Gender Discrimination in the Family

Prashant Bharadwaj, Gordon B. Dahl, Ketki Sheth

University of California, San Diego
Abstract
This chapter examines whether parents discriminate among their sons and daughters in
ways that could have important and long-lasting effects. We begin by laying out the
methodological challenges of measuring discriminatory treatment. We then review the
empirical evidence for both the developed and developing world. The research for
developed nations documents that daughters are more likely to grow up in larger families,
in fatherless households, and in poverty. Child gender also affects parents labor force
participation and paternal involvement. Many of these same patterns, often magnified,
are found in developing countries. There is evidence of sex selective abortions,
differential investments in childhood health, and gender differences in household inputs
such as educational investments, parental time, and food.

1. Introduction
Gender differences found in adulthood are in part a product of an individual's
family environment during childhood. The leading difference studied by economists has
been the gender wage gap. In both developed and developing countries alike, wages for
female adults are significantly lower than their male counterparts (World Bank 2001,
OECD 2008). One explanation is discrimination in the workplace. But part of the gap can
also be explained by the different characteristics of women versus men, such as
occupational choice and the decision to remain in the labor force (Altonji and Blank
1999). Are these differing characteristics a reflection of innate differences between men
and women, or could they at least partly be consequences of differential treatment and
allocation of resources within the family setting? A better understanding of the existence
and magnitude of gender discrimination early in life is crucial for addressing the
pervasive gender inequities which persist into adulthood. This chapter focuses on
whether parents discriminate among their sons and daughters in ways that could have
1

important and long-lasting effects.

More attention has been paid to discrimination against young girls in developing
countries than in richer industrialized nations. Yet, recent studies have shown that though
the manifestation of gender discrimination differs in these two settings, there are small,
but significant, gender differences in early childhood inputs in more developed countries
as well. Critical factors such as marital stability, family size and sibling gender, labor
force participation of parents, and parental involvement have been found to differ in both
developed and developing countries. These inputs are significant factors in a childs
upbringing. The fact that ones gender influences them suggests that the childhood
environment of boys and girls differs even in developed countries, creating gender
differences at very early ages in life. These documented differences may also be
indicative of other ways in which parents treat sons and daughters unequally.
This chapter will review the literature which documents the existence of gender
differences that occur early in life through the family. We will discuss the empirical
evidence on how offspring gender influences parental decisions and how this may impact
investments during childhood and later in life. Though most studies are able to document
the existence of gender differences, an understanding of the reasons behind these
differentials is limited. For this reason, this chapter is focused on documenting the
existence and extent of discrimination, without much discussion on the possible
mechanisms for why such differences might occur.
Before discussing the empirical evidence, we first lay out the methodological
challenges of measuring discrimination. Following a description of methodological
concerns and challenges to interpretation, we review the empirical evidence for gender
differences early in life in the family context. We split the discussion into two parts. We
first review research for developed western nations, where the added complication of sexselected abortions is not an important consideration. We then turn to a review of the
literature in the South Asian context, where gender differences are larger and sex-based
abortion is prevalent (Das Gupta and Li 1999). Following the review of empirical
evidence, we discuss possible interpretations, as well as directions for future research and
policy implications.
2

2. Methodological Issues
Isolating the existence and estimating the extent of gender discrimination is a
difficult task. One cannot simply conclude that differences between males and females
are due to gender-based discrimination. Consider the setting of wage differentials
between men and women. The fact that women earn less than men in the labor market
could be attributed to discrimination in the workplace (e.g., women may have a harder
time getting a job or being promoted). But as researchers have shown, part of the gap is
also due to the differing characteristics of women versus men, including the choice of
occupation, level of education, attachment to the labor market, and assertiveness (e.g.,
Altonji and Blank 1999). Some of these differences may be due to non-discriminatory
differences between men and women, but they could also be influenced by gender
discrimination earlier in life. For example, women could choose certain careers which
pay less because they value the flexibility or nature of the jobs. But women could also
earn less because of lower investments by parents in girls education.1
If the decision to enter the labor market is affected by gender, an additional
complication occurs. A selection bias arises, because nave estimates of labor market
returns are based only on those who choose to enter the labor force. To see this, imagine
that all men enter the labor force, but only low-skilled women are in the labor force. A
comparison of wage differentials between men and women would reveal that women are
paid substantially less than males. While it is accurate to say that women in the labor
force are paid less than men on average, it is unclear whether this is due to discrimination
in wages, discrimination in earlier investments by parents (and society), or simply a
function of selection bias into the sample of labor force participants.
Analogous to the labor market example, this section provides a simple structured
framework for outlining the various problems in pinpointing the occurrence of
discrimination in families and correctly interpreting any observed differences between
sons and daughters.
1

Discrimination could also operate on a larger societal level due to societal norms on gender roles.

2.1 Identifying gender discrimination

A simple model for gender discrimination is given by the following equation:

where y is the outcome of interest (e.g. parental investment) for child i, Female is an
indicator that takes on the value 1 if the child is a girl, X is a vector of additional factors
which influence the outcome, and is an error term. Since the variable Female is binary,
the coefficient captures the average difference in the outcome between daughters and
sons, controlling for the variables contained in X.
To begin, consider the case where having a son or a daughter is random. At first
glance, this might seem plausible, since whether one has a boy versus a girl seems largely
outside of parents control. However, this is only the case if:
(i) child gender is biologically unrelated to parental characteristics
(ii) parents do not have sex-selective abortions
(iii) parents do not base fertility decisions on the gender of previous children.
In the case where all three conditions are met, simple comparisons of mean differences
between sons and daughters would capture the causal effect of being male versus female.
In the context of equation (1), random assignment of the Female variable implies that for
consistent estimation of , it is not necessary to control for any of the variables contained
in X as these variables will be balanced on average across gender. The only reason to
control for additional covariates is for efficiency, i.e., more precise estimates of .
2.2 Strategies to deal with non-randomness
Estimation of equation (1) is complicated by the fact that the three conditions for
randomness listed above do not always hold. To start, there is some evidence that child
gender could be biologically related to parental characteristics. Biologically, older women
and women of different races are slightly more likely to have daughters (Almond and
Edlund 2007). Additionally, certain animal species appear to have adapted to regulate

offspring sex to adjust to prenatal conditions. Charnov et al. (1982) and Norberg (2004)
argue that similar mechanisms might be present in humans. To account for observable
biological differences, controls for characteristics such as maternal age and race can be
added to the regression. In practice, these controls generally do little to alter estimates.
However, it is more difficult to account for evolutionary biology explanations, and most
researchers ignore this possibility.
Abortions also create a problem for identification of gender discrimination. As we
will discuss later, parents of daughters are more likely to have abortions in many Asian
and developing countries. This in and of itself is evidence of gender discrimination.
Unfortunately, it complicates estimation for other outcomes, since what we observe is a
non-random sample of sons and daughters. To understand the selection bias problem,
suppose parents who have the most bias against daughters abort girls but that unbiased
parents do not abort based on gender. When comparing later outcomes we are left with
selected samples of sons versus daughters. By construction, the sample of children who
are not aborted includes fewer daughters born to gender biased parents compared to the
overall fraction of gender biased parents. Moreover, the characteristics of gender biased
parents may be correlated with investments. For example, if gender biased families who
abort daughters are poorer on average this will make it appear that sons receive fewer
investments on average. Research looking at developing countries therefore controls
extensively for as many parental characteristics as possible. Additionally, in the presence
of sex-selective abortions, any estimated effect is often argued to represent a lower
bound. This is because the parents most biased against girls are less likely to be observed
having daughters and often more likely to have fewer resources.
There is no evidence for widespread selection bias due to abortions within North
America.2 For this reason, the North American context is somewhat simpler to analyze.
However, as we will discuss later, in both developing and developed countries fertility
decisions are a function of the gender of previous children. To understand how this
creates a selection bias, consider the stylized example depicted in Figure 1.

Some subpopulations in the U.S., including first generation immigrants from Asian countries, show
evidence for sex selective abortions (Almond et al. 2009, Abrevaya 2009).

Imagine that half of the couples in the population are gender neutral and want to
have two children, regardless of the gender mix, while the other half of couples are
gender biased and stop having children following the birth of a son or two children,
whichever comes first. For simplicity, assume no sex-selective abortions. In our example,
couples have a maximum of 1 unit of investment they can make in their children, which
they can split among their children however they want. Parents can also choose not to
invest in any of their children. Suppose gender neutral couples split the maximum
investment equally between their two children regardless of gender (0.5 units of
investment for each child), while those who are gender biased invest the maximum in
their son but invest nothing in daughters (1 unit of investment in the son, and 0 units of
investment in any daughters).

