Son Preference, Number of Children, Education and Occupational Choice in Rural Nepal

• Author: Magnus Hatlebakk
• DOI: http://doi.org/10.1111/rode.12237
• Published date: 01 February 2017
• Date: 01 February 2017

Son Preference, Number of Children, Education and

Occupational Choice in Rural Nepal

Magnus Hatlebakk*

Abstract

A unique family survey was conducted in Nepal to investigate the economic consequences of having a

first-born girl. Women have more children, but we find no causal effect of number of children on

economic outcomes, but independently of the number of children there is a positive effect on boys’

education of having a first-born sister, who presumably takes care of household work so the boys can

focus on school. This indicates a stronger son preference in Nepal than that found in studies from

neighboring countries.

1. Introduction

For South Asia it is well known that son preference explains selective abortion of

girls (at least in India), or additional children if the first born turns out to be a girl.

1

We focus on the potential costs of having too many children in Nepal, where, as we

shall see, abortion of girls is still not a problem. We investigate whether son

preference leads to more children and thus less investment in each child. Inspired

by the literature we started out with the random gender of the first born as an

exogenous instrument for number of children, but we discovered that it is not the

number of children that matters for the investments in each child, it is rather the

instrument itself that has a direct and positive effect, in contrast to the potential

negative effect of number of children. If the first born is a girl, then later-born boys

receive better education. This finding, which has no parallel in the literature as far

as we know, is reported together with a number of robustness checks and analysis

of additional economic outcomes.

We combine two strands of literature in a unified framework. Angrist and Evans

(1998) is the seminal contribution to the literature where the gender of first-born

children is used as an instrument for the number of children. Their work led to an

extensive literature focusing on the effect of family size on female labor supply, see

for example Cruces and Galiani (2007) on data from Mexico and Argentina, and

Daouli et al. (2009) on Greece. Within this strand there is a more limited literature

that looks at other economic outcomes, including Gupta and Dubey (2006) on

poverty in India. Our focus is on education, and similarly to us, most studies find

no causal effect of number of children on children’s education, although there are

*Hatlebakk: Chr. Michelsen Institute (CMI), P.O. Box 6033, 5892, Bergen, Norway. Tel: +47-4793-8000;

Fax: +47-4793-8001; E-mail: magnus.hatlebakk@cmi.no. The research is part of a joint project with

Yogendra Gurung and Bal Kumar KC, IPDS in Kathmandu. The research is funded by the Research

Council of Norway through the EconPop program that, in turn, is part of the international PopPov

network. Thanks go to Anjushree Pradhan from New Era who advised on the survey, as well as Sachin

Shrestha, also New Era, who supervised the survey together with us. Thanks also go to Arild Aakvik and

Vincent Somville for useful comments and suggestions.

Review of Development Economics, 21(1), 1–20, 2017

DOI:10.1111/rode.12237

©2016 John Wiley & Sons Ltd

two recent studies that find a negative effect by Ponczek and Souza (2012) and

Kugler and Kumar (2015). Ponczek and Souza also have a good review of this

literature.

Within this strand of the literature the seminal theoretical model of the link

between number of children and investments in each child is due to Becker (1960,

see his footnote 10 for the formal model). Becker describes a quality–quantity

trade-off in preferences where the two variables are simultaneously determined,

subject to the costs of raising and educating children. If a women for some reason

(and we focus on the random event that she has a girl first) decides to have more

children, she may at the same time decide to compensate by investing less in each

child, but that would be a causal effect of having a girl first on both quality and

quantity of children. When we attempt below to identify a causal effect of number

of children on their education, then we implicitly assume a recursive structure

where the women first observe the gender of the first born, then decide on the

number of children and finally decide on their education. This is not an unlikely

sequence of events, but as said, we do not find a significant effect in the last part of

the chain of decisions, only the first effect of the gender of the first born on number

of children is significant. This may reflect that a first-born girl simultaneously affects

both the number of children and their education, but we focus below on a more

direct interpretation of the finding, as it is likely that the first-born girl takes care of

household work so that her brothers can focus on school. This interpretation is

supported by Edmonds’ (2006) analysis of time use among children in Nepal where

he, in fact, finds that older girls work more than their brothers and even more if

they have additional younger siblings.

The second strand of the literature applies birth order and gender as direct

explanatory variables. A robust finding from poor countries seems to be that

siblings of first-born girls receive more education than others. We have, however,

an even more specific finding in our data, as only male siblings receive more

education if the first born is a girl. We thus find a stronger son preference, since

only boys benefit, than in similar studies from neighboring countries where later-

born children of both gender benefit; see Sawada and Lokshin (2009) on Pakistan,

and Ota and Moffatt (2007) on India. This male only effect is the main contribution

of the paper. There is a related literature where the focus is only on birth order

and a possible interaction with own gender (but without looking at the gender of

the siblings), see for example Ejrnæs and P€

ortner (2004) who find that later-born

boys (in the Philippines) spend more time in school, and Emerson and Souza

(2008) who find that daughters (in Brazil) who are first born are less likely to go to

school than later-born daughters.

We consider it value added to use data from Nepal, where the literature on son

preference is limited, but see Koolwal (2007) who analyzes the link between son

preference and child labor in Nepal. There is, however, a related literature on son

preference elsewhere, in particular Basu and de Jong (2010) on son preferences in

India, Das Gupta et al. (2003) on determinants of son preference in China, India

and South Korea, and Edlund and Lee (2009) on theory and evidence for son

preference in South Korea.

The fact that Nepali women have more children if the first born is a girl is well

known—see for example Gudbrandsen (2013), who uses the 2006 demographic and

health survey (DHS).

2

The DHS data has, however, only limited information on

economic outcomes. An alternative is the Nepal Living Standards Surveys (NLSS,

1995, 2003, 2010), which are of high quality and contain information on a number

2Magnus Hatlebakk

©2016 John Wiley & Sons Ltd