GIRLS' SCHOOLING CHOICES AND HOME PRODUCTION: EVIDENCE FROM PAKISTAN

• Date: 01 May 2020
• DOI: http://doi.org/10.1111/iere.12440
• Published date: 01 May 2020
• Author: Hugo Reis

INTERNATIONAL ECONOMIC REVIEW

Vol. 61, No. 2, May 2020 DOI: 10.1111/iere.12440

GIRLS’ SCHOOLING CHOICES AND HOME PRODUCTION: EVIDENCE

FROM PAKISTAN∗

BYHUGO REIS1

Banco de Portugal and Universidade Cat´

olica Portuguesa, Cat´

olica Lisbon School of Business

and Economics, Cat ´

olica Lisbon Research Unit in Business and Economics, Portugal

The article develops and estimates a dynamic structural model of girls’ school-going decisions and mother’s

labor market participation. It seeks to determine the causes of low school participation and to evaluate alternative

public policies. The model incorporates mother’s education, school availability, the productivity of the girl when

engaged in household production, and the potential trade-off between mother’s and daughter’s housework

decisions. Our findings suggest that school construction is the most cost-effective program. When using monetary

incentives, our results highlight the effectiveness of conditionality, as opposed to unconditional transfers, and

the existence of a trade-off between maternal employment and daughter’s schooling.

1. INTRODUCTION

It is widely acknowledged that human capital plays a critical role in developing countries’

economic growth, which has motivated social programs aimed at enhancing the educational

participation of children, especially of girls. The social benefits associated with women’s educa-

tion are noteworthy. An educated mother tends to have more influence in household decisions,

investing more resources in the education and health of her children. There is extensive empir-

ical literature on the strong association between maternal schooling and children’s outcomes in

developed and developing countries, which is surveyed in Behrman (1997). An educated girl

is more likely to be self-confident and to participate in the formal labor market. Investment in

girls’ education, especially beyond primary school, has been identified as one of the greatest

challenges for low-income countries. If nothing changes in the society, the current daughter will

face the same decision problem when she has a daughter of her own.

Many lower income countries have made great progress in terms of primary and secondary

school enrollment rates. However, Pakistan still lags far behind, especially regarding girls in

rural areas.2Net enrollment rates for girls in those areas are below 60% and 35% for primary

and secondary education, respectively. It is therefore important to understand the reasons for

the prevailing low levels of investment in girls’ education.

∗Manuscript received October 2017; revised August 2019.

1The helpful and constructive comments of the three anonymous referees and of the Associate Editor, Christopher

Flinn, as well of those of Orazio Attanasio, Pedro Carneiro, M´

onica Costa Dias, and Petra Todd are gratefully

acknowledged. The article has also benefited from insightful comments made by Lars Nesheim, Michael Keane, Aureo

de Paula, Pierre-Andr´

e Chiappori, Jishnu Das, Pat Kehoe, Dan Black, Paulo Rodrigues, Sankar Mukhopadhyay,

and participants at the IZA/SOLE Transatlantic Meeting of Labor Economists and at the Northeast Universities

Development Consortium Conference. The excellent research assistance by Marta Lopes and the support of Fundac¸˜

ao

paraaCi

ˆ

encia e Tecnologia, World Bank, University College London, and Banco de Portugal is also acknowledged.

Please address correspondence to: Hugo Reis, Economics and Research Department, Banco de Portugal, Av. Almirante

Reis 71, 1150-012 Lisboa, Portugal. Phone: +351-213130282. E-mail: hfreis@bportugal.pt.

2Despite the efforts, educational participation in Pakistan is far from the targets proposed by several international

institutions, like the United Nations, especially for girls. For example, Pakistan is far behind the Millennium Devel-

opment Goals (MDG) regarding Universal Primary Education. According to the MDG (2015) report, in 2015, the

Pakistani net primary enrollment ratio and the literacy rate were below 60% overall (below 50% for girls in both cases).

