New Zealand Economic Papers

ISSN: 0077-9954 (Print) 1943-4863 (Online) Journal homepage: https://www.tandfonline.com/loi/rnzp20

The effect of the price or rental cost of housing on

family size: a theoretical analysis with reference to
New Zealand

Mimi Liu & Jeremy Clark

To cite this article: Mimi Liu & Jeremy Clark (2017) The effect of the price or rental cost of housing
on family size: a theoretical analysis with reference to New Zealand, New Zealand Economic
Papers, 51:3, 281-301, DOI: 10.1080/00779954.2016.1199591

To link to this article: https://doi.org/10.1080/00779954.2016.1199591

Published online: 28 Jun 2016.

Submit your article to this journal

Article views: 473

View related articles

View Crossmark data

Citing articles: 2 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=rnzp20
NEW ZEALAND ECONOMIC PAPERS, 2017
VOL. 51, NO. 3, 281
301
http://dx.doi.org/10.1080/00779954.2016.1199591

The effect of the price or rental cost of housing on family size:

a theoretical analysis with reference to New Zealand
Mimi Liu and Jeremy Clark
Department of Economics and Finance, University of Canterbury, Christchurch, New Zealand

ABSTRACT ARTICLE HISTORY

We investigate the effect of higher house prices or rental costs on family Received 23 June 2015
size. We provide a static model of a household’s choice of family size Accepted 7 June 2016
assuming constant elasticity of substitution preferences between children, KEYWORDS
leisure, and other goods, and Cobb
Douglas household production of Economic determinants of
children using parental time and housing. We then explore the wealth and fertility; housing prices;
substitution effects that changing house prices or rental costs have on home ownership; wealth
desired family size. Renters are predicted to have fewer children in effects
response to higher rents. Home owners are predicted to have more
children in response to higher house prices only if they have sufficient
initial housing and low substitution between family size and other
consumption, and fewer children otherwise. Finally, we provide
exploratory correlations between the change in lagged house prices or
rental costs and the change in number of children born to women using
aggregated census units in New Zealand. We find weak negative
correlations in both cases.

I. Introduction
Some developed countries are concerned they do not have enough population growth, with fertility
rates being below replacement level, while others have the reverse concern. Along with immigration
policy, policy-makers are interested in whether policy levers that change economic incentives are
effective in influencing fertility. This has been the focus of many studies in economics and demogra-
phy to date; for example, the responsiveness of fertility to parental leave or child tax credits (Feyrer,
Sacerdote, & Stern, 2008; Lalive & Zweim€ uller, 2009; Ridao-Cano & McNown, 2005; Whittington,
1992; Whittington, Alm, & Peters, 1990).
One area of price signal and economic incentive that has not received as much attention in the
literature is the extent to which fertility responds to changes in the price of housing. Since housing
is needed to have children, and is a major source of both cost and store of wealth for home-owning
families, then if fertility responds to economics incentives at all, it seems likely that it will respond to
changes in the price of housing. Finding out the extent to which this is true would be helpful for
anticipating whether fertility would be influenced by other policy levers that affect economic incen-
tives, and particularly policy levers that change the price of housing.
In theory, high rental costs increase the cost of living for renters, reducing the spending available
for raising children. The space needed to raise children also becomes more expensive, increasing the
cost of having children. Thus, an exogenous increase in the price of rental housing could be expected
to have a negative effect on fertility for renters. Conversely, as recognised by others (Dettling &

CONTACT Jeremy Clark jeremy.clark@canterbury.ac.nz

© 2016 New Zealand Association of Economists Incorporated
282 M. LIU AND J. CLARK

Kearney, 2014; Lovenheim & Mumford, 2013), housing also represents one of the single largest
stores of wealth for home owners. Provided children are normal goods, an increase in the purchase
price of housing thus creates a positive wealth effect for current home owners, potentially increasing
the number of children they wish to have. With offsetting wealth and substitution effects for home
owners, the net effect of an increase in the price of housing on fertility is uncertain, and should be
sensitive to the initial amount of housing already owned. Less recognised, home owners’ responses
to changing house prices will depend on their elasticity of substitution between family size and other
‘goods’.
The rest of this paper will proceed as follows. Section 2 provides a review of the existing empirical
literature on the response of fertility to economic incentives, with a focus on the price of housing.
Section 3 provides a theoretical model of a representative household’s choice of family size, focusing
on the price of rental and owner-occupied housing. While a proper empirical test of our theory is
beyond the scope of this paper, Section 4 provides simple correlations between changes in housing
price or rental costs and changes in fertility for filtered aggregated census units in New Zealand.
Section 5 provides a conclusion.