Type of family:

Gender neutral

1st child:

1st child, 2nd child:

BB

Gender biased

BG

GB

GG

GB

GG

Figure 1: Hypothetical fertility patterns for gender neutral and biased parents.
Notes: B and G indicate boy and girl children, respectively. Gender neutral couples
have two children, regardless of child gender, while the gender biased parents stop after
they have a boy or two children, whichever comes first.

In this example, first-born females receive 0.25 units of investment on average

and first-born boys receive 0.75 on average (i.e., = -0.5). But if one were to compare all
boys versus all girls in our sample, the estimate of would be approximately -0.38,
which is less than -0.5. This is because by including all children, we are oversampling
unbiased couples relative to those who stop having children following a boy. While our
population of parents is evenly split between each type of couple, the sample of children

shifts the weight towards those who are unbiased. Since these types of families do not
discriminate, they push down the estimate of .
The selection problem is readily seen by considering families with exactly two
children. Those households that had a first born son and stopped having children are left
out of the sample entirely. Thus, second born females receive an average investment of
0.33, while second born males receive an average investment of 0.67, leading to an
estimated of -0.33. While it would be accurate to state that second born females on
average receive investments which are half as large as males, it does not accurately
estimate the magnitude of gender discrimination.3 This is because the estimate is based
on a selected sample, where the selection itself is related to the effect we are trying to
estimate. This is particularly problematic for studies which attempt to condition on the
number of children or gender mix in a family.
In addition to looking across families, another method is looking at differences
within the family. This family fixed effects approach is identified by comparing siblings
within a family. There are several drawbacks to this method. First, it fails to include
families with no variation in child gender. Using the example depicted in Figure 1 above,
households that stopped having children after a son (and families that have two sons or
two daughters) would be excluded from the estimation. This means there will be twice as
many gender neutral families with one boy and one girl compared to gender biased
families in the sample. Thus the two family types are represented in the estimation
sample at a proportion that is not representative of the overall population. Again, for the
same reasons described above, this approach will underestimate gender discrimination in
our sample, with equal to -0.33 in our simple example.
One strategy to address this type of selection bias is to look at outcomes for firstborn children only. This gets around the issue of endogenous fertility stopping rules. In
the example described in Figure 1, the estimated effect would be consistently estimated

Of course the problem of selection due to son-based stopping rules is further exacerbated when
considering higher order births of 3 or 4 children.

as -.5 because the gender of the first child is like a random coin flip.4
Gender based stopping rules can affect investment, even if gender biased couples
split their investments equally among their children, regardless of gender. This is because
larger families must share resources among more children. Gender biased couples that
stop having children after a first-born boy will concentrate all of their investment on one
child, rather than needing to split the investment among two children. It is important to
recognize that gender based stopping rules do not affect the consistency of estimates for
first-born children samples. Rather, the fact that first-child gender influences subsequent
fertility is simply part of the mechanism through which differential treatment occurs.
The selection problems created by abortion and fertility stopping rules are similar
in many respects. They both induce a bias in the estimated effect of being a girl versus a
boy. In many cases, the estimate can be argued to plausibly represent a lower bound. One
difference is that a mothers abortion history, and more importantly the gender of an
aborted fetus, is rarely observed in datasets. Rather, one makes inferences about abortion
and female infanticide by comparing infant sex ratios.
In the real world, the strategy to focus on first-born children works as long as
biological determinants are not important factors and parents do not selectively abort. As
discussed above, it is unclear how large a role evolutionary biology plays and difficult to
account for such a mechanism. But at least in developed countries there is no evidence
for sex-selective abortion, which makes the first-born child strategy especially attractive.
Of course, for small datasets, sometimes researchers must use children of any parity to
increase precision.
Our discussion of methodological issues is meant to highlight the difficulties in
testing for gender discrimination due to the fact that child gender may not be random. In
what follows, we focus on the main substantive findings found in the economics
literature. We remind the reader upfront to remember the various issues created by
biology, abortion, and fertility, as we will spend less time on these issues in what follows.

One limitation of studies looking at first-born children only is that they do not estimate effects for higher
order births.

The lack of sex-selective abortion makes the North American context easier to analyze,
so that is where we begin our discussion. We then turn to the South Asian context, where
gender discrimination is more prevalent and has arguably more severe consequences.

3. Gender Differences in Developed Countries

We begin our discussion of gender preferences within families by looking at
studies in developed countries, most of which use U.S. data. While our survey is broad, it
is not comprehensive, but rather focuses on some of the recent research within
economics.5
In the U.S. and other Western countries, the ratio of male versus female births is
consistent with the naturally occurring ratio based on biology.6 This balanced gender ratio
is taken as evidence that parents do not influence their childs gender through sexselective abortions. When compared to the unbalanced gender ratios observed in many
developing countries (which are indicative of either sex-selective abortions or female
infanticide), gender discrimination in Western families is a much smaller concern.
However, it would be a mistake to conclude that there are no differentials based on
gender in developed countries. As we document in this section, differences based on a
childs gender influence a variety of outcomes, including stated preferences; fertility and
adoption decisions; marriage, divorce, and custody; time use in the labor market and with
children; and educational investments.
3.1 Stated Preferences
The starting point for ascertaining whether gender preferences exist is to ask
people directly about their preferences for sons versus daughters. From 1941 to the
present, the Gallup organization has intermittently collected survey data on this topic.
While the wording of the question has changed over time, respondents were asked a
5

For example, there is a large literature in sociology, including influential work by Philip Morgan and coauthors (1988, 2002, 2003).
6
Biologically, male births are slightly more likely than female births, but male offspring also have lower
survival rates (Sen 1990).

question similar to Suppose you could only have one child. Would you prefer that it be a
boy or a girl? There is a consistent pattern in responses over time in the data: Americans
say they would prefer to have a son over a daughter. The time trend in survey responses is
plotted in Figure 2.7 In 2011, the most recent survey year, 40 percent of respondents
stated a preference for a boy compared to 28 percent of respondents who preferred girls
(the remaining 32 percent either had no opinion or were unsure). These stated preferences
are remarkably stable over time. In 1941, the first year Gallup polled individuals on this
topic, 38 percent said they preferred boys compared to 24 percent who preferred girls,
percentages which are not that different from those in 2011.
The most recent 2011 survey conducted by the Gallup Poll provides further
insight into what types of individuals are most gender-biased. Men prefer sons over
daughters by more than a two-to-one margin (49 percent boy to 22 percent girl), while
women are evenly divided in their preferences (31 percent to 33 percent). Lower
educational attainment is also associated with a strong preference for sons, with a big
difference in stated preferences for those with no more than a high school degree (44
percent versus 25 percent), but essentially no difference for those with an advanced
degree (32 percent versus 33 percent). Respondents who identify themselves as
politically conservative express a strong preference for sons (41 percent to 25 percent),
while self-identified liberals are evenly split in their preferences (36 percent to 37
percent). Interestingly, as individuals get older, they prefer boys less, perhaps indicating
that lifes experiences (including raising girls) help temper any bias.
The survey results provide a preliminary indication that, at least on average,
segments of the American population have a bias towards sons. However, whether and
how people act on these stated preferences are arguably the more important questions.
Indeed, economists have traditionally been more interested in studying revealed
preferences instead of stated preferences. We therefore turn to a variety of choices made
by parents to see if these stated preferences translate into different consequences for boys
versus girls.
7

Surveys were conducted on nationally representative random samples of adults (18 years or older). The
surveys typically have 1,000 or more completed interviews per year, which implies a conservative sampling
error of plus or minus three percentage points.

