women rule: preferences and fertility in australian households · women rule: preferences and...
TRANSCRIPT
Department of Economics
Issn 1441-5429
Discussion paper 57/10
Women Rule: Preferences and Fertility in Australian Households*
Elliott Fan† and Pushkar Maitra‡
Abstract Using a unique panel data set from Australia, we investigate how individual fertility
preferences translate into fertility realizations. We examine whether the dominance
over fertility decisions is held by the wife, the husband, or whoever with higher (or
lower) fertility desire. We find consistent evidence that the wife's preference is more
important than the husband's preference in predicting subsequent births, no matter
whether her initial fertility desire is higher or lower than that of her partner. We
also explore the implications of the introduction of so called Baby Bonus on fertility
outcomes. The results suggest that the effect of Baby Bonus is heterogeneous - it has
no effect on birth decisions for couples sharing a similar level of fertility desire, but it
appears to increase the influence of whoever has a higher level of fertility desire for
couples whose desires differ significantly.
Key Words: Fertility Preference, Births, Australia
JEL Codes: J12, J13, O12
* We would like to thank seminar participants at Monash University for comments and suggestions. The usual caveat
applies. † Elliott Fang, SPEAR Centre, Research School of Economics, ANU College of Business and Economics, The Australian
National University, Canberra ACT 0200, Australia. Email: [email protected] ‡ Pushkar Maitra, Department of Economics, Monash University, Wellington Road, Clayton Campus, VIC 3800, Australia.
Email: [email protected]
© 2010 Elliott Fang and Pushkar Maitra
All rights reserved. No part of this paper may be reproduced in any form, or stored in a retrieval system, without the prior
written permission of the author.
1 Introduction
Analysis of household behavior has traditionally been based on the idea that family mem-
bers maximize a single utility function. This is known as the unitary household or common
preference model. Underlying such models, the assumption of common preference order-
ing among family members can be traced back to Becker (1960). While this approach
has been proven useful for its elegance and analytical tractability, the hypothesis of a
single utility function encompassing all family members has been increasingly challenged
in recent years. Such challenges have included attempts at modelling individual utility
to incorporate divergent and conflicting preference of different family members (see, for
example, Manser and Brown, 1980; McElroy and Horney, 1981; Chiappori, 1988, 1992).
Crucial to the notion of non-unitary models of the household is the notion of power (Pol-
lak, 1994). Much of the empirical work using bargaining models has tested the resource
pooling implication of the unitary model. Failure to accept the hypothesis of resource
pooling generally leads to the conclusion that there exists some sort of bargaining process
within the household.
This literature in recent years has been extended to examine the fertility effects of spousal
differences. In the demography literature it has long been argued that males and females
differ in their desires regarding fertility and family planning (see Mason and Taj, 1987;
Pritchett, 1994). Empirically, it has been observed that men’s and women’s preferences
both affect fertility and family planning (see Freedman and Thornton, 1980; Thomson
et al., 1990; Bankole, 1995; Thomson, 1997; Dodoo, 1998). It is not clear, however, how
large the magnitude of the effect of differing preferences is. Further, whether women and
men have differing preferences over the number of children in general and if so how large
the difference is remain as yet unanswered questions.
In purely economic terms, the demand for children originates from the utility that parents
derive from children. Different preferences typically arise from differences in costs that
accrue to men and women. Women have to carry the burden of child bearing and in
most cases also the burden of child rearing. However knowledge of different preferences is
not sufficient, because ultimately having one more child is a joint decision irrespective of
2
whether or not preferences differ. How is this decision achieved and how relevant are the
preferences of women in this respect? The answer to this question is important because it
has significant implications for the design of population policies. If women have stronger
preferences compared to men then increasing the influence of women in the context of
decision making will help increase the number of births. This explanation is particularly
relevant in the context of most developed countries where the important issue facing policy
makers is to raise the birth rate. The problem is surely very different for policy makers
in developing countries, but the relative preferences of the husband and the wife remain
important in trying to understand the effects of fertility related policies.
To fully analyze the relationship between preferences and fertility one needs longitudinal
data where one can follow the couple over time to examine whether their preferences
are realized or not. There exist very few longitudinal data sets around the world that
collect data on both fertility preferences and realized birth outcomes. One of the few
that do is the Household Income and Labour Dynamics of Australia (HILDA) survey that
has been conducted annually from 2001. The HILDA data contain a unique measure of
fertility desire for both males and females aged 18− 55. This information, together with
information about birth outcomes, is collected annually in the survey, making it possible
to track the changes in both variables over time.
The advantage of using the HILDA data set is that it allows us to examine the following
issues:
1. the extent to which husband’s or wife’s preferences affect actual fertility outcomes;
2. whether husband’s or wife’s preference is more important in determining birth out-
comes;1 and
3. whether the answer to question 2 depends on the sign of the within-couple difference
in fertility preference.
Examining the last issue is of particular interest as it involves testing on another possi-
1Despite the fact that around 25% of couples in our estimating sample are in a de facto relationship,we use the generic terms – husbands and wives – to denote the male and the female partners respectively.
3
ble aspect of intra-household decision making: the dominance on fertility decision is not
unconditionally acquired by the wife or the husband, but by whoever with the higher or
lower fertility desire.2
Further, the time span of the data from HILDA allows us to explore a related and impor-
tant policy issue: to what extent will a public income transfer directed towards rewarding
births affect fertility rates? Economic theory (from Becker, 1960 onwards) supports this
possibility, postulating a direct relationship between the costs associated with having
children, and empirical evidence often suggests that public income incentives enhance fer-
tility (see also, Rosenzweig, 1999; Milligan, 2005; McDonald, 2006; Cohen et al., 2007;
Laroque and Salanie, 2008). Further, it is often argued that cash transfers targeted to-
wards women is associated with a change in household consumption patterns (Duflo, 2003;
Ashraf et al., 2009) and in fertility outcomes (Eswaran, 2002; Seebens, 2005), due to an
increase of women’s bargaining power over household decisions. In this paper, we explore
these possibilities using the introduction and reform of the Australian Maternity Payment
(popularly known as the Baby Bonus) scheme as a natural experiment. The scheme, as
a head-counted, non-means-tested program, commenced in 2004 with scheduled increases
in later years, offers a fixed stipend to the primary carer of each new born child. To gauge
the policy effects, our strategy focuses on estimating the impact of the program on fertility
realizations, and examining whether the impact is associated with within-couple disparity
in fertility desire.3
We find consistent evidence suggesting that wife’s fertility desire is more important in
forecasting birth outcomes. After controlling for husband’s initial preference, the proba-
bility of having an additional child in the next 4 or 6 years is significantly higher if the
2The literature on the effect of preferences on fertility is not surprisingly quite scarce. Partly this isdriven by the lack of data availability. The paper that comes closest to this study is a working paperby Hener (2010). Using the child preference data from the German Socio-Economic Panel, he developsa strategy that makes use of within-couple preference disparity to test the unitary model against thebargaining model. He finds that, for a couple where the woman has a greater preference to having childrenthan her partner, fertility increases with the income share of the woman. Our paper does not explicitlyexamine the issue of bargaining; rather we examine the effects of preferences directly.
3The true effects of the Baby Bonus program are still open to debate. Gans and Leigh (2009) argued thatthe main impact of the program was the delay in births by a day/few days before the policy commenced,so couples could take advantage of the program. Drago et al. (2009) find that fertility intentions rose afterthe announcement of the program and the birth rate is estimated to have risen moderately as a result ofthe program.