783

C

(2020) by the Economics Department of the University of Pennsylvania and the Osaka University Institute of Social

and Economic Research Association

784 REIS

Housework activities, especially for older girls, play an important role in explaining a girl’s

schooling decisions, as presented and discussed in Andrabi et al. (2010) and Andrabi et al.

(2012) using a child time allocation survey. The presence of a trade-off between housework and

school for the older girls (aged 12–15) creates an important economic disincentive to invest in

education. At the same time, South Asia has one of the lowest female labor force participation

(LFP) rates in the world. For instance, despite the increasing trend, in Pakistan, the female LFP

was only about 25% in 2017.3Efforts currently underway seek to improve the labor supply of

women, but if maternal labor supply and girls’ schooling are substitutes in home production,

such efforts may have unintended adverse consequences on girls’ schooling. Other determinants

of girls’ schooling could be poverty, cultural considerations, or an inadequate supply of schools.

In this article, we examine these issues through a behavioral model.

In most developing countries, women are still situated at the lower end of the educational

system (and consequently also in the labor market) in comparison to their male counterparts,

particularly in rural and suburban areas. Indeed, there is a large literature suggesting very low

female LFP as a general and critical issue for these countries. Pakistan is a particularly good

example but there are also striking examples in India, Afghanistan, and in the Middle East and

North Africa.4Therefore, the importance of this issue goes clearly beyond Pakistan, as also

highlighted in the World Development Report (2012) on gender differences in employment

and why they matter. In particular, this report provides a comprehensive discussion specifically

on women’s home production and employment, emphasizing how still important and critical

these issues are for the development of many low-income countries as they are likely to be

reproduced and passed on to the next generation.

We develop and estimate a dynamic discrete choice model over a finite horizon to understand

the determinants of girls’ schooling decisions. Our behavioral model spells out the interrelations

between a girl’s schooling and her mother’s labor market decision, incorporating the role of

home production. It enables us to carry out the impact analysis of alternative policies that aim

to increase girls’ school enrollment beyond primary education. The empirical analysis is based

on detailed information on the school choice decisions in the rural Punjab province, the largest

state in Pakistan, obtained from the Learning and Educational Achievement in Punjab Schools

(LEAPS) data set, for girls in the primary, elementary, and secondary education level.

In the model, the household makes the employment decision for the mother and the schooling

decision for one girl aged 6–15 under a process similar to that proposed by Bernal (2008) and

Todd and Wolpin (2006). In Bernal (2008), the mother makes childcare and labor market

participation (part time vs. full time) decisions, whereas in our case, the mother can opt for

only the full-time solution. In Todd and Wolpin (2006), married couples are assumed to make

sequential decisions over a finite horizon about their children’s school attendance and labor

market participation, and about the timing and spacing of births. Parents’ labor market decisions

are not included in their model. Taking into account fertility effects is potentially interesting.

Nevertheless, the unresponsiveness of fertility to the program incentives in Todd and Wolpin

(2006) should be pointed out. Moreover, we do not observe the family decision from the start of

the marriage. Therefore, in contrast to Todd and Wolpin (2006), but in line with Attanasio et al.

(2012), we are not considering a childbirth decision in the model. The main feature of our model

is the explicit consideration of the role of home production, modeling the trade-off between

a girl’s schooling and housework decisions and the potential implications of mother’s labor

market participation on the girl’s schooling decision. Moreover, it is important to highlight that

the more natural trade-off between generations is mother–daughter and not daughter–father

(see, e.g., Lillard and Willis, 1994; Thomas, 1994). Including the father’s choices in the model

would make the model considerably more complex without obvious gains in terms of the more

natural and important trade-off regarding girls’ schooling decisions.

3This figure is similar to India, below Bangladesh and Sri Lanka levels (just above 30%), and above countries like

Afghanistan (around 15%).

4See, for example, Jayachandran (2015) for a discussion on the roots of gender inequality in developing countries.