10

3.2 Fertility and Adoption

While there is no evidence that parents in Western countries selectively abort
based on gender, as we show in this section, a childs gender does affect parents'
subsequent fertility decisions. These fertility decisions can have long lasting effects on all
of the children in a family as it changes family composition and the amount of available
resources.
If parents have a preference for sons, then we would expect to observe fertility
stopping rules based on the gender of existing children. A simple example of this
behavior was provided in Figure 1. The intuition for why a son-based stopping rule
increases fertility is that parents who want a son, but have not yet had one, will be more
likely to continue having additional children in the hopes of having a male child. Of
course, some parents may also prefer daughters and have a daughter-based stopping rule.
The point of looking at aggregate data is to see which family type dominates.8
Consistent with a son preference, Dahl and Moretti (2008) find that in the U.S.
during the period 1960-2000, married women were more likely to have additional
children if their first born was a daughter versus a son. They estimate that first born
daughters caused approximately 5,500 more births per year. When the sample is restricted
to women still in their first marriage, this effect on fertility following a first born daughter
is even greater. Andersson et al. (2006) find that in Nordic countries, looking at all
mothers (including cohabiting and single mothers), there is no effect on fertility as a
function of the first childs gender, but that having two sons is more likely to lead to a
third child relative to having two daughters in Denmark, Norway and Sweden (with the
opposite being true in Finland). As discussed in the previous section, the effect of two
children of the same gender on the decision to have more children is trickier to interpret
since the fertility decision to have a second child may be a function of the gender of the
first child.
Even with a fertility rule favoring boys within couples, Ichino et al. (2011)

Empirically, the most dominant preference seems to be for a mix of child gender among offspring
(Angrist and Evans 1998).

11

illustrate that competing factors (such as divorce) may reduce the actual impact on
fertility. When looking at all women regardless of marital status, Ichino et al. (2011) find
that having a first born boy increases the probability of having more children relative to
having a first born girl. They argue this is due to the increased marital stability caused by
sons and the fact that average fertility is higher among married versus single women. We
discuss the effect of child gender on martial stability in the next section. Their study uses
data from the U.S., the U.K., Italy, and Sweden.
While the relationship between a childs gender and fertility is consistent with a
son-based stopping rule, other interpretations are also plausible. For example, daughters
may be less costly to raise or result in a more positive childrearing experience for parents.
If children are normal goods, then relatively-inexpensive daughters could encourage
higher fertility, even in the absence of gender preferences.
The finding of increased fertility for married women following the first birth of a
daughter implies that these daughters are more likely to reside in larger families. Even if
parents treat all of their children equally, family size may have its own negative effect on
a childs upbringing. Becker (1981) theorizes that as families choose to have more
children, each child receives less investment, resulting in a quality-quantity tradeoff.
While a correlation exists between family size and lower educational outcomes (Black et
al. 2005, Black et al. 2010, Deschnes 2007, Hanushek 1992, Mogstad and Wiswall
2010), studies attempting to show a causal link reach different conclusions. Mogstad and
Wiswall (2010) estimate the marginal effect of additional siblings and find that education
is lowered for children when they belong to larger families. Black et al. (2005), however,
find the effect of family size disappears after controlling for birth order.
The adoption market, which accounts for 2.5 percent of all children within
families in the U.S., provides an additional setting to examine revealed preferences. The
adoption market is unique in that parents can choose the gender of their child. While the
fertility evidence is broadly consistent with a preference for boys, Baccara et al. (2010)
find that adoptions run counter to this finding. Potential parents seeking to adopt are more
likely to apply for girl babies. When the authors convert this preference into a dollar
amount, they conclude that parents are willing to pay approximately $19,500 more in
12

adoption costs for a girl versus a boy baby. However, those who are interested in adoption
are a select subset of individuals, making it unclear to what extent such preferences are
present in the larger population.
3.3 Family Structure: Marriage, Divorce, and Custody Channels
Dahl and Moretti (2008) document that compared to first born sons, first-born
daughters in the U.S. are 3.1 percent less likely to be living with their father, a gap that
has persisted in every decade from 1960 to 2000. They estimate that approximately
52,000 more first-born daughters (and all their siblings) would have had a father living
with them had they been first-born sons instead. Choi et al. (2008) find similar results in
Germany, where a first-born son is 25 percent less likely to have a father leave the
household compared to a first-born daughter.
The gender differential in having an absentee father can be decomposed into three
factors: initial entry into marriage, divorce, and custody. We first explore each of these
channels in this section. Then we review the literature which discusses why growing up
without a father might negatively impact children.
Differences in marriage rates after the birth of a boy or a girl child is the first
channel which could affect the chances a father is present in the household. Dahl and
Moretti (2008) find that having a first born daughter increases the probability a woman
never marries by 1.4 percent. Additional evidence for the U.S. and other countries
confirms that mothers of daughters are less likely to ever have been married (Ananat and
Michaels 2008, Choi et al. 2008, Lundberg and Rose 2003).
Interestingly, differences in marital status based on child sex occur even before
the baby is born. Using birth certificate data from California, Dahl and Moretti (2008)
find that mothers who have an ultrasound during their pregnancy and have a daughter
versus a son are less likely to be married at the time of birth. With high accuracy,
ultrasounds reveal the gender of a fetus. Hence, they interpret their results as indicative
that child gender matters for shotgun marriages. Their data comes from a period (1989
1994) when ultrasounds were not as common as today. There is no effect for women who
do not have an ultrasound, suggesting that it is knowledge about the gender of the child
13

driving the results. Lundberg and Rose (2003) further find that having a son versus a
daughter speeds the transition to marriage if biological parents are unmarried at the time
of birth.
The second channel by which a daughters chances of living with her father are
reduced is through divorce. A number of sociologists find a positive correlation between
having a daughter and divorce (Spanier and Glick 1981, Morgan, Lye and Condran 1988,
Katzev, Warner and Acock 1994, Morgan and Pollard 2002). In contrast, another set of
papers in sociology using data from a variety of countries finds no statistically significant
difference in divorce based on child gender (Mauldon 1990, Devine and Forehand 1996,
Wu and Penning 1997, Diekmann and Schmidheiny 2004). The main reason the second
set of studies do not find a significant effect appears to be due to small sample sizes. For
example, Diekmann and Schmidheiny find no divorce effect in eighteen countries, but
their sample sizes are around 2,000 for each country and therefore they are unable to
precisely identify small effects. Using large census samples, Dahl and Moretti (2008) find
that first born daughters increase the probability of divorce by 1.3 percent over the period
1960-2000. The effects in the more recent years of their study, while statistically
significant, are much smaller, pointing out the need for large datasets to identify these
relatively small effects. In the economics literature, many other researchers have
documented similar divorce effects with relatively large datasets (e.g., Ananat and
Michaels 2008, Bedard and Deschenes 2005).
Higher paternal custody of sons is the third channel that causes boys to be more
likely to live with their fathers. Dahl and Moretti (2008) and Mammen (2008) both find
that boys in the U.S. are relatively more likely to live in single-fathered homes than
daughters. While it is true that mothers are more likely to gain custody of their children
after a marital dissolution, when fathers do obtain custody, they are almost 3 percent
more likely to obtain custody of their sons versus daughters.
The relative importance of the marriage, divorce, and custody channels have
changed over time (Dahl and Moretti 2008). While divorce was the biggest reason for the
gender gap in having a resident father in 1960, by the year 2000 the custody channel
accounted for the majority of the difference. Given the large secular changes in family
14

structure over time, it is difficult to compare the amount of gender bias through these
various channels over time. But what one can conclude is that differential access to
fathers for sons versus daughters has remained sizable over time.
3.3.1 Consequences of Growing Up in a Fatherless Household
This literature shows that girls are less likely to live with their fathers because of a
lower probability of marriage, an increased probability of divorce, and a decreased
probability of paternal custody. We now turn to the literature which documents that
growing up without a father present in the household has a variety of negative
consequences for children.
The first concern is that single-mother households have significantly less income
than households with a male figure, implying that girls are disproportionately likely to be
living in a low-income household. As Dahl and Moretti (2008) document, families with
first-born daughters have higher poverty rates and less available family income. They
find that family income is reduced by 50 percent and the odds of being in poverty
increase by 34 percent due to the first-born daughter effect. Mamman (2008) shows that
girls are 5 percent more likely to live in single family households, which have an average
income that is significantly less than households with a father figure, whether or not the
comparison is to single father households or married households.
While first born daughter households clearly have less income, whether the effect
operates through the divorce channel is less clear. Ananat and Michaels (2008) and
Bedard and Deschnes (2005) use a first born daughter indicator as an instrument for
divorce. Using this instrumental variables strategy, they find that divorce has no effect on
a womans mean household income. Bedard and Deschnes (2005) find that, if anything,
these households are slightly better off, with the women earning $5,000 more annually as
a result of increased work hours. Ananat and Michael (2008), however, find that while
there is no difference in average income, a divorce due to a first born daughter does affect
the probability of a woman having a very low or very high income. Of course, for these
findings to be interpreted as causal, it must be that a first-born daughter does not have
any other direct effects on household income.