4
wife’s initial desire is higher, and significantly lower if wife’s initial desire is lower. Despite
the fact that the two effects are in opposite directions, they have a similar magnitude. As
to the effect of Baby Bonus, we find that it casts no impact on birth decisions of couples
having a similar level of fertility desire. For couples whose desires differ significantly, how-
ever, the provision of the Baby Bonus increases the influence of whoever has a higher level
of fertility desire. This result is consistent with the hypothesis that the public income can
be used by the partner with higher fertility desire to bargain with the other partner over
fertility decisions.
The rest of the paper is organized as follows. In the next section (Section 2), we intro-
duce the data set and provide detailed definitions of the key variables, followed by an
introduction of the policy background of Baby Bonus. In Section 3, we present the em-
pirical strategies and the hypotheses to be tested in this paper. The empirical results are
presented in Section 4. Finally, Section 5 concludes.
2 Data and Policy Background
The data used in this paper are from years 2001 to 2007 of the Household Income and
Labour Dynamics of Australia (HILDA) survey that has been conducted annually from
2001. It is a nationally representative household longitudinal survey, of which the opening
sample consists of 19,914 individuals in Australian households in 2001. The survey in-
cluded modules on family and household formation, sources of income, and labour supply
behaviours. The collection of each wave of the data took place between August of the
given year and March of the following year.
One unique aspect of the HILDA data set that we utilize in this paper is that it contains
measures of fertility desire and expectations for both males and females aged 18−55. This
information is collected by asking the respondents the following two questions.
1. Pick a number between 0 and 10 to show how you feel about having [a child/more
children] in the future. This is fertility desire.
5
2. Pick a number between 0 and 10 to show how likely you will have [a child/more
children] in the future. This is fertility expectation.
While in this paper we restrict ourselves to fertility desire, it needs to be mentioned that
Fan and Ryan (2009) have shown that the two measures share a similar predictive power
of subsequent birth realizations. A larger number of this measure represents a higher level
of fertility desire.
In our estimating sample we include couples who are married or are in a de facto relation-
ship at the time of the survey. When a couple separates or get divorced, they are excluded
from the sample. We also restrict our sample to couples where the wife is aged 18 − 40
in any survey round, so couples are trimmed out of the sample when the wife reaches age
41. At the same time, there are other reasons for sample attrition, and a small number of
couples enter the sample due to family formation. Table 1 presents the selected descriptive
statistics for the estimating sample (of 1, 636 couples). Among them around 75% of the
couples are legally married, and the average number of children per couple is 1.64. In
general the husbands are slightly older than the wives (the average age of the husbands in
our sample is 35 years, compared to 32 for the sample of wives); 90% of the wives are in
good health (are free from long term health condition, disability or impairment) compared
to 87% of the husbands. The yearly income of the husbands is three times that of the
wives. The average fertility desire of the wives is slightly higher (4.42) than the figure of
the husbands (4.33).4
Figure 1 illustrates the average fertility desire for females aged 18 to 40 and their partners
using the 2001 data.5 The two ‘age profiles’ are fairly close to each other, with the only
difference being that women’s profile is slightly steeper. This implies that their fertility
desire decays faster than their partners if we ignore the possible cohort effects. Figure 2
shows the relationship between 2001 desire and the incidence of having one child/more
children in the following six years. The graph presents a clear positive (unconditional)
correlation between fertility desire and subsequent births for both males and females,
4We provide the definitions of the variables F more and F less in Section 3.1.5Both curves are generated by using the STATA command lpoly, which performs a kernel-weighted local
polynomial regression of the y variable on the x variable and displays a graph of the smoothed values.
6
though there is a gap between the two curves when the value of desire is between around 4
and 8. Finally, Figure 3 displays the density distribution of the within-couple disparity in
fertility desire, computed as the wife’s measure minus the husband’s measure. It is clear
that the density is heavily concentrated on the middle points, and is roughly symmetrical
with respect to point zero.
On May 11, 2004, the Australian government announced the Maternity Payment scheme,
commonly referred to as Baby Bonus.6 The primary aim of this program was to stop the
slide in fertility rate in Australia and it was hoped that this kind of a cash payment would
serve as an incentive to have children. Mr. Peter Costello, the then Federal Treasurer
was unequivocal about this intention: he strongly urged Australian couples to have “one
[baby] for your husband, one for your wife and one for the country.”7
Baby Bonus was implemented in July of 2004, less than two months after it was announced,
so that there was very little time gap between the announcement and the implementation
of the program. As a start, the program provided a fixed amount of AU$3, 000 to the
primary carer of a newborn child born on or after July 1, 2004. The most important
aspect of this program was that it was not means tested. The payment is granted to the
primary carer of each new-born child, regardless of the child’s birth order, the income of
the household, or any other observable characteristics of the parents and/or the household.
The benefit level was later raised by 26% on July 1, 2006.8
6One needs to bear in mind that effectively the Baby Bonus scheme replaced two existing payments −the Maternity Allowance (MA) and a “baby bonus” administered through the Australian Tax Office. TheMA was a relatively modest payment (a maximum of AU$842.64 per child at the time the scheme cameto an end) which was restricted to women who were eligible for Family Tax Benefit, and thus by extensionlived in households with at most modest incomes. On the other hand, the bonus was administered throughthe tax system. While being potentially much more generous (with a maximum sum of up to AU$12,500per child available over a 5-year period following a birth) than the MA, the bonus seems to have not beenwidely utilized. Low utilization rates were probably due to the program functioning as a complicated anddelayed tax rebate system. Indeed, the most substantial payments were reserved for women with relativelyhigh employment income in the year prior to birth, who subsequently remained out of the workforce for atotal of five years.
7As in many other developed countries, after the post World War II boom, the total fertility rate declinedin Australia, from a peak of 3.5 children per woman in 1961 to 1.75 in 2003 (http://www.abs.gov.au). Theseconditions triggered significant public debates in Australia about the causes of this decline and appropriatepolicy responses to reverse this trend (Gray et al., 2008).
8Along with the first announcement in May 2004, it was announced that the amount of the baby bonuswould go up in the future, but the jump of $834 that happened in 2006 was unexpected in terms of theamount of the increase.
7
3 Strategies
The starting point of our analysis is to examine the role of wife’s fertility desire in pre-
dicting birth realizations, and compare it to the role of the husband’s desire. A straight
forward way to do so is to estimate the following family model:
bit = α+ βfFDfi,2001 + βmFDm
i,2001 + Xfitγf + Xm
it γm + Zitδ + ηY eart + εit (1)
The dependent variable in the regression is whether the couple (indexed i) has a child
in year t (t = 2002, ...., 2007): bit = 1 if the couple had a child in year t, and bit = 0
otherwise; FDfi,2001 and FDm
i,2001 refer to fertility desire measured in 2001 for the wife
and the husband respectively; Xfit and Xm
it are row vectors of individual characteristics
for the wife and the husband, including age, age squared, education level, health, private
income, and number of existing children; Zit contains state dummies and a dummy variable
indicating whether the couple are married or in a de facto relationship; Y eart refers to the
time trend; and εit is the error term. Here, βf captures the marginal effect of the wife’s
fertility desire in 2001 on the incidence of having a birth in any year during 2002 to 2007,
controlling for her husband’s fertility desire in 2001. Symmetrically, βm represents the
marginal effect of the husband’s initial fertility desire. Through a direct comparison of βf
and βm, we will be able to test whether the predictive power of the wife’s initial desire is
different from that of her husband’s.
