Is son preference disappearing from Bangladesh?

Highlights

• Using a survey of women born between 1975 and 1994, we assess how son preference is evolving in Bangladesh.

• We find that, among women of childbearing age, son preference is giving way to a desire for gender balance.

• We explore factors behind the trend: increased female education and employment, and decline in joint family living.

• In contrast to stated fertility preferences, actual fertility decisions are still shaped by son preference.

Abstract

Historically, son preference has been widely prevalent in South Asia, manifested in the form of skewed sex ratios, gender differentials in child mortality, and worse educational investments in daughters versus sons. In the present study, we show, using data from a purposefully designed nationally representative survey for Bangladesh, that among women of childbearing age, son bias in stated fertility preferences has weakened and there is an emerging preference for gender balance. We examine a number of different hypotheses for the decline in son preference, including the increasing availability of female employment in the manufacturing sector, increased female education, and the decline of joint family living. Using survival analysis, we show that in contrast to stated fertility preferences, actual fertility decisions are still shaped by son preference.

Introduction

The phenomenon of “son preference” has been widely documented in different parts of the world, most notably in East and South Asia, the Middle East, and North Africa. We use the term to refer to any situation where parents value sons over daughters along some dimension and make choices – for example, relating to fertility or investments in children – on the basis of these preferences. These practices have potentially far-reaching economic and demographic consequences (Edlund, 1999) including, for example, excess female adult and maternal mortality (Milazzo, 2018), sex-selective abortions (Jha et al., 2011), gender differences in breastfeeding (Hafeez and Quintana-Domeque, 2018, Jayachandran, 2015), intra-household gender bias in food allocation (Rahman, 2018), gender differentials in infant and child mortality (Rose, 1999), imbalanced sex ratios and shortages of marriageable women in the population (Hudson & den Boer, 2004).

In South Asia, son preference has historically manifested itself in the form of imbalanced sex ratios. In India, sex ratios were persistently (and increasingly) imbalanced during the twentieth century, and stood at 933 females per 1000 males in 2001 (Pande & Astone, 2007). In recent years, there has been a worsening of child sex ratios in India (Jha et al., 2011). These trends have been attributed to a combination of a decline in desired family size, the diffusion of prenatal sex diagnostic technologies, and sex-selective abortion (see Bhalotra and Cochrane, 2010, Jha et al., 2011, Jayachandran, 2017).

However, these recent trends are not shared across all of South Asia: Bangladesh has experienced a decline in fertility together with an improvement in child sex ratios (Kabeer, Huq, & Mahmud, 2014). Data on declared preferences for sons and daughters indicate a steady decline in son preferences among women in Bangladesh and – to a lesser extent – in India and Nepal (these trends are discussed in more detail in Section 3). Whether and to what extent these declared child sex preferences are shaping actual fertility decisions remains, however, an open question.

In this paper, we use a purposefully-designed survey to assess independently whether and to what extent son preference has declined among women in Bangladesh. In addition, we address the following questions. Is son preference giving way to indifference regarding the sex composition of children or some other type of preference such as a desire for both sons and daughters? What are the drivers of the decline in son preference? Are the changes in stated child sex preferences reflected in actual fertility decisions such as birth spacing and the number of children?

The survey was conducted in 2014 with a nationally representative sample of women in Bangladesh of childbearing age and contains information on the respondents’ fertility history, their desire for future sons and daughters and other socio-economic characteristics. We use a regression framework to investigate how the birth of a son versus a daughter affects respondents’ stated desires for future sons and daughters. This approach allows us to infer whether the population, on average, exhibits son preference, a ‘balance’ between sons and daughters or indifference between sons and daughters. Furthermore, we use survival/duration analysis to investigate how the birth of a son or daughter affects the decision to have another child and the associated birth-spacing. This analysis addresses the question whether women’s stated preferences regarding sons and daughters are reflected in their actual fertility decisions.

Our analysis indicates a strong desire among women in Bangladesh for children of both sexes. Among respondents who have not yet had a child, the proportions indicating a desire for sons and daughters are almost identical. Among respondents with one or two children, the presence of a son has a strong negative effect on the desire for additional sons, and the presence of a daughter has a strong negative effect on the desire for additional daughters.

We use the same approach to explore heterogeneity in preferences within the population. We find some evidence that the desire for gender balance in child sex composition is stronger among women who have completed secondary school and those who live in areas with more opportunities for female paid work, specifically in the ready-made garments sector. The desire for gender balance in the sex composition of children is, surprisingly, stronger among women who are co-resident with their mothers-in-law.

The survival analysis indicates that actual fertility decisions are still shaped by son preference. The model estimates indicate that respondents who have no sons among their first two children are significantly more likely to have another child in any subsequent time period relative to those who do. The absence of a daughter among the first two children, on the other hand, has no corresponding effect on the decision to have another child.

Thus, our analysis reveals a discrepancy between the child sex preferences of women in Bangladesh and their fertility behaviour: while they express a desire for both sons and daughters, it is only the desire for sons that shape actual fertility decisions. We also uncover some suggestive evidence that fertility is affected by constraints on women’s access to birth control. Among older cohorts of women in the sample with two or more children, the husband’s opposition to birth control increased the risk of a third birth in a specific time period by 50%. Issues related to access to birth control methods may explain the discrepancy between child sex preferences and actual fertility decisions.

The question as to whether son preference is weakening in Bangladesh has previously been raised in the literature. Kabeer et al. (2014) document the phenomenon in Bangladesh using qualitative interviews and a quantitative survey conducted in 2008 in 8 districts in Bangladesh. Based on their quantitative measures, they report son preference among 40% of respondents, daughter preference among 7%, and indifference among the rest. We contribute to the existing work with evidence from a more recent, nationally representative survey and show that (i) son preference has, on average, given way to a desire for gender balance in child sex composition and (ii) actual fertility decisions are lagging behind the evolution in women’s child sex preferences.

Our contribution is also distinctive from a large number of existing studies that measure child sex preferences using questions on the ideal number of daughters and sons, or the desire for sons versus daughters in a hypothetical situation (see, for example, Clark, 2000, Pande and Astone, 2007, Kabeer et al., 2014). In contrast to this approach, our methodological approach allows a focus on women who are physically able to bear children and their desire for future sons and daughters.¹

Our survival/duration analysis using the fertility history of respondents follows previous work in the literature that have used this approach to show that the sex composition of existing children affects subsequent fertility decisions for Shanxi province in China (Tu, 1991), Bangladesh (Rahman & DaVanzo, 1993), India (Arnold, Choe, & Roy, 1998), Vietnam (Haughton & Haughton, 1998), China (Poston, 2002) and Pakistan (Javed & Mughal, 2020). Unlike the existing literature, we investigate fertility decisions in a population where son preference has definitively given way to a desire for children of both sexes as far as stated preferences are concerned. Nevertheless, in line with much of the existing literature, we find that actual fertility decisions of women in Bangladesh continue to be shaped by son preference.