GIRLS’SCHOOLING AND HOME PRODUCTION 785

Girls are the focus of our analysis and the unit of observation is one girl and her mother. In

addition to the impact of housework on girl’s decisions, the model allows for other household

characteristics to affect the decision, even though as exogenous determinants. In particular,

father’s income is exogenous, as well as the presence in the household of both boys of schooling

age and very young children (aged 0–5). Potential birth order effects are also considered through

a variable that identifies whether the girl is the oldest daughter in the household. The model also

allows for important features that influence preferences, such as monetary and nonmonetary

costs of not having a school (public or private) beyond the primary education level in the village.

The parameters of the model are estimated by simulated maximum likelihood and fit the data

well. Overall, the estimated parameters have the expected signs, and the model replicates the

observed distribution of schooling choices for each specific girl’s age. The model accounts for

the significant decrease in girls’ school enrollment rates after age 12 and for the relation between

mother’s working and girls’ schooling. The model shows that home production is an important

factor underlying the utility of sending a girl to school. Moreover, in line with key findings in the

education and development economics literatures, the results also show the important role of

mother’s education, and school availability beyond primary level in the girl’s schooling decision.

The schooling labor decision model developed in this article builds on several strands of the

literature, including dynamic models of occupational choices as in Keane and Wolpin (1997),

dynamic models of employment–childcare decisions of women as in Bernal (2008), and on

closely related papers by Attanasio et al. (2012) and Todd and Wolpin (2006), in which a dynamic

schooling behavioral model is used to evaluate the impact of monetary incentives provided to

families to increase children’s school attendance.5We extend the literature by combining the

mother’s and girl’s decisions in the setting of a low-income developing country, and use the

model to simulate girls’ school participation under different scenarios. In terms of policy, we

focus mainly on interventions aimed at increasing investment in girls’ education beyond primary

school. Those policies seek to reduce poverty and/or alleviate credit constraints through cash

transfer programs, and to reduce the distance to schools by establishing new schools beyond

primary education.

Indeed, governments have implemented several measures to provide incentives for attracting

and retaining students, including gender-targeted conditional cash transfer (CCT) programs that

explicitly address intrahousehold disparities in human capital investment. Such programs are

also intended to reduce poverty and alleviate credit constraints (see Fiszbein and Schady, 2009,

for a comprehensive survey).6However, in a context of inadequate supply of schools, to expand

education beyond primary education, it is important to understand the cost effectiveness of

monetary incentive schemes to potentially balance policies addressing poverty and alleviating

financial constraints for the most disadvantaged population with the establishment of new

schools. Our simulations compare the cost effectiveness of these different programs.

From a policy perspective, simulations suggest that in a context of inadequate supply of

schools, the school construction program is more cost effective than any cash transfer program.

In particular, the impact of the school construction program on school enrollment for girls aged

14–15 is close to four times greater than the CCT program. Our findings indicate that cash

transfer programs can be an inefficient public expenditure unless there are enough schools that

can absorb the increase in demand. Nevertheless, to create better incentives for girls’ investment

in education, it is desirable that the policy become increasingly demand-driven. Our results

highlight the effectiveness of conditionality when using monetary incentives, as emphasized by

Baird et al. (2011), Attanasio et al. (2015), and Del Boca et al. (2016). On average, the increase

in older girls’ school enrollment of the unconditional program is only about 10% as large as the

CCT scheme. Finally, we also address the cost effectiveness of the timing of the transfers. Our

5Aguirregabiria and Mira (2010), Keane et al. (2011), and Todd and Wolpin (2010) present a good survey of these

types of behavioral models.

6In Pakistan, incentives toward girls’ education have targeted the primary level. An important exception is the

Girls Stipend Conditional Cash Transfer in Punjab (2004), which sought to increase girls’ schooling progression after

completing the primary level and it was performed after the baseline survey from LEAPS.