One problem with using the preference variables to determine births is that preferences
might themselves be affected by birth outcomes. The desire to have another child is
therefore endogenous, especially in the cases that panel data are employed to conduct
the estimation. This endogeneity makes the interpretation of the results difficult. This
is why in Equation (1) we restrict preferences to those in 2001, which we assumed to
be pre-determined and is not affected by birth outcomes in the subsequent years. This
specification, however, is subject to a limitation that both partners’ fertility desires would
change after a child is born. Thus, the predictive power of 2001 desires is likely to be
limited to the first subsequent birth. After having one birth, the couple’s fertility desires
could reform and this breaks the linkage between the initial desires and later births. Thus,
to a large extent the model only allows us to estimate the average effect of the initial desires
8
on the first birth that happens in the subsequent years.9
3.1 Effect of Differences in Preferences
The model specified in Equation (1) does not allow estimating the effect of within-couple
disparity in fertility desire, nor does it allow an investigation of whether the sign of the
disparity matters. To address these questions, we estimate a model that incorporates
the within-couple disparity in both positive and negative directions, denoted by Diff ,
which is defined as the value of the wife’s desire minus the husband’s desire. We then
define three dummy variables as: (1) female wanting more (F more) = 1 if Diff > 2; (2)
female wanting less (F less) = 1 if Diff < −2; and (3) no conflict if Diff ∈ [−2, 2].10
Throughout the regressions analyses below, the no conflict group always serves as the
reference category. We then estimate the following regression:
bit = α+ βmFDmi,2001 + θ1F morei,2001 + θ2F lessi,2001 (2)
+ Xfitγf + Xm
it γm + Zitδ + ηY eart + εit
Once again the within couple difference in desire is defined as of 2001, which is assumed
to be pre-determined and unaffected by subsequent outcomes.
The difference between equations (1) and (2) lies in the inclusion of F morei,2001 and
F lessi,2001, both measured at 2001, and the exclusion of FDmi,2001 in Equation (2). These
changes make Equation (2) more flexible than Equation (1) by allowing the effect of the
wife’s desire to be different when her desire is at least two points higher, two points lower,
or within a disparity of two points when compared to her husband’s desire.11
9We also estimate a modified version of Equation (1) where the dependent variable is replaced by adummy variable indicating whether a couple have any births during the period of 2002 to 2007, whileFDf
i,2001 and FDmi,2001 continue to remain as the key regressors.
10Figure 4 presents the average values of F more and F less over the different survey years. It is worthnoting that the proportion of couples with F more = 1 and F less = 1 remain relatively stable over thedifferent survey years.
11Our empirical results are highly robust to an alternative categorization of the female wanting moreand the female wanting less groups based on a lower degree of disparity in fertility desire, where the threegroups are categorized by (1) female wanting more (F more) = 1 if Diff > 1; (2) female wanting less(F less) = 1 if Diff < −1; and (3) no conflict if Diff ∈ [−1, 1].
9
Results from Equation (2) are particularly helpful for testing the following alternative
models of decision making within the household:
Model 1: If the wife’s preference affects birth decisions despite whether the wife’s desire is
higher or lower than that of her husband, θ1 is expected to be positive. A positive θ1
implies that, compared to the no conflict group, the female wanting more (F more)
group has a higher probability of having more children during 2002 to 2007 (Note
that the husband’s initial desire, FDmi,2001, is controlled for in Equation (2)). At the
same time, θ2 should be negative because a negative θ2 suggests that the fertility
incidence is lower for the female wanting less (F less) group. When we find θ1 > 0
and θ2 < 0, a test on θ1 + θ2 = 0 tells us whether the two effects are symmetric.
Model 2: In the case that the husband’s desire dominates and wife’s preference does not
matter, both θ1 and θ2 should be zero because any deviation of the wife’s desire from
that of her husband does not have any effect on birth outcomes (that is, θ1 = θ2 = 0).
Model 3: It is possible that the dominance over birth decisions is not acquired by either
the wife or the husband, but by whoever has the higher fertility desire. In this case,
θ1 ought to be positive because higher wife’s desire drives up fertility. Meanwhile,
θ2 is expected to be zero because, after the husband’s initial desire is controlled for,
lower wife’s desire does not lead to a lower fertility incidence. So we have θ1 > 0
and θ2 = 0.
Model 4: It is also possible that the birth decisions are dominated by whoever has the
lower fertility desire. If so, θ1 and θ2 are expected to be zero and negative, respec-
tively, i.e., θ1 = 0 and θ2 < 0.
It needs to be noted that equation (2) shares the same limitation of equation (1): it does
not allow reforming of fertility desires for both partners after they give a birth. Thus,
since we use 2001 within-couple difference in desire to predict birth realizations in the
subsequent years, its predictive power is plausibly restricted to the first subsequent birth.
As a result, θ1 and θ2 only capture the mean difference in the incidence of having the first
10
birth in the subsequent years between the F more group and the no conflict group, and
between the F less group and the no conflict group, respectively.
3.2 Exploring the Effect of the Baby Bonus
It is often argued that cash transfers targeted towards women (of the kind we have as
a result of the baby bonus program) is often associated with a change in the spending
pattern of households and in number of births. The argument here is that transfers to
women increases their bargaining power, which in turn differently affects the behavior
of men and women. A secondary aim of this paper is to explore household bargaining
over birth decisions using the Australian Baby Bonus program as an experiment. In this
case, though in practice either mother or father can claim the payment, the mother is
usually assumed to be the principal applicant of the benefit. This might to some degree
provide privilege to mothers.12 In this scenario, the provision of the Baby Bonus payment
might enhance the bargaining power of the wife. However, another possible effect is that
it enhances the bargaining power of the partner, either the wife or the husband, whose
fertility desire is higher than that of the other partner. In this Section, we examine these
possible effects using the following regression function, which is modified from the family
model given by Equation (2) in Section 3.1:
bit = α+ βmFDmi,2001 + θ1F morei,2001 + θ2F lessi,2001 (3)
+ φY 04 + θ′1 (F morei,2001 × Y 04) + θ′2 (F lessi,2001 × Y 04)
+ Xfitγf + Xm
it γm + Zitδ + ηY eart + εit
Equation (3) is different from equation (2) in the inclusion of two additional terms: the
interaction terms of F more dummy and F less dummy with a dummy variable, denoted by
Y 04, indicating post-program years, that is, post 2004 (Y 04 = 1), or otherwise (Y 04 = 0).
12Typically it is the primary carer who receives the payment. In the vast majority of cases, it isthe mother of the child who is the eligible person. Using the infant cohort of the LSAC data setfor Australia, Harrison et al. (2010) find that out of a sample of 5,107 infants aged between 3 and19 months, 98% had the mother as the primary carer. This proportion goes down to 97.3% forchildren aged between 51 and 67 months (sample size is 4,983). If it is the mother who is mak-ing the claim, she does not need to provide her spouse’s details. More information is available at:http://www.ato.gov.au/taxprofessionals/content.asp?doc=/content/38285.htm
11
In this specification, the coefficients θ1 and θ2 represent the effects of F more dummy and
F less dummy on the annual probability of having an additional child during 2002 and
2003 (pre-program years), while the coefficients θ′1 and θ′2 captures how the two effects
change from pre-program years to post-program years, respectively. If the provision of
Baby Bonus did not alter the role of the within-couple desire differentials in forecasting
birth outcomes, both θ′1 and θ′2 are expected to be zero.