15

In addition to lower household income, another area of concern for daughters

living in single-mother households is the role of paternal involvement and child support.
There is limited evidence that divorced and unmarried fathers discriminate against their
daughters. Lundburg et al. (2007) find that sons born to unmarried parents are more likely
than daughters to receive the fathers surname, but that there is no difference in the
fathers monetary and time investments one year after birth for unmarried fathers.
Mammen (2008) finds no support for the idea that mothers of sons are more likely to
receive child support payments, whether or not the parents were never married or
divorced.
More generally, related research documents that children in single-mother
households are more likely to drop out of high school, become teenage parents, and
experience unemployment (e.g., McLanahan and Sandefur 1994, Haveman and Wolfe
1994, Case and Paxson 2001).
3.4 Time Use: Labor Market Participation and Time Spent with Children
Do families with both parents present allocate time differently in response to a
childs gender? Several studies have shown that both fathers and mothers participation
in the labor force depends on a childs gender. There is also some evidence that fathers,
but not mothers, spend more time with their sons versus their daughters.
Studies in the U.S. and Germany find that fathers increase their labor supply more
if they have a son versus a daughter (Lundberg and Rose 2002, Choi et al. 2008). Both of
these studies find that having a child increases a mans hourly wage and hours of work,
but that the increase is greater in response to having a son. For example, Lundberg and
Rose (2002) find that men with a first born son had wages which were 5.3 percent higher
for some cohorts and that fathers worked 63 more hours per year if they had a son for
some cohorts. However, using time diaries in the U.S., Lundberg et al. (2007) fail to find
an effect of fathers labor supply in married couples when comparing families with boys
versus girls. In related work, Pabilonia and Ward-Batts (2007) find that Asian fathers in
the U.S. work fewer hours relative to white fathers if they have a son instead of a
daughter, which the authors interpret as son preference. In these studies, education

16

appears to be a critical factor in determining a fathers labor market response to having a

son or daughter.
Lundberg and Rose (2002) find no difference for mothers labor supply in
response to child gender. In contrast, Ichino et al. (2011) find that mothers in U.S., U.K.,
Italy, and Sweden with first born sons are less likely to work in a typical week and are
more likely to work fewer hours. They attribute this to increased marital stability, and
thus a reduced need for income and increased fertility. Their estimates suggest that in the
U.S. 50,000 women (ages between 18 and 55) with at least one child do not work simply
because their first born was a male. Lundberg (2005) finds that single mothers spend
more time in market work and less time with their child if they have a son relative to a
daughter.
A consistent finding in the literature is that fathers spent more of their non-work
hours with sons compared to daughters (Yeung et al. 2001, Lundberg et al. 2007,
Mammen 2008, Price 2008). Yeung et al. (2001) find that in intact families in the U.S.,
fathers spend 18 more minutes per weekend day in playtime and companionship with
their sons relative to daughters. Lundberg et al. (2007) also find that fathers spend more
time with sons when comparing families of similar size. They find the increased time
spent with sons is compensated by a reduction in the fathers leisure. Mammen (2008)
observes that siblings benefit by having brothers since fathers spend approximately 23
more minutes in a two child family with at least one son, relative to two daughters.
However, the increased time spent with daughters comes from an increase in watching
television together and therefore may be of limited value.
Lundberg et al. (2007) suggests that if the increased paternal time with sons stems
from sons requiring more parental time, then one would suspect that single mothers also
spend more time with their sons relative to daughters. However, she finds the opposite;
single mothers with only one child spend more time with daughters than sons. Unlike
fathers, this increased time that single mothers spend with daughters stems from
decreased labor market participation as opposed to leisure.
3.5 Additional Areas of Gender Differences

17

In addition to the empirical evidence on family structure and parental time,

researchers have reviewed prenatal care, household activities, and bequests as areas of
potential gender differences. Lhila and Simon (2008) find no difference by prenatal care
by gender for U.S. parents who choose to go to full term, including for Indian and
Chinese immigrants. Time diaries of older children suggest that household chores are
more likely to be done by daughters, and that girls and boys are more likely to do chores
that fit in traditional gender roles (e.g., cooking and cleaning as opposed to yard work)
(Gager et al. 2009, Bianchi and Robinson 1997).
One final area which deserves brief mention is educational investments. At early
ages in life, investments in human capital are a decision that is made by parents on behalf
of their children. If households have a limited amount of resources and prefer to invest
more in sons, we would expect that girls with brothers are worse off and brothers with
sisters to be the most favored. However, Butcher and Case (1994) find conflicting results.
Analyzing data in the U.S. from 19201965, they find that a daughter with a brother
receives half a year more education than if she had a sister. Conley (2000) finds the
opposite effect, where girls educational attainment is lowered by the presence of
brothers. These conflicting results could be the result of selection based on family size, or
the way in which joint investments are made in a family. Focusing just on the arguably
random gender of the first born child, Dahl and Moretti (2008) find that siblings with a
first born sister have lower education on average compared to siblings with a first born
brother.

4. Gender Differences in Developing Countries

Unlike in developed countries, the number of females relative to males is much
lower in the developing world, implying that gender discrimination has larger effects on
female mortality. Hence, health investments are an important dimension along which
girls are discriminated against in the developing country setting. We first examine
differences in health investments that might result in excess mortality among women.
Later in this section, we examine gender differences in household inputs such as
educational investments, parental time and food expenditures.
18

4.1 Differences in Health Investments: Excess Female Mortality

In 1990, Sen (1990) highlighted the incredibly low female to male ratios in
countries like China (.94), India (.93) and Pakistan (.90) relative to the gender ratios seen
in developed countries (1.05) and estimated that there were some 100,000 million
missing women. At all ages, womens biological ability to survive is higher than mens
(Sen 1990), indicating that women must be missing women due to social and cultural
practices. Though a proper counterfactual is difficult to determine (due to such factors as
countries different life expectancies, incidences of illnesses, and genetic differences), the
large differences in gender ratios suggest unnatural numbers of missing women in
developing countries (Sen 1990, Anderson and Ray 2010). The potential negative
consequences of missing females go beyond the individual disutility of the women
directly affected. Ebenstein and Jennings (2009) discuss the lack of women in the
Chinese setting and estimate that 10.4 percent of Chinese men born between 1980 and
2000 will never marry. Using historical and current trends, they argue that this could lead
to potential lack of care for the elderly (in particular unmarried men who do not have
offspring), increased crime, and increases in prostitution and sexually transmitted
diseases. Using province level implementation of Chinas one child policy, Edlund et al.
(2007) argue there exists a causal link between rising male ratios in China and the
incidence of crime. Excess female mortality can occur via sex-selective abortions, which
is prevalent in this setting, as well as through post natal discrimination in health
investments. Below, we summarize the findings of papers that examine each.
4.1.1 Sex-Selective Abortions
The assumption that parents do not influence their childs gender is no longer
innocuous in the developing country setting, especially in South Asia. Evidence from
birth rates suggests a relatively high incidence of sex-selective abortions in Asia.
Following the availability of prenatal sex diagnoses in these regions, the gender ratio at
birth began to skew towards boys, especially for children in higher birth orders,
suggesting sex-selective abortions that favor male children (Bhalotra and Cochrane 2010,
Ebenstein et al. 2012, Arnold et al. 2002). Bhalotra and Cochrane (2010) estimate
480,000 girls were aborted in India over the decade 19952005. Their estimates run
19

counter to the popular belief that households desire at least one son, but find that families
tend to favor two boys and a girl. Arnold et al. (2002) find that even within India, large
regional variation exists. In northern states,(like Punjab, Haryana and Rajasthan, which
are known for greater son preference), in the late 1990s when prenatal sex tests were
becoming prevalent, couples who had a first born daughter were more likely to abort their
next conception than if they had a first born son. Even though over 94 percent of women
receiving an ultrasound do not report an abortion, they still estimate 106,107 as a lower
bound for sex-selective abortions of female fetuses each year (this is despite the fact that
revealing the sex of the fetus is illegal in India). They further find that the male to female
ratio in northern states for couples after an abortion is 1.58 and for last born children is
1.88much higher than the country average of 1.07suggesting that many of these
couples had an abortion in hopes that the next child would be a son.
Ebenstein (2010) documents the extent to which the one-child policy in China
increased sex-selective abortions. He finds that in areas where the policy was more
strictly enforced, male-female sex ratios are higher and is driven by an increase in the
prevalence of sex selection within the first two births. He further finds that the interval
between births is longer for sons than daughters, suggesting those couples who have a
son had past failed pregnancies, and that the fraction of male children increases for higher
order births. Ebenstein et al. (2012) further find that the diffusion of ultrasound access in
China led to an increase of girls death in the first four weeks following birth, suggesting
gender differences in prenatal investments.
The phenomenon of sex-selective abortion makes the detection of gender
discrimination even more difficult; we can no longer compare differences between girls
and boys, even at birth, since gender is no longer a random variable. Still there are
researchers who have argued that sex-selective abortion is rare for first born births in
India, and hence the gender of the first born is still an exogenous variable and allows for
valid comparisons., Bhalotra and Cochrane (2010) find that more educated and wealthier
households are more likely to undergo sex selective abortions. This suggests that girls are
more likely to be born into richer families This is likely to create an upward bias in the
estimated gender discrimination against girls since there is a positive correlation between

