parts of India. Thus, multiple references have been cited here.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Rural–Urban Fertility Convergence,Differential Stopping Behavior,and Contraceptive Method Mix in West Bengal,India: A Spatiotemporal Analysis

Abstract

Keywords

Introduction

Works Which Have Led up to the Study

Materials and Methods

Data and Variables

Analytical Model

Results

Trends and Differentials of District-level Fertility Rates and Percentage of Urban Population in West Bengal

Changes of Sample Characteristics between Rounds

Changes in the Preference for Contraceptive Methods

Changes in Fertility Preferences

Changes in Contraceptive Adoption and Choice by Gender-parity Composition

Econometric Analysis

Discussions

Footnotes

Authors’ Note

Declaration of Conflicting Interests

Funding

Notes