The major limitation of Equation (3) being used to identify the policy effect lies in lacking
a proper comparison group, which is required to satisfy two assumptions – the control
group is free from the effect of the policy, and no other factors affect the birth decisions
of the treatment and control groups in the same way as Baby Bonus does. Since Baby
Bonus is a nation-wide program, it is difficult to find a legitimate control group that
well satisfies the two validity requirements. Instead of trying to locate such a control
group, we argue that θ′1 and θ′2 legitimately represent the effects of Baby Bonus; because
after examining other child-related public programs in Australia such as the Family Tax
Benefit, the Maternity Immunization Allowance, and the Child Care Benefit, we do not
find any significant changes in these programs that could plausibly induce changes in
fertility parallel to the possible impact of the Maternity Payment around 2004.
4 Results
We present the empirical results from estimating Equations (1), (2), and (3) in Sections
4.1−4.3. Throughout all estimations, we focus on a sample comprised of married and de
facto couples, of which the wife is aged 18− 40 in any of the survey years. We also show
results based on sub-samples that consist of females of various age groups or samples of
different years.
4.1 Effect of Fertility Desire on Birth Outcomes
Table 2 shows the results from Probit estimation of equation (1). All reported coefficient
estimates are marginal effects of the corresponding regressors. In column (1), estimates
12
are obtained using couples with the wife aged 26 to 40, which are considered as the
group facing the largest likelihood to give births. The estimated effects suggest that,
after controlling for husband’s desire in 2001, an unit increment of wife’s desire in 2001 is
associated with an increase of 1.1 percentage points in the incidence that the couple will
have one more child in any of the years between 2002 and 2007. Assuming that this effect
can be linearly extrapolated along the values of FDfi,2001, this estimate implies that the
annual probability that a wife with maximum fertility desire (FDfi,2001 = 10) in 2001 will
have a child during 2002 and 2007 is 11 percentage points higher, relative to a wife with
minimal fertility desire (FDfi,2001 = 0) in 2001. Compared to the mean (12.2%) of the
dependent variable, this estimate implies a fairly large marginal effect of the wife’s initial
fertility desire. The corresponding estimate for husbands, however, is only 0.4 percentage
point, suggesting a weaker predictive power of the husband’s initial desire, though the
estimate is still statistically significant. Not surprisingly the null hypothesis of equality of
the effect of the wife’s and the husband’s desire is rejected (p− value = 0.0317).
Columns (2) and (3) suggests that the estimates of FDfi,2001 and FDm
i,2001 are fairly robust
to an extended sample that includes wives aged 21 to 25 (column (2)), and to an even
larger sample that additionally incorporates those aged 18 to 20 (column (3)). All these
results show an estimate of FDfi,2001 that is more than twice as large in magnitude as the
estimate of FDmi,2001. Meanwhile, the null hypothesis of equality of the effect of the wife’s
and the husband’s desire is rejected in all cases.
In the results presented in columns (1) - (3), the dependent variable is whether the couple
i had a child in year t. In column (4), we present the estimates from the Probit regression
with the dependent variable replaced by a dummy variable indicating whether the couple
had any (at least one) child during the entire 2002 − 2007 period. Thus, the regression
reduces to a cross-sectional format using only the 2001 data. As shown in column (4), the
results are qualitatively similar to those displayed in columns (1) − (3). The estimated
effects suggests that, after controlling for husband’s desire in 2001, an unit increment of
wife’s desire in 2001 is associated with an increase of 3.6 percentage points in the incidence
that the couple will have at least one child during the period of 2002 − 2007. The effect
of an unit increase in the husband’s desire is lower (2.3 percentage points) though the
13
null hypothesis of the equality of the wife’s and the husband’s desire cannot be rejected
(p− value = 0.1364).13
4.2 Effect of Differences in Preference on Birth Outcomes
Table 3 presents the results estimated using Equation (2). All coefficients are the marginal
effects from a standard Probit regression of bit. Note that the covariates Xfit, X
mit , and Zit
remain the same as those in Equation (1), and the sample refers to couples with the female
partner aged 18 to 40 in any survey round during 2002 and 2007. Again, our analysis here
focuses on testing the following four hypotheses:
1. Wife’s preference affects birth decisions in both directions (θ1 > 0 and θ2 < 0);
2. Husband’s preference dominates over birth decisions (θ1 = θ2 = 0);
3. The dominance belongs to whoever has the greater desire for children (θ1 > 0 and
θ2 = 0);
4. The dominance belongs to whoever has the lower desire for children (θ1 = 0 and
θ2 < 0).
In column (1), the estimate of θ1 is 0.053, which is positive and statistically significant.
This implies that the annual probability of having more child(ren) during 2002 and 2007
is 5.3 percentage points higher for the female wanting more group, compared to the no
conflict group. Also, the estimate of θ2 is −0.045, also statistically significant, suggesting
that the annual birth incidence for the female wanting less group is 4.5 percentage points
lower relative to the no conflict group. The null hypothesis of θ1 = −θ2 cannot be
rejected (p− value = 0.7257), suggesting that the effects are roughly symmetric (though
in opposite directions) in that the magnitudes of these two effects are not significantly
different. Thus, these results are consistent with the hypothesis that the wife’s fertility
preference matters and it matters in both directions: the wife’s preference drives up fertility
13Note that the number of observations in column (4) is only about one third of that in column (3).This should lead to a much larger standard error and thus the marginally insignificant t value.
14
when her preference is higher, and drags down fertility when her preference is lower.
Finally, the magnitudes of the two estimates are relatively large provided that the mean
of the dependent variable is 0.122.
As argued earlier, one concern about our methodology is that the predictive power of
fertility preference on birth realizations might fade out over time, i.e., the initial fertility
desire is not so relevant in forecasting the birth decisions in later years of the sample as
in earlier years. To explore this issue, we shorten the sample by eliminating the data of
the last two years (2006 and 2007) and replicate the estimation of Equation (2) using
this restricted sample. The results are presented in column (2) of Table 3. Compared
to the full sample results in column (1), the estimate of θ2 is virtually unchanged, while
the estimate of θ1 becomes larger (0.09, compared to 0.053 in column (1)). The outcomes
suggest that, while the female wanting more effect appears to fall over time, the female
wanting less effect remains highly persistent.14 These findings are in fact sensible as the
female wanting more group have a higher birth incidence in the subsequent years, and
the initial fertility desire (as well as the within-couple disparity) would lose its predictive
power of birth decisions after the first birth realization, which might explain the fading
effect of θ1. On the other hand, the female wanting less effect is persistent because this
group faces a much lower fertility rate over the same period of time.
We further explore the robustness of the results by examining another sub-sample that
consists of only 2004 to 2007 data. To make the results comparable to those based on
the earlier sub-sample (2002 to 2005 data), we control for male’s desire in 2003 (that is,
FDmi,2003) as the initial value instead of the 2001 desire used in the full sample and the
earlier sub-sample. Similarly F more and F less are accordingly redefined in terms of the
2003 preferences. The estimated model can be specified as:
bit = α+ βmFDmi,2003 + θ1F morei,2003 + θ2F lessi,2003 (4)
+ Xfitγf + Xm
it γm + Zitδ + ηY eart + εit
We present the results in column (3) of Table 3, which tell a very similar story – significant
θ1 and θ2 but in opposite directions, and the difference in magnitude of the two effects is
14Despite the increased gap between θ1 and θ2, a formal test still cannot reject the hypothesis of equalityof the two estimates (p− value = 0.3650).
15
insignificant.