20

education and wealth and sons. On the other hand, to the extent that we believe these
households that perform sex-selective abortions would have discriminated more against
girls relative to boys, our estimates may be underestimating discrimination.
Areas where sex-selective abortions are prevalent typically also exhibit son
preference-based fertility stopping rules. In line with families practicing son preference
based fertility stopping rules, Arnold et al. (2002) find that girls are more likely to live in
larger families. Barcellos et al. (2012) show that while the youngest children in a family
are 51 percent likely to be male when aged zero to five months, youngest children aged
54 to 59 months are 58 percent likely to be male, suggesting that the households decide to
have another child if their previous child was a girl. This fertility stopping pattern has
also been linked to a significant fraction of Indias excess female mortality in the recent
work of Rosenblum (2012). Many of the results in this chapter are suggestive of
increased discrimination at higher birth orders, when the pressure of having a son is
greater (e.g., Jayachandran and Kuziemko 2011).
4.1.2 Missing Women after Birth: Health and Shocks
While much attention is given to sex-selective abortions, the missing women
phenomenon is not limited to sex ratios at birth. Anderson and Ray (2012) document
missing women in India, China, and Sub-Saharan Africa (hereafter SSA) by comparing
male-female death ratios at different age groups with those in developed countries and
find that the majority of missing women occur at later ages as opposed to at birth.9 They
show that unlike in developed countries, India and SSA have a male to female death ratio
that is under one until women reach their mid-30s. Of the total missing women in 2000,
less than 11 percent was due to prenatal factors in India (less than 40 percent in China).
Though SSA does not appear to have missing women at birth, the region has a total of
approximately 1.53 million missing women, making it comparable to the total missing
women in India (1.71 million) and China (1.73 million).

In general, creating the correct counterfactual for male-female death ratios is difficult and perhaps using
developed countries as a comparison is not an exact counterfactual. Thus, while these numbers may not be
wholly accurate due to ambiguous counterfactuals, they do provide strong evidence of a significant number
and trends of missing women.

21

The reason for so many missing women in later ages can be due to a host of
factors. While one mechanism may be direct discrimination (e.g., health care differences
or violence against women), it is also possible that there are genetic gender differences in
the incidence of illnesses, or cultural norms or poverty that lead to nuanced gender
differences in illnesses and death. While it is clear that across all age groups the relative
death rate of women to men is higher in developing countries (see Anderson and Ray,
2012), it is difficult to pin down the mechanism as to why this difference occurs and how
much of the phenomenon can be attributed to gender discrimination.
Researchers have made progress towards understanding this phenomenon by
examining shocks to income and testing for a gender specific response. Under equal
treatment, barring any genetic pathways by which women might react differently to
economic shocks, we should find that men and women are affected similarly by shocks to
income. However, using data of rural Indian children born between 1961 and 1964, Rose
(1999) shows the female-male survival ratio increases during times of favorable rain fall
(i.e. when economic conditions are more favorable). Miguel (2005) looks at rainfall data
in Tanzania in the 1990s and finds that in times of poor rainfall shocks, more witches
(nearly all of whom are elderly women) are accused and killed. These studies suggest that
when times are economically poor, discrimination against women is likely to be
exacerbated.
Anderson and Ray (2012) decompose male-female death ratios in developing
countries by disease, and find that at younger ages, infectious, nutritious and reproductive
ailments cause excess female deaths, and at older ages, chronic and degenerative diseases
dominate. Perhaps suggesting fairly high levels of discrimination is the simple statistic
that in India, a significant number of excess female deaths occur from injuries, with
approximately 30,000 extra female deaths due to intentional injuries. Observing infants,
Khanna et al. (2003) estimates that in Delhi, India from 1997 to 2001 mean infant
mortality was 1.3 times higher among females than among males. While twice as many
infant girls less than a year old died from diarrhea (responsible for 22 percent of infant
deaths overall), a relatively preventable illness, there were no significant gender
differences in deaths that are less preventable (birth asphyxia, septicemia, prematurity,

22

and congenital anomalies).

While Khanna et al. (2003) provide suggestive evidence of parental
discrimination by gender by comparing mortality on the basis of their preventability,
others have attempted to estimate direct differences in health care investments for boys
versus girls. Asfaw et al. (2010), using the 2004 Indian National Sample Survey, find that
for children under nine, boys are much more likely to be hospitalized than girls, and that
though families use similar amounts of income and savings in health expenditures, they
are more willing to use drastic financial measures -- such as borrowing and selling of
assets -- for boys health expenditures. This gender gap in health care is higher for poorer
families, where budget constraints are tighter. While this suggests families discriminate in
health care by gender, alternative explanations are possible, including different incidences
of illnesses by gender or different household composition for households with boys
versus girls.
Bharadwaj and Nelson (2012) investigate the extent of differential health care
treatment even before birth (for the subset of children who are born) in India using data
from the 1990s and 2000s. They find that mothers who have sons are 1.8 percent more
likely to receive prenatal care at least twice and significantly more likely to receive a
tetanus shot. They estimate that the gender gap in receiving a tetanus shot accounts for
between 410.5 percent of excess female neonatal mortality in India. In northern India,
these differences are even greater, with women being 16 percent more likely to deliver
their babies outside of their home when having a boy. Bharadwaj and Nelson (2012) find
similar suggestive evidence in China, Bangladesh, and Pakistan.
Barcellos et al. (2012) examine children in 1992 in India (before sex selective
technology became prevalent) at a young age (zero to 15 months), before parents have
another child, as a way to overcome the endogenous fertility choice of parents based on
previous childrens gender. They find that boys are more likely to be breastfed longer,
given vaccinations, and given vitamin supplements, although similar to other studies,
they find no differences in anthropomorphic indicators (weight for height z scores) by
gender. They find no evidence that these differences stem from differential gender needs
of infants, and argue that these results are due to discrimination by parents based on child
23

gender. As none of the above studies directly control for sex-selective abortions, these
estimates may be considered lower bounds, assuming that including such abortions would
show higher levels of discrimination.
Jayachandran and Kuziemko (2011) find that girls in India are breastfed for
approximately .37 months less than boys and argue that this is a response to sonpreference fertility rules given the different patterns exhibited by birth order.
Breastfeeding inhibits fertility, and so households that have not met their desired number
of sons will stop breastfeeding children earlier in hopes that the next child will be a son.
Breastfeeding is essential in these settings. They estimate that the reduced time spent
breastfeeding could account for 8,000 to 21,000 missing girls in India each year,
explaining approximately 9 percent of the gender gap in child mortality.
4.2 Differences in Other Inputs
While health is a major factor along which discrimination favors boys over girls
in developing countries, there are other dimensions along which parents favor boys. Lack
of data has made gender differences in household expenditure in developing countries a
difficult task to document. In this section, we discuss education, food, and parental time
as possible areas of gender discrimination in the family.
4.2.1 Education
Girls education has historically been significantly lower than boys education,
though rapid advances have been made in the past two decades to reduce the gender gap.
In the 1990s, a series of papers discussed the large gender gaps in educational
enrollments (e.g., Filmer and Pritchett 1998) but as Figure 3 shows, drastic reductions
have been made in the gender gap in school enrollment over time in developing
countries, at least at the primary level. Though progress has been made to close the
gender gap, school dropouts are still more likely to be girls, especially in secondary
school (UNESCO 2012, ASER 2010). Azam and Kingdon (2011) use data from the 2005
India Human Development Survey and observe no evidence of gender bias in school
expenditure or enrollment at the primary school level (age five to nine). However, they
do find a difference in education expenditure by ages 1014, and a difference in
24

