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: Elliott.fan@anu.edu.au ‡ Pushkar Maitra, Department of Economics, Monash University, Wellington Road, Clayton Campus, VIC 3800, Australia.

Email: Pushkar.Maitra@monash.edu

© 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