Finally in column (4) we present the marginal effects from a probit estimation with the
dependent variable replaced by a dummy variable indicating whether the couple had any
child during the 2002 − 2007 period. Birth incidence for the female wanting more group
is significantly higher compared to the no conflict group, while it is significantly lower for
the female wanting less group. The effects are approximately symmetric and a formal test
cannot reject the null hypothesis of equality of the two magnitudes (p− value = 0.9429).
Combining the results from this section and previous section, we conclude that wife’s
preference dominates over birth decisions. After controlling for the husband’s initial desire
to have more children, the probability of having an additional child in the following 4 or
6 years is significantly higher if the wife’s initial desire is greater than that of her partner
(relative to the no conflict group where couples share similar degrees of desire) and is
significantly lower if the wife’s initial desire is weaker (again, relative to the reference
group).
4.3 Policy Effect: Did the Baby Bonus Scheme Make a Difference?
Table 4 presents the marginal effects estimated from Equation (3) based on the Probit
model. Recall that both the announcement and the implementation of the baby bonus
program took place between the collections of 2003 and 2004 HILDA, so the Y 04 dummy
neatly separates the pre- and post-program time periods. In the specification of Equation
(3), θ1 captures the effect for female wanting more relative to no conflict in pre-2004
period, while θ1 + θ′1 gives the effect for female wanting more relative to no conflict in
post-2004 period. Meanwhile, θ2 and θ2+θ′2 are defined accordingly for the female wanting
less variable.
As pointed out in the previous section, it is possible that the policy effects captured by
θ′1 and θ′2 are confounded with the fading effect of the initial preferences. For example,
couples in the female wanting more group (as of 2001) might act fast to have a child, say
in 2002 or 2003. Since their fertility plan is realized, the 2001 desires are unable to forecast
16
birth decisions in the later years. To reduce the concern about this problem, we estimate
Equation (3) using only the two sub-samples (the 2002− 2005 and the 2004− 2007 data),
which are the same two sub-samples as used in the previous section. To apply the second
sub-sample, again, we set fertility desire in 2003 as the initial value, based on which we
calibrate the within-couple difference in fertility desire and construct the female wanting
more and female wanting less dummies as of 2003. Also, a dummy variable, Y 06, is
defined to indicate years 2006 and 2007 (Y 06 = 1) as the post-program years. Note that
the two sub-samples share a very similar structure, as there are two pre-program years of
data followed by two post-program years of data in each of the sub-sample. The first sub-
sample excludes 2006 and 2007 data, reducing the concern about the fading-out effects.
Using the second sub-sample, we are able to exploit the possible lagged effect, if lagged by
two years, of the introduction of Baby Bonus in 2004. Also, if there is only an immediate
effect that takes place instantly after the policy commenced, the second sub-sample can
be used to examine the 2006 policy expansion that raised the benefit level of Baby Bonus
by AU$834, which constitutes a second (though possibly weaker) experiment to explore
the policy effect. Based on the second sub-sample, we estimate the following regression:
bit = α+ βmFDmi,2003 + θ1F morei,2003 + θ2F lessi,2003 (5)
+ φY 06 + θ′1 (F morei,2003 × Y 06) + θ′2 (F lessi,2003 × Y 06)
+ Xfitγf + Xm
it γm + Zitδ + ηY eart + εit
The results using these two sub-samples are presented in columns (1) and (2) of Table 4,
respectively. In column (1), the estimates of θ1 and θ2 are both statistically significant, and
they suggest that female wanting more is associated with a 8.5 percentage point increase
in the probability of having another child per year in 2002 and 2003, while female wanting
less is associated with a 5.4 percentage point reduction in the probability during the same
period of time. The estimate of the Y 04 dummy is small and not statistically significant,
indicating a limited program effect on birth incidence for couples in the no conflict group.
The estimate of θ′1 is positive, but not statistically significant, indicating that the positive
effect of female wanting more is actually strengthened post 2004. At the same time, very
17
interestingly, the estimate of θ′2 is also positive, though not statistically significant, and
the null hypothesis of θ2 + θ′2 = 0 is rejected (p − value = 0.140). The results therefore
imply that while wife’s preference drives fertility decisions pre-program, post-program it
is the preference for the partner with greater initial fertility desire that is driving fertility
decisions: θ1 + θ′1 > 0 implies that higher wife’s desire drives up fertility; on the other
hand, θ2 + θ′2 = 0 suggests that once the husband’s initial desire is controlled for, lower
wife’s desire does not lead to a lower fertility incidence. The post-program outcome is
consistent with Model 3 described in Section 3.1, which states that the dominance over
fertility decisions belongs to the partner with higher desire when the coefficient of female
wanting more variable is positive and the coefficient of female wanting less variable is zero.
The results displayed in column (2) present a similar picture based on the second sub-
sample, though some important differences are worth noting. First, the estimate of
θ1(0.041) is much smaller than its counterpart in column (1) and is not statistically signif-
icant, implying that the effect of female wanting more (as of 2003) on the birth incidence
of the following two years (that is, 2004 and 2005) is smaller (less than half in terms of
magnitude) than the effect of female wanting more (as of 2001) on the birth incidence of
2002 and 2003. Second, the estimate of θ2(−0.077) is larger in magnitude than the cor-
responding value in column (2), suggesting a stronger deterring effect of female wanting
less as of 2003 than that as of 2001 on birth incidence in the next two years. Third, and
most importantly, both the estimates of θ′1(0.31) and θ′2(0.135) are positive and relatively
large in magnitude (note in particular that the estimate of θ′2 is now significant at the 10%
level). Thus, the total effect of female wanting more is significantly positive in post-2006
period, but the total effect of female wanting less is not significant in the same period.
Finally, similar to the estimate of Y 04 based on the first sub-sample, the estimate of the
Y 06 dummy is small and insignificant, implying that the effect on birth incidence for
couples in the no conflict group is little.
Essentially, these results suggest that post-program the dominance over birth decisions
is not driven by either the wife or the husband, but by whoever has the higher fertility
desire. How should we interpret these results? One possible channel that Baby Bonus
affects a couple’s birth decisions is through reducing the cost of having children (that is,
18
making children ‘cheaper’). To the extent that a child is a normal good, both the income
effect and substitution effect caused by this ‘price’ change should increase the probability
of having more children, and these effects should be applicable to all couples in the female
wanting more, female wanting more, and no conflict groups. Another possible channel
that is only relevent of the first two groups, not the no conflict group, operates through
household bargaining – Baby Bonus enhances the bargaining power of the partner with
higher fertility preference by making children less costly, because children are ‘goods’ that
this partner prefers. Our results using the two sub-samples show that the Y 04 dummy in
column (1) and the Y 06 in column (2) are not statistically significant, implying that the
baby bonus program did not have a significant effect on the no conflict group. Therefore,
it is difficult to argue that the first channel is at work in a significant way. At the same
time, especially based on the 2004-2007 sub-sample, the findings of positive θ′1 and θ′2
support the second possibility that Baby Bonus enhances the bargaining power of the
partner with higher fertility desire.
One might wonder that if it is true that the Baby Bonus has changed fertility outcomes
by altering what drives fertility outcomes within the household, why is the effect so much
stronger in 2006 relative to 2004, when the program was initially introduced and the
amount of incentive was much greater? One possible explanation is that the implemen-
tation of Baby Bonus could have had a delayed effect on fertility outcomes. After all, an
additional birth is not something that happens over night, and it is reasonable to consider
that it takes around 1.5 to 2 years to complete a birth plan.15 Therefore, in terms of
fertility outcomes, it might well be the case that the Baby Bonus is effective only with
a lag and the lagged effect, if of around 2 years, is better picked up by the 2004 − 2007
sample than by the 2002− 2005 sample.