enrollment (though not on expenditure conditional on enrollment) by ages 15 19. One

of the reasons behind the different expenditures is that boys are more likely to be sent to
private schools, which is generally considered to be of higher quality than free public
schools.
An underlying reason for the different investments in education may be due to
different requirements of boys and girls. Many have argued for gender specific policy
proposals to attract female students, such as recruiting more female teachers, separate
bathroom facilities for girls, or reducing distances to schools. However, evidence on
whether reductions in such barriers disproportionally increase female educational
attainment is sparse. Burde and Linden (2012) find that reducing the distance to schools
in Afghanistan by building local village schools drastically increases older girls
enrollment rates and test scores, which they speculate is driven by increased safety for
girls. Though both genders experienced increased enrollment and test scores, they find
that girls are more sensitive to distance, leading to higher effects for girls to the point of
virtually closing the gender gap and reducing the test score gap in enrollment by one
third. However, a significant number of other papers have found little evidence of
enrollment increases as a function of distance to school (Filmer 2007, Handa 2002).
Muralidharan and Sheth (2012) find that girls with female teachers do experience a very
small increase of approximately .025 standard deviations in test scores than when they
matriculate with male teachers, but find limited evidence of increased attendance when
taught by a female teacher, suggesting a limited role of parents reacting to female
teachers. Thus, while there is some evidence that girls face differential requirements to
succeed in schools, overall not much is known about the extent to which this drives
observed differences in educational attainment and investment.
One leading theory for why son preferences exist is that the household rate of
return for boys is higher than for girls. Differences in educational attainment may be one
factor behind the lower rates of returns to girls, though it is difficult to disentangle the
direction of causality. Differences in educational investment may stem from different
rates of return to human capital by gender, or difference in educational investments may
lead to different rates of return by gender.

25

Jensen (2012) finds that increasing awareness of employment opportunities for

women resulted in increased enrollment and body mass index of younger school-aged
girls (ages 515), suggesting that parents are willing to invest in girls more when the
potential of their future returns are improved. The study finds that recruiters advertising
employment opportunities for females in rural India closed 60 percent of the baseline
gender gap in enrollment and 1520 percent in BMI for girls aged 515, and increased
the household budget share for girls schooling for girls aged 5 15. Munshi and
Rosenzweig (2006) find that increasing returns to English combined with caste-based
networks are at the heart of increased school enrollment and investment for girls in cities
in India. The idea in their paper is that while boys occupation should conform to caste
normsdue to network effects, girls occupation choices are not similarly constrained.
Hence, when returns to English language skills increased in India, so did girls
educational investments. Jayachandran and Lleras-Muney (2010) show in a different
context (Sri Lanka) that when female adult mortality declines (specifically through
reductions in maternal mortality), schooling of girls increases, since increased life
expectancy increases the returns to education. Their study suggests that increasing returns
to education for women impacts educational investments.
4.2.2 Household Expenditures
Identifying gender differences in food purchases or other expenditure has been
challenging due to lack of detailed data on consumption or expenditures within the
household. Some of the earlier studies resort to creative ways of empirically documenting
whether resource allocation within the household is driven by gender discrimination. A
classic early study in this fashion is by Deaton (1989), who finds that in Thailand there is
a small, though statistically insignificant bias in favor of boys in household expenditure.
Deaton (1989) does this by inferring that for a given level of income, adults in families
with children will spend less on goods for themselves (adult goods) in order to
purchase things that matter more for children (childrens goods). Discrimination can be
detected if for a given family size, families with boys make smaller expenditures on adult
goods than families with girls.
However, even having individual level data on expenditure and consumption does
26

not solve the problem of inferring whether gender discrimination occurs. If men and
women have different caloric needs, then gender differences in food expenditure and
consumption can occur due to such needs rather than due to discrimination. In order to
solve these issues, researchers in the past have opted to proxy for these caloric needs by
using a measure of health endowment. After accounting for this proxy, however, any
remaining differences in allocation could be attributable to discrimination. Pitt et al.
(1990) find that boys (age 12 and older) consume more food than girls (30 percent more)
in Bangladesh. They find that the households are generally averse to inequality and it is
the different caloric needs of men versus females that drive the results. This result is
supported by a lack of difference in mean calories consumed for children ages seven to
11 (since presumably for younger children, caloric needs across boy and girls could be
quite similar), though under age six, boys do receive more calories.
A different approach to this problem is to examine whether the allocation of
income within the household (to the mother or the father) affects nutrients or
expenditures on girls versus boys. Thomas (1994) finds suggestive evidence that parental
gender matters for resource allocation within the household. Using data from three
different countries, Thomas finds that resources are better allocated to girls when mothers
have better bargaining power within the household. This better resource allocation is
reflected by girls having better health outcomes, as proxied by height. Duflo (2000) finds
similar results in South Africa using exogenous variation in female pension incomes. She
finds that pensions that go to grandmothers raise the anthropometrics of young girls in the
household. No such effect is found for boys.
Another aspect of household expenditure is the opportunity cost of time of the
parents. Since parental investments are important for determining later life outcomes,
researchers have examined whether parental time inputs vary by the gender of the child.
Barcellos et al. (2012) look at young children before son preference fertility stopping
rules result in additional children and find that parents spend significantly more time with
boy children in the first year of life than with girls. They find that boys have
approximately 30 minutes more a day than girls (14 percent more time), and that the
quality of care for boys is higher. While lacking direct data on time spent with children,

27

Rose (2000) finds similar evidence by examining labor market choices of women after
the birth of a boy compared to labor market choices after the birth of a girl. She finds that
in credit constrained households, women work more after giving birth to daughters. Since
this increase in labor market hours occurs when the child is quite young, we can presume
that this implies less parental time spent with the child during her formative years.
4.3 South Asians in North America: Cultural Persistence of Son Preference
While gender discrimination exists in all parts of the world, the extent of
discrimination is arguably much greater in developing regions. To what extent are the
increased gender gaps we see in developing countries a function of living in a lower
income society? Do gender gaps in immigrant families disappear when they move to
developed countries and live in a more prosperous economic climate, or is the preference
for a son deeply entrenched in cultural and social norms that persist despite changes in
living standards? When observing immigrants in developed countries, there is suggestive
evidence of stronger sex preferences than non-immigrant populations in the country.
Higher male-female birth ratios, particularly for higher birth orders, suggest gender bias
among Asian immigrants in the U.S. and Canada (Almond and Edlund 2008, Almond et
al. 2009, Abrevaya 2009, Lhila and Simon 2008). In particular, Almond et al. (2009) find
that fertility stopping rules drive the sex composition for first generation Asian families,
while sex-selective abortions are the more predominant method for second generation
Asian families in Canada. Abrevaya (2009) estimates over 2,000 missing Chinese and
Asian Indian baby girls in the U.S. between 1991 and 2004.
In addition to fertility and abortion choices, Pabilonia and Ward-Batts (2007) find
that for married parents with an only child less than two years old, Asian men in the US
work less relative to white men when they have a son rather than a daughter. One
explanation is that there is an increase in Asian mens greater desire to spend more time
with sons, especially young sons. These studies suggest that even when removed from a
developing country context, cultural norms are persistent and son preference is evident.
Though the effects of son preference tend not to be as large as in China and India, this
could be because of a selection bias in the type of person who chooses to immigrate, as
opposed to a reduction in gender preference
28