Finally, recall that while the program can have an effect on fertility directly, it can in
addition have an effect on fertility indirectly by changing the effect of preferences. So
far we have not examined how the program and preferences interact so that the program
15There is possibly a considerable amount of time spent on planning and even after conception theoutcome is realized after 9 months, not to mention that an average couple is likely to conceive within12−15 months of starting to try. Additionally it might take some time for individuals to be fully convincedabout the nature of the program.
19
has the indirect effect on fertility behavior. To examine this indirect effect, consider the
following regression where we regress the difference in preferences (difference in the desire
for having another child/children) for couple i in year t on a set of initial preferences.
Specifically we consider the following regression:
Diffit = α+ βmFDmi,2001 + θ1F morei,2001 + θ2F lessi,2001 + φY 04 (6)
+ Xfitγf + Xm
it γm + Zitδ + ηY eart + εit
The results are presented in Table 5, where we show that Baby Bonus does not have
a signficant effect on preferences, using either the full sample or the two sub-samples
(2002 − 2005 and 2004 − 2007).16 The Y 04 dummy in columns (1) and (2) and the Y 06
dummy in column (3) are never statistically significant, indicating that there is no effect
for the no conflict group.
5 Conclusion
To briefly summarize our results, we find consistent evidence that the fertility preference
of wives is more important in predicting birth outcomes, compared to the preference of
husbands . After controlling for the initial preferences for men, the probability of having
an additional child in the next 4−6 years is significantly lower if the wife’s initial desire is
higher, and significantly lower if the wife’s initial desire is lower. As to the effect of Baby
Bonus, we find that it is unlikely that the payment increases the decision power of the
wife of a family, but increases the influence of whoever has a higher fertility desire. This
result is consistent with the hypothesis that the availability of Baby Bonus benefit can be
used by the partner with higher level of fertility desire to convince the other partner to
have more children.
Our findings are intriguing from a policy point of view. First, it is likely that the effect
of Baby Bonus is heterogeneous, as the effect depends on the fertility preferences of the
16As before for the second sub-sample the initial conditions are defined as of 2003, not 2001. In thiscase, instead of using Equation (6), we estimate the following regression:
Diffit = α+ βmFDmi,2003 + θ1F morei,2003 + θ2F lessi,2003 + φY 06Xf
itγf +Xmit γm + Zitδ + ηY eart + εit
20
husband and the wife. For couples who share a similar degree of fertility desire, and
possibly have formed their birth plans, the policy effect appears to be little. This result is
consistent with a common speculation that the benefit level (AU$3,000 to 4,000) of Baby
Bonus is not significant enough to induce a large number of births purely due to income
effect. However, we should not ignore the possibility that the income incentive operates
more effectively through household bargaining, given that there are many couples with
different levels of fertility preference and our findings seem to suggest that the availability
of Baby Bonus payment strengthens the influence of the partner with higher fertility
desire at least in the short term. Second, from a broader perspective, the existence of
household bargaining should be an important consideration for policy makers. After all,
most household decisions are jointly made by the couple, and in general both partners’
preferences matter if they are not completely consistent with each other. Thus, knowing
about the within-couple disparity in preferences and how the two partners translate their
preferences into decisions would help advance our understanding and prediction of the
effects of public policies.
References
Ashraf, N., E. Field, and J. Lee (2009). Household bargaining and excess fertility: An
experimental study in zambia. Mimeo, Harvard University.
Bankole, A. (1995). Desired fertility and fertility behaviour among the yoruba of nigeria:
a study of couple preferences and subsequent fertility. Population Studies 49 (2), 317 –
328.
Becker, G. (1960). An economic analysis of fertility. In Demographic and Economic Change
in Developed Countries, pp. 209 – 231. Princeton, NJ, Princeton University Press.
Chiappori, P. (1988). Rational household labor supply. Econometrica 56 (1), 63 – 89.
Chiappori, P. (1992). Collective labor supply and welfare. Journal of Political Econ-
omy 100 (3), 437 – 467.
Cohen, A., R. Dehejia, and D. Romanov (2007). Do financial incentives affect fertility?
Mimeo, NBER Working Paper # 13700.
Dodoo, F. N. A. (1998). Men matter: additive and interactive gendered preferences and
reproductive behavior in kenya. Demography 35 (2), 229 – 242.
21
Drago, R., K. Sawyer, K. Sheffler, D. Warren, and M. Wooden (2009). Did australia’s
baby bonus increase the fertility rate. Mimeo, MIAESR Working Paper # 1/09.
Duflo, E. (2003). Grandmothers and granddaughters: Old age pensions and intra-
household allocation in south africa. World Bank Economic Review 17 (1), 1 – 25.
Eswaran, M. (2002). The empowerment of women, fertility, and child mortality: Towards
a theoretical analysis. Journal of Population Economics 15 (3), 433 – 454.
Fan, E. and C. Ryan (2009). Child subsidy, within-couple inconsistency and fertility plans.
Mimeo, SPEAR Centre, Australian National University.
Freedman, R., D. S. F. and A. D. Thornton (1980). Changes in fertility expectations and
preferences between 1962 and 1977: Their relation to final parity. Demography 17 (4),
365 – 378.
Gans, J. and A. Leigh (2009). Born on the first of july: An (un)natural experiment in
birth timing. Journal of Public Economics 93 (1-2), 246 – 263.
Gray, M., L. Qu, and R. Weston (2008). Fertility and family policy in australia. Working
paper, Australian Institute of Family Studies, Melbourne, Research Paper # 41.
Harrison, L. J., J. A. Ungerer, G. J. Smith, S. R. Zubrick, and S. Wise (2010). Child care
and early education in australia - the longitudinal study of australian children. Working
paper, Social Policy Research Paper # 40, FaHCSIA.
Hener, T. (2010). Do couples bargain over fertility? evidence based on child preference
data. Mimeo, Ifo Institute for Economic Research, University of Munich.
Laroque, G. and B. Salanie (2008). Does fertility respond to financial incentives? Mimeo,
Discussion Paper # 0708-05, Columbia University.
Manser, M. and M. Brown (1980). Marriage and household decision making: a bargaining
analysis. International Economic Review 21 (1), 31 – 44.
Mason, K. O. and A. M. Taj (1987). Differences between women’s and men’s reproductive
goals in developing countries. Population and Development Review 13 (4), 611 – 638.
McDonald, P. (2006). Low fertility and the state: The efficacy of policy. Population and
Development Review 32 (3), 485 – 510.
McElroy, M. and M. J. Horney (1981). Nash-bargained household decisions: Towards a
generalization of the theory of demand. International Economic Review 21 (2), 333 –
349.
Milligan, K. (2005). Subsidizing the stork: New evidence on tax incentives and fertility.
Review of Economics and Statistics 87 (3), 539 – 555.
22
Pollak, R. A. (1994). For better or worse: the roles of power in models of distribution
within marriage. American Economic Review: Papers and Proceedings 84 (2), 148 –
152.
Pritchett, L. H. (1994). Desired fertility and the impact of population policies. Population
and Development Review 20 (1), 1 – 55.
Rosenzweig, M. R. (1999). Welfare, marital prospects and nonmarital childbearing. Jour-
nal of Political Economy 107 (6), S3 – S32.
Seebens, H. (2005). Bargaining over fertility in rural ethiopia. Mimeo, University of
Goettingen and Center for Development Research, Bonn.
Thomson, E. (1997). Couple childbearing desires, intentions, and births. Demogra-
phy 34 (3), 343 – 354.
Thomson, E., E. McDonald, and L. L. Bumpass (1990). Fertility desires and fertility:
hers, his, and theirs. Demography 27 (4), 579 – 588.
23
Figure 1: Age Profiles of Fertility Desire for Couples, 2001
Source: Household Income and Labour Dynamics of Australia, 2001
24
Figure 2: Fertility Desire and Subsequent Birth Realisation for Couples
Note: The fertility desire is measured at 2001Source: Household Income and Labour Dynamics of Australia, 2001
25
Figure 3: Within-couple Difference in Fertility Desire, 2001
Note: The difference is measured as wife’s desire minus husband’s desire.
Source: Household Income and Labour Dynamics of Australia, 2001
26
Figure 4: Proportions of F more and F less groups by Year
0.000
0.020
0.040
0.060
0.080
0.100
0.120
0.140
2001 2002 2003 2004 2006 2007
%
Year
F_moreF_less
Note: The 2005 values are dropped as questionnaires asking for fertility desire are different fromthose of other years making the definitions of F more and F less for 2005 incomparable
Source: Household Income and Labour Dynamics of Australia, 2001-2004 and 2006-2007.
27
Table 1: Summary Statistics for Couples Using 2001 Data
Female MalesMean s.d. Mean s.d.
Age 31.99 5.54 34.87 6.86Education (%)Grad diploma, grad certificate 0.06 0.23 0.04 0.20Bachelor or honours 0.16 0.37 0.15 0.36Adv diploma, diploma 0.11 0.31 0.08 0.27Cert III or IV 0.12 0.32 0.32 0.47Cert I or II 0.01 0.11 0.01 0.10Cert not defined 0.00 0.05 0.00 0.00Year 12 0.20 0.40 0.12 0.32Year 11 and below 0.32 0.47 0.24 0.43Health∗ 0.90 0.31 0.87 0.34Fertility desire 4.42 4.34 4.33 4.23Yearly income (×10−6) 0.02 0.02 0.05 0.05F more 0.12 0.33F less 0.12 0.32In de facto relationship 0.25 0.43Total number of children 1.64 1.38Observations 1,636
Notes:∗ Health refers to a dummy variable indicating whether the
respondent is free from long term health condition,disability or impairment (equal to 1) or otherwise (equal to 0)
All incomes are deflated to 2001 value.
28
Tab
le2:
Par
sim
onio
us
regr
essi
onre
sult
s
(1)
(2)
(3)
(4)
Wifeaged
26
to40
Wifeaged
21
to40
Wifeaged
18
to40
Wheth
erhad
any
child,2002
to2007a
Mea
nof
the
dep
end
ent
vari
ab
le0.1
22
0.1
22
0.1
22
0.3
26
Wif
e’s
fert
ilit
yd
esir
ein
2001
0.0
11***
0.0
11***
0.0
11***
0.0
36***
(0.0
02)
(0.0
02)
(0.0
02)
(0.0
05)
Hu
sban
d’s
fert
ilit
yd
esir
ein
2001
0.0
04**
0.0
04**
0.0
05***
0.0
23***
(0.0
02)
(0.0
02)
(0.0
02)
(0.0
05)
Wif
e’s
age
0.0
61***
0.0
72***
0.0
64***
0.1
61***
(0.0
19)
(0.0
13)
(0.0
13)
(0.0
27)
Wif
e’s
age
squ
are
d-0
.001***
-0.0
01***
-0.0
01***
-0.0
03***
(0.0
00)
(0.0
00)
(0.0
00)
(0.0
00)
Wif
e’s
hea
lthb
0.0
32***
0.0
31***
0.0
32***
-0.0
39
(0.0
12)
(0.0
12)
(0.0
12)
(0.0
51)
Hu
sban
d’s
hea
lthb
0.0
20
0.0
18
0.0
16
0.0
58
(0.0
13)
(0.0
12)
(0.0
12)
(0.0
40)
Ind
efa
cto
rela
tion
ship
-0.0
28***
-0.0
32***
-0.0
32***
-0.0
83***
(0.0
11)
(0.0
10)
(0.0
10)
(0.0
30)
No.
of
exis
tin
gch
ild
ren
=1
0.0
10
0.0
19
0.0
19
0.0
52
(0.0
13)
(0.0
13)
(0.0
13)
(0.0
43)
No.
of
exis
tin
gch
ild
ren
=2
or
more
-0.0
87***
-0.0
77***
-0.0
77***
-0.0
40
(0.0
15)
(0.0
14)
(0.0
14)
(0.0
41)
Wif
e’s
yea
rly
inco
me
(×10−6)
-0.5
65***
-0.6
18***
-0.6
32***
0.3
34
(0.2
03)
(0.2
04)
(0.2
06)
(0.5
97)
Hu
sban
d’s
yea
rly
inco
me
(×10−6)
-0.0
09
-0.0
26
-0.0
28
0.3
77
(0.1
01)
(0.1
00)
(0.1
00)
(0.3
74)
Tim
etr
end
-0.0
02
-0.0
02
-0.0
02
–(0
.003)
(0.0
03)
(0.0
03)
–T
est
on
equ
ality
of
the
coeffi
cents
of
0.0
317
0.0
435
0.0
468
0.1
364
male
an
dfe
male
’sd
esir
es(p
-valu
e)O
bse
rvati
on
s5,1
07
5,4
41
5,4
88
1,6
36
Note
s:C
oeffi
cien
tsare
imp
ute
dm
arg
inal
effec
ts.
Rob
ust
stan
dard
erro
rsin
pare
nth
eses
.∗∗∗p<
0.0
1,∗∗p<
0.0
5,∗p<
0.1
a:
Wif
eaged
18
to40
b:
Hea
lth
refe
rsto
ad
um
my
vari
ab
lein
dic
ati
ng
wh
eth
erth
ere
spon
den
tis
free
from
lon
gte
rmh
ealt
hco
nd
itio
n,
dis
ab
ilit
yor
imp
air
men
t(e
qu
al
to1)
or
oth
erw
ise
(equ
al
to0)
All
inco
mes
are
defl
ate
dto
2001
valu
e.O
ther
contr
ol
vari
alb
esin
clud
est
ate
du
mm
ies
an
dd
um
mie
sfo
rb
oth
hu
sban
d’s
an
dw
ife’
sed
uca
tion
level
s.