5. Conclusion
Our paper summarizes recent economic work that documents the existence of
gender discrimination along various dimensions, and in particular differential outcomes
during early childhood in developed and developing countries. Our assessment of the
literature suggests that girls everywhere experience lower levels of parental investments
when they are younger, be this via family structure, health investments, parental time or
educational investments. While we think this plays a part in the differences in mortality
(in the developing country case) or labor market consequences we see between adult men
and women, the extent to which these differences are a causal factor in that process is
difficult to determine.
Most importantly, our chapter says very little about why discrimination, defined as
differential treatment based on gender, occurs in the first place. Why is there a preference
for sons? Research on this important question is rather sparse. Understanding the reasons
underlying preferential treatment of sons is key for crafting effective policy aimed at
reducing the gender gap in investments. Numerous competing explanations have been
posited to explain the gender preferences we observe. Theories include gender
differences in actual taste-based preferences, parents having a comparative advantage in
raising a child of their own gender, monetary returns to inputs which vary by gender (e.g.,
education), needs and requirements related to gender (e.g. increased security, nutritional
requirements, old age insurance), and costs (e.g., dowries). Though researchers have been
able to document variety of gender differences, little progress has been made in
deciphering the possible mechanisms behind the observed gender gaps.
Recent research from developing countries however, provides some evidence that
economic factors play an important role in the formation of son preference. Such research
could be translated into policy aimed at reducing gender gaps. In recent work, Qian
(2008) finds that increasing the return to female labor income while holding male income
constant increases female child survival in China. The work of Lleras-Muney and
Jayachandran (2010) also suggests that increasing the rate of return to female education
increases investments in girls. The previously discussed work of Jensen (2012) is another
example of how gender gaps in investments could potentially be reduced by simply
29

providing families with relevant information on the rate of return to education. All of
these studies have direct policy relevance in that they suggest that changing economic
opportunities for women would potentially reduce discrimination against girls when they
are young. However, more work needs to be done in establishing whether such policies
indeed translate into better childhood investments for girls.

30

References
Abrevaya, Jason. 2009. "Are there missing girls in the United States? Evidence from
Birth Data." American Economic Journal: Applied Economics 1(2): 134.
Almond, Douglas, Lena Edlund, and Kevin Milligan. O Sister, where art thou? The Role
of Son Preference and Sex Choice: Evidence from Immigrants to Canada. 2009.
NBER Working Paper Number 15391.
Almond, Douglas and Lena Edlund 2007. TriversWillard at Birth and One Year:
Evidence from US Natality Data 19832001. Proc. R. Soc. B. 274(1624):
24912496.
Altonji, Joseph G. and Rebecca M. Blank. 1999. Race and Gender in the Labor Market,
In Handbook of Labor Economics, Edition 1, Volume 3, Chapter 48, ed. O.
Ashenfelter and D. Card, 31433259. Elsevier: Amsterdam.
Ananat, Elizabeth and Guy Michaels. 2008. "The Effect of Marital Breakup on the
Income Distribution of Women with Children." Journal of Human Resources
43(3): 611629.
Anderson, Siwan and Debraj Ray. 2010. "Missing Women: Age and Disease." Review of
Economic Studies 77: 12621300.
Andersson, Gunnar, Karsten Hank, Marit Rnsen, and Andres Vikat. 2006. "Gendering
Family Composition: Sex Preferences for Children and Childbearing Behavior in
the Nordic Countries." Demography 43(2): 255267.
Angrist Joshua D. and Bill Evans. 1998. Children and Their Parents Labor Supply:
Evidence from Exogenous Variation in Family Size.
American Economic Review 88(3): 450477.
Arnold, Fred, Sunita Kishor, and T. K. Roy. 2002. "SexSelective Abortions in India."
Population and Development Review 28(4): 759785.
ASER. 2010. Annual Status of Education Report (Rural) 2009. Annual Status of
Education Report. New Delhi.
Asfaw, A., F. Lamanna, and S. Klasen. 2010. "Gender Gap in Parents' Financing Strategy
for Hospitalization of Their Children: Evidence from India." Health Economics
19(3): 265279.
Azam, Mehtabul and Geeta Gandhi Kingdon. 2011. Are Girls the Fairer Sex in India?
31

Revisiting Intrahousehold Allocation of Education Expenditure. CSAE

Working Paper Series 201110. Oxford: University of Oxford.
Baccara, Mariagiovanna., Allan CollardWexler, Leonard Felli, and Leeat Yariv. 2010.
ChildAdoption Matching: Preferences for Gender and Race. NBER Working
Paper No. 16444.
Barcellos, Silvia Helena, Leandro Carvalho, and Adriana LlerasMuney. 2012. "Child
Gender and Parental Investments in India: Are Boys and Girls Treated
Differently?" NBER Working Paper No. 17781.
Becker, Gary S. 1981. A Treatise on the Family. Cambridge, MA: Harvard Univ. Press.
Bedard, Kelly and Olivier Deschnes. 2005. "Sex Preferences, Marital Dissolution, and
the Economic Status of Women." The Journal of Human Resources 40(2): 411
434.
Bhalotra, Sonia and Tom Cochrane. 2010. Where Have All the Young Girls Gone?
Identification of Sex Selection in India. The Center for Market and Public
Organization 10/254. Bristol, UK: University of Bristol.
Bharadwaj, P. and L. Nelson. 2012. "Discrimination Begins in the Womb: Evidence of
Sex Selective Prenatal Investments." Forthcoming, Journal of Human Resources.
Bianchi, Suzanne and John Robinson. 1997. What Did You Do Today? Children's Use of
Time, Family Composition, and the Acquisition of Social Capital. Journal of
Marriage and Family 59: 332334.
Black, Sandra E., Paul J. Devereux, Kjell G. Salvanes. 2004. "The More the Merrier? The
Effect of Family Size and Birth Order on Children's Education." Quarterly
Journal of Economics 120(2): 669700.
Black, Sandra E., Paul J. Devereux, and Kjell G. Salvanes. 2010. "Small Family, Smart
Family? Family Size and the IQ Scores of Young Men." Journal of Human
Resources 45(1): 3358.
Burde, Dana and Leigh L. Linden. 2012. The Effect of VillageBased Schools:
Evidence from a Randomized Controlled Trial in Afghanistan. NBER Working
Paper No. 18039.
Butcher, Kristin F. and Anne Case. 1994. "The Effect of Sibling Sex Composition on
Women's Education and Earnings." The Quarterly Journal of Economics 109(3):

32

531563.
Case, Anne and Christina Paxson 2001. Mothers and Others: Who Invests in Children's
Health? Journal of Health Economics 20(3): 301328.
Charnov, E., R.L. LosDen Hartogh, W.T. Jones, and J. Van Den Assem. 1981. "Sex
Ratio Evolution in a Variable Environment." Nature 289: 2733.
Choi, HungJai, Jutta M. Joesch, and Shelly Lundberg. 2008. "Sons, Daughters, Wives,
and the Labour Market Outcomes of West German Men." Labour Economics
15(5): 795811.
Conley, Dalton. 2000. "Sibship Sex Composition: Effects on Educational Attainment."
Social Science Research 29: 441457.
Conley, Dalton and Rebecca Glauber 2006. "Parental Educational Investment and
Children's Academic Risk: Estimates of the Impact of Sibship Size and Birth
Order from Exogenous Variation in Fertility." Journal of Human Resources 41(4)
722737.
Dahl, Gordon B. and Enrico Moretti. 2008. "The Demand for Sons." Review of Economic
Studies (75): 10851120.
Das Gupta, Monica and Li Shuzhuo. 1999. Gender Bias in China, South Korea and India
19201990: Effects of War and Famine. Development and Change 30(3): 619
652.
Deaton, Angus. 1989. Looking for BoyGirl Discrimination in Household
Expenditures. World Bank Economic Review 3(1): 115.
Devine, D. and R. Forehand. 1996. Cascading Towards Divorce: The Roles of Marital
and Child Factors. Journal of Consulting and Clinical Psychology 64(2):424
427.
Diekmann, Andreas and Kurt Schmidheiny 2004. "Do Parents of Girls Have a Higher
Risk of Divorce? An Eighteen Country Study." Journal of Marriage and Family
66(3): 651660.
Duflo, Esther. 2003. Grandmothers and Granddaughters: Old Age Pension and Intra
household Allocation in South Africa. World Bank Economic Review 17(1): 1
25.
Ebenstein, Avraham. 2010. "The Missing Girls of China and the Unintended