29
Tab
le3:
Exam
inin
gth
eeff
ect
ofd
isp
arit
yof
fert
ilit
yd
esir
eon
fert
ilit
yre
aliz
atio
ns
(1)
(2)
(3)
(4)
Years2002
to2007
Years2002
to2005
Years2004
to2007
Wheth
erhad
any
childa
Mea
nof
the
dep
end
ent
vari
ab
le0.1
22
0.1
26
0.1
34
0.3
26
Hu
sban
d’s
fert
ilit
yd
esir
ein
2001
0.0
15***
0.0
17***
0.0
55***
(0.0
01)
(0.0
02)
(0.0
04)
Hu
sban
d’s
fert
ilit
yd
esir
ein
2003
––
0.0
20***
––
–(0
.002)
–F
more
0.0
53***
0.0
92***
0.0
58**
0.1
87***
(0.0
19)
(0.0
24)
(0.0
26)
(0.0
47)
Fless
-0.0
45***
-0.0
45***
-0.0
47***
-0.1
46***
(0.0
10)
(0.0
12)
(0.0
14)
(0.0
29)
Wif
e’s
age
0.0
63***
0.0
63***
0.0
48***
0.1
63***
(0.0
13)
(0.0
15)
(0.0
13)
(0.0
28)
Wif
esage
squ
are
d-0
.001***
-0.0
01***
-0.0
01***
-0.0
03***
(0.0
00)
(0.0
00)
(0.0
00)
(0.0
00)
Wif
e’s
hea
lthb
0.0
35***
0.0
40***
0.0
34**
-0.0
33
(0.0
11)
(0.0
13)
(0.0
14)
(0.0
50)
Hu
sban
d’s
hea
lthb
0.0
16
0.0
17
0.0
17
0.0
52
(0.0
12)
(0.0
15)
(0.0
15)
(0.0
40)
Ind
efa
cto
rela
tion
ship
-0.0
31***
-0.0
38***
-0.0
25**
-0.0
83***
(0.0
10)
(0.0
12)
(0.0
12)
(0.0
30)
No.
of
exis
tin
gch
ild
ren
=1
0.0
23*
0.0
15
0.0
11
0.0
48
(0.0
14)
(0.0
15)
(0.0
15)
(0.0
43)
No.
of
exis
tin
gch
ild
ren
=2
or
more
-0.0
80***
-0.0
71***
-0.0
62***
-0.0
56
(0.0
14)
(0.0
16)
(0.0
16)
(0.0
41)
Wif
e’s
yea
rly
inco
me
(×10−6)
-0.6
63***
-0.7
28***
-0.5
39**
0.4
43
(0.2
09)
(0.2
70)
(0.2
34)
(0.6
07)
Hu
sban
d’s
yea
rly
inco
me
(×10−6)
-0.0
36
-0.0
33
-0.0
55
0.4
07
(0.1
05)
(0.1
31)
(0.1
11)
(0.3
79)
Tim
etr
end
-0.0
01
0.0
06
0.0
00
–(0
.003)
(0.0
05)
(0.0
05)
–T
est
on
equ
ivale
nce
of
coeffi
cents
ofF
more
an
dF
less
(p-v
alu
e)0.7
257
0.3
650
0.8
110
0.9
429
Ob
serv
ati
on
s5,4
88
4,0
55
3,6
86
1,6
36
Note
s:C
oeffi
cien
tsare
imp
ute
dm
arg
inal
effec
ts.
Rob
ust
stan
dard
erro
rsin
pare
nth
eses
.∗∗∗p<
0.0
1,∗∗p<
0.0
5,∗p<
0.1
a:
Yea
rs2002
to2007
b:
Hea
lth
refe
rsto
ad
um
my
vari
ab
lein
dic
ati
ng
wh
eth
erth
ere
spon
den
tis
free
from
lon
gte
rmh
ealt
hco
nd
itio
n,
dis
ab
ilit
yor
imp
air
men
t(e
qu
al
to1)
or
oth
erw
ise
(equ
al
to0)
All
inco
mes
are
defl
ate
dto
2001
valu
e.O
ther
contr
ol
vari
alb
esin
clu
de
state
du
mm
ies
an
dd
um
mie
sfo
rb
oth
hu
sban
d’s
an
dw
ife’
sed
uca
tion
level
s.
30
Table 4: Estimating the impact of Baby Bonus on household bargaining
(1) (2))Years 2002 to 2005 Years 2004 to 2007
Mean of the dependent variable 0.126 0.134Husband’s desire in 2001 0.017***
(0.002)Husband’s desire in 2003 0.020***
(0.002)Initial F more 0.085*** 0.041
(0.028) (0.031)Initial F less -0.054*** -0.077***
(0.014) (0.014)Y04 -0.015
(0.022)Y06 -0.013
(0.024)(Initial F more)× Y04 0.011
(0.033)(Initial F less)× Y04 0.034
(0.040)(Initial F more)× Y06 0.031
(0.043)(Initial F less)× Y06 0.135*
(0.074)Wife’s age 0.063*** 0.048***
(0.015) (0.013)Wife’s age squared -0.001*** -0.001***
(0.000) (0.000)Wife’s health 0.040*** 0.033**
(0.013) (0.014)Husband’s health 0.018 0.016
(0.015) (0.015)In de facto relationship -0.038*** -0.025**
(0.012) (0.012)No. of existing children = 1 0.016 0.012
(0.015) (0.015)No. of existing children = 2 or more -0.070*** -0.060***
(0.016) (0.016)Wife’s yearly income(×10−6) -0.727*** -0.527**
(0.270) (0.232)Husband’s yearly income (×10−6) -0.032 -0.048
(0.132) (0.107)Time trend 0.010 0.001
(0.010) (0.010)Test on θ1 + θ′1 = 0 (p-value) 0.002 –Test on θ2 + θ′2 = 0 (p-value) 0.140 –Test on θ1 + θ′1 = 0 (p-value) – 0.017Test on θ2 + θ′2 = 0 (p-value) – 0.601Observations 4,055 3,686
Notes:Coefficients are imputed marginal effects. Robust standard errors in parentheses.∗∗∗p < 0.01,∗∗ p < 0.05,∗ p < 0.1∗ Health refers to a dummy variable indicating whether the respondent is free from long term health condition,
disability or impairment (equal to 1) or otherwise (equal to 0)All incomes are deflated to 2001 value.Other control varialbes include state dummies and dummies for both husband’s and wife’s education levels.
31
Table 5: Estimating the impact of Baby Bonus on within-couple difference in fertilitydesire
(1) (2) (3))Years 2002 to 2007 Years 2002 to 2005 Years 2004 to 2007
Y04 0.025 0.023(0.027) (0.035)
Y06 -0.009(0.045)
Husband’s desire in 2001 0.002 0.004(0.002) (0.003)
Husband’s desire in 2003 -0.005(0.003)
Initial F more 0.217*** 0.244*** 0.272***(0.026) (0.031) (0.033)
Initial F less -0.306*** -0.357*** -0.341***(0.026) (0.031) (0.034)
Wife’s age -0.052*** -0.047** -0.032*(0.018) (0.020) (0.019)
Wife’s age squared 0.001*** 0.001** 0.000(0.000) (0.000) (0.000)
Wife’s health -0.007 -0.022 0.026(0.025) (0.029) (0.032)
Husband’s health 0.011 0.006 0.026(0.023) (0.029) (0.027)
In de facto relationship -0.005 -0.001 -0.026(0.021) (0.025) (0.024)
No. of existing children = 1 0.047** 0.059** 0.040(0.024) (0.027) (0.027)
No. of existing children = 2 or more 0.059*** 0.073*** 0.045(0.023) (0.028) (0.028)
Wife’s yearly income(×10−6) 0.092 0.177 -0.021(0.243) (0.365) (0.266)
Husband’s yearly income(×10−6) -0.384*** -0.360** -0.351**(0.145) (0.159) (0.167)
Time trend -0.009 -0.011 -0.002(0.008) (0.018) (0.019)
Observations 5,488 4,055 3,686
Notes:Coefficients are imputed marginal effects. Robust standard errors in parentheses.∗∗∗p < 0.01,∗∗ p < 0.05,∗ p < 0.1∗ Health refers to a dummy variable indicating whether the respondent is free from long term health condition,
disability or impairment (equal to 1) or otherwise (equal to 0)All incomes are deflated to 2001 value.Other control varialbes include state dummies and dummies for both husband’s and wife’s education levels.
32