33

Consequences of the One Child Policy." Journal of Human Resources 45(1): 87

115.
Ebenstein, Avraham and Ethan Jennings Sharygin. 2009. "The Consequences of the
Missing Girls of China." World Bank Economic Review 23(3): 399425.
Ebenstein, Avraham, Hongbin Li, and Lingsheng Meng. 2012. The Impact of Ultrasound
Technology on the Status of Women in China. Working Paper.
Filmer, Deon. 2007 If You Build It, Will They Come? School Availability and School
Enrollment in 21 Poor Countries. Journal of Development Studies 43(5):901
928.
Filmer, Deon and Lant Pritchett. 1999. "Educational Enrollment and Attainment in India:
Household Wealth, Gender, Village, and State Effects." Journal of Educational
Planning and Administration 13(2): 135164.
Gager, Constance.T., Laura A. Sanchez, and Alfred Demaris. 200. Whose Time Is It?:
The Effect of Gender, Employment, and Work/Family Stress on Childrens
Housework.Journal of Family Issues 30(11): 1459-1485. .
Handa, Sudhanshu. 2002. Raising Primary School Enrolment in Developing Countries:
The Relative Importance of Supply and Demand. Journal of Development
Economics 69:103-28.
Hanushek, Eric. 1992. The TradeOff Between Child Quantity and Quality. Journal of
Political Economy 100(1): 84117.
Haveman, Robert and Barbara Wolfe. 1995. The Determinants of Childrens
Attainments: A Review of Methods and Findings. Journal of Economic
Literature 33: 1829 1878.
Ichino, Andrea, Elly-Ann Lindstrm, and Eliana Viviano. 2011. Hidden Consequences
of a FirstBorn Boy for Mothers. IZA Discussion Papers No. 5649.
Jayachandran, Seema and Adriana LlerasMuney. 2009. Life Expectancy and Human
Capital Investments: Evidence from Maternal Mortality Declines. Quarterly
Journal of Economics 124(1): 349397.
Jayachandran, Seema and Ilyana Kuziemko. 2011. "Why Do Mothers Breastfeed Girls
Less than Boys: Evidence and Implications for Child Health in India." Quarterly
Journal of Economics 126(3): 14851538.

34

Jensen, Robert. 2012. "Do Labor Market Opportunities Affect Young Women's Work and
Family Decisions? Experimental Evidence from India." Quarterly Journal of
Economics 127(2): 753792.
Katzev, Aphra, RebeccaL. Warner, and Alan C. Acock. 1994. Girls or Boys?
Relationship of Child Gender to Marital Instability. Journal of Marriage and the
Family 56: 89100.
Khanna, R., A. Kumar, J.F. Vaghela, V. Sreenivas, andJ.M.Puliyel. 2003. Community
based retrospective study of sex in infant mortality in India. BMJ 327:126.
Lhila, Aparna and Kosali I. Simon. 2008. "Prenatal Health Investment Decisions: Does
the Child's Sex Matter?" Demography 45(4): 885905.
Lundberg, Shelly. 2005. "Sons, Daughters, and Parental Behavior." Oxford Review of
Economic Policy 21(3): 340356.
Lundberg, Shelly, Sara McLanahan, and Elaina Rose. 2007. "Child Gender and Father
Involvement in Fragile Families." Demography 44(1): 7992.
Lundberg, Shelly and Elaina Rose. 2002. "The Effects of Sons and Daughters on Men's
Labor Supply and Wages." The Review of Economics and Statistics 84(2): 251
268.
Lundberg, Shelly and Elaina Rose. 2003. "Child Gender and the Transition to Marriage."
Demography 40(2): 333349.
Lundberg, Shelly, Sabrina Wulff Pabilonia, Jennifer WardBatts, et al. 2007. "Time
Allocation of Parents and Investments in Sons and Daughters." Forthcoming.
Mammen, Kristin. 2008. "The Effect of Children's Gender on Living Arrangements and
Child Support." American Economic Review: Papers and Proceedings 98(2):
408412.
Mauldon, J., 1990, "The Effect of Mar ital Disruption on Children's Health,"
Demography , 27(3).
McLanahan, Sara and Gary .D. Sandefur. 1994. Growing Up With a Single Parent: What
Hurts and What Helps. Cambridge, Mass.: Harvard University Press.
Miguel, Edward. 2005. "Poverty and Witch Killing." Review of Economic Studies 72(4):
11531172.

35

Mogstad, Magne and Matthew Wiswall 2010. Testing the QuantityQuality Model of
Fertility: Linearity, Marginal Effects, and Total Effects. Working Paper.
Morgan, S. Philip, Diane .N. Lye, and Gretchen .A. Condran. 1988. Sons, Daughters,
and the Risk of Marital Disruption. American Journal of Sociology 94(1): 110
129.
Morgan, S. Philip., and Michael .S. Pollard. 2002. Do Parents of Girls Really Have a
Higher Risk of Divorce? Paper presented at the 2002 Annual Meeting of the
Population Association of America, Atlanta.
Munshi, Kaivan, and Mark Rosenzweig. 2006. "Traditional Institutions Meet the Modern
World: Caste, Gender, and Schooling Choice in a Globalizing
Economy."American Economic Review, 96(4): 12251252. Norberg, Karen. 2004.
"Partnership Status and the Human Sex Ratio at Birth." Proc. R. Soc. B.
271(1555): 24032410.
Pabilonia, Sabrina Wulff and Jennifer WardBatts. 2007. "The Effect of Child Gender on
Parents' Labor Supply: An Examination of Natives, Immigrants, and their
Children." The American Economic Review 97(2): 402406.
PetterssonLidbom, Per and Peter Skogman Thoursie. 2009. Does Child Spacing Affect
Childrens Outcomes? Uppsala: Institute for Labor Market Policy Evaluation.
Working Paper No. 2009:7.
Price, Joseph. 2008. "Parent Child Quality Time: Does Birth Order Matter?" Journal of
Human Resources 43(1): 240265.
OECD. 2008 OECD Employment Outlook, 2008 Edition. OECD. Multilingual
Summaries.
Pitt, Mark M., Mark R. Rosenzweig, and Nazmul Hassan. 1990. Productivity, Health,
and Inequality in the Intrahousehold Distribution of Food in LowIncome
Countries. The American Economic Review 80(5): 11391156.
Qian, Nancy. 2008. Missing Women and the Price of Tea in China: The Effect of Sex
Specific Income on Sex Imbalance. The Quarterly Journal of Economics 123(3):
12511285.
Rose, Elaina. 1999. "Consumption Smoothing and Excess Female Mortality in Rural
India." The Review of Economics and Statistics 81(1): 4149.

36

Rosenblum, Daniel. 2012. Economic Incentives for Sex Selective Abortion in India.
Working Paper.
Sen, A. 1990. "Missing Women." British Medical Journal 304: 587588.
Spanier, Graham B. and Paul C. Glick. 1981. Marital Instability in the United States:
Some Correlates and Recent Changes. Family Relations 31:32938.
Thomas, Duncan. 1994. Like Father, Like Son; Like Mother, Like Daughter: Parental
Resources and Child Height. Journal of Human Resources 29(4): 950988.
UNESCO. 2012. Enrollment and Gender Trends: Primary Education. World Atlas of
Gender Equality in Education. UNESCO.
World Bank. 2001. Engendering Development through Gender Equality in Rights,
Resources, and Voice. New York: Oxford University Press.
Wu, Zheng and Margaret J. Penning. 1997. Marital Instability After Midlife. Journal of
Family Issues 18(5): 459478.
Yeung, W. Jean, John F. Sandberg, Pamela E. DavisKean, and Sandra L. Hofferth. 2001.
"Children's Time with Fathers in Intact Families." Journal of Marriage and
Family 63: 136154.

37

15

24
23

38

40

27

25

1947

1990

24

34

38

Boy

Girl

1996

25

31

41

Either/doesn't matter
41

29

25

1997
Year

2000

25

27

42
38

28
27

2003

No opinion

Figure 2. Trend in Preferences for Sons versus Daughters in the U.S., Gallup Poll Data, 1941-2011.

45

40

35

30

25

20

15

10

0
1941

40

28

37

28

26

2011

26

2007

Questions: 2000, 2003, 2007, 2011: Suppose you could only have one child. Would you prefer that it be a boy or a girl?
1996, 1997: If you were to have a child right now, would you rather have a boy or a girl?
1947, 1990: If you had another child would you rather have a boy or a girl?
1941: If you could have only one (one more) child, which would you prefer to have--a boy, or a girl?

Figure 3. Gross Enrollment Rates in Primary School by Gender in Developing Countries.

120
100
80
Female

60

Male
40
20
0
1970s

1980s

1990s

2000s

Calculations based on 92 countries from Africa, Asia, and Latin America. Averages are taken over decade per
each country, and then averaged over the countries. Source: UNESCO Institute for Statistics.