2. Literature review
Researchers in economics and demography have long been interested in the possible economic
influences on fertility. The effect on fertility of female labour force participation and employment,
parental leave, and tax exemptions, to name a few, have been studied in detail in the historical litera-
ture. The concept of house prices directly affecting fertility has been emerging in more recent stud-
ies, but few have emphasised the theoretical importance of home ownership status.
Historically, women have been responsible not only for giving birth to children, but taking a pri-
mary role in raising them as well. Hence, the opportunity cost of a woman’s time is a major determi-
nant of a family’s opportunity cost of having children. Since substantial time is required to look after
children, mothers face a significant trade-off between childcare services and market work. Thus,
female labour force participation and female employment affect a couple’s choice of family size, and
vice versa. Estimating female labour force participation and fertility simultaneously, early studies by
Cain and Dooley (1976) and DeFronzo (1980) find that female labour force participation has a sta-
tistically significant negative effect on fertility. Breaking down the analysis to first- and second-order
births, Kim (2014) finds that female labour force participation has no statistically significant effect
on the probability of having a first child, but has a negative effect on the probability of having a sec-
ond. A meta-analysis conducted by Matysiak and Vignoli (2008) on 30 articles investigating the
relationship between women’s employment and fertility using European and United States data find
that women’s employment has a statistically significant negative effect on fertility in countries with
conservative, liberal, or family-centred welfare regimes.
Child support policies, such as parental leave and tax exemptions, also affect a family’s opportu-
nity cost of having children. As more childcare support is given to mothers, it reduces the opportu-
nity cost of having children and should result in a positive effect on fertility. Indeed, Lalive and
Zweim€ uiller (2009), and Feyrer et al. (2008) find that increased parental leave provision is associated
with higher levels of fertility. Similarly, studies by Ridao-Cano and McNown (2005), Whittington
(1992), and Whittington et al. (1990) on the impact of the United States personal tax exemption of
dependents on fertility unanimously find a statistically significant positive relationship between the
two.
Of particular interest here are the empirical studies which examine the effect of housing prices on
fertility. As early as 1986, B€orsch-Supan investigates the effect of high housing prices on people’s
choice of housing alternatives. He raises the issue that high housing prices may lead to several cou-
ples living together, or married couples choosing to live with their parents, discouraging household
formation.1 Using data from the United States, B€orsch-Supan finds evidence to support his
 NEW ZEALAND ECONOMIC PAPERS 283

hypothesis. Clark (2012), Giannelli and Monfardini (2003), and Hughes (2003) have also docu-
mented the effect of high housing prices on delayed family formation.
Taking family formation as a given, Walker (1995) develops a model of the ‘shadow price’ of fer-
tility to help explain variations in Swedish fertility. He constructs a shadow price of fertility, com-
posed of direct expenditures (which includes the additional expenditure on housing needed to have
children), childcare, foregone earnings, and forgone human capital accumulation. Walker finds a
negative correlation between fertility and its shadow price, but the extent to which this is driven by
housing expenditures alone is not identified.
Following the same theoretical framework as Walker, Curtis and Waldfogel (2009) test for factors
which affect the cost of having children. Namely, they test for the impact of labour markets, housing
costs and availability, and welfare policies in the United States on the likelihood of unmarried and
married mothers having additional children. Using data from the Fragile Families and Wellbeing
Study and three variables to control for housing (housing costs, availability of subsidised housing,
and an interaction between the two), Curtis and Waldfogel find that only the interaction term is sta-
tistically significant, suggesting unmarried mothers are more likely to have additional children when
more subsidised housing is available and housing costs are high.
Other studies have also highlighted that fertility is affected by the housing market, but no
emphasis has been put on the price of housing specifically. Kostelecky and Vobecka (2009)
investigate the impact of measures of housing affordability on the total fertility rate in the Czech
Republic.2 Using two different indicators (the price-to-income ratio of the average house price
with the average income of an average household, and the percentage of households that qualify
for a mortgage of an apartment of average size and price), the authors find that the total fertility
rate is lower when housing is less affordable. Similarly, Vignoli, Rinesi, and Mussino (2013)
argue that housing tenure, housing prices, housing supply, and access to home loans are all
aspects of housing market conditions which simultaneously affect fertility. They examine how
the degree to which couples feel secure about their housing situation influences their short-term
fertility plans in Italy. Using data from the household multipurpose survey Family and Social
Subjects, the authors find that women who feel more secure about their housing conditions are
more likely to plan to have their first child.
In recent years, a small number of studies have tested directly for the effect of housing price on
fertility. Pan and Xu (2012) categorise the effect of house prices on fertility via two channels: a hous-
ing price constraint and a housing area constraint. High house prices compete with spending avail-
able for children (the housing price constraint) and they reduce the size of housing occupied by
households (the housing area constraint), where space is needed to have children. Using the real
interest rate as an inverse proxy for housing prices in China, Pan and Xu find a statistically signifi-
cant positive correlation between the real interest rate and fertility, in support of their hypothesis of
a housing price constraint. In addition, the authors find that housing area per capita and the number
of rooms per capita are positively correlated with fertility, in support of their hypothesis of a housing
area constraint. Yi and Zhang (2010) investigate the impact of housing prices on fertility in Hong
Kong using annual aggregate-level census data and the house price index for the period 1971
2005.
They find that house prices are negatively related to the total fertility rate, controlling for female
labour force participation, and female and male wage rates. Simon and Tamura (2009) and Feyrer
et al. (2008) investigate the links between house price and fertility in the United States’ context.
Simon and Tamura use the Integrated Public Use Microdata Samples (IPUMS) from the census for
the period 1940
2000, with a dependent variable of either the number of children currently living
in the household, or the number of children ever born to women aged 40 years and above. The
median Consolidated Metropolitan Statistical Area (CMSA) rental rate per room is used as the
main measure of housing price. They find a statistically significant negative relationship between the
price of living space and the number of children living in households, and the total number of chil-
dren born to older women. In contrast, Feyrer et al. (2008) also use IPUMS data from the 1980 and
2000 United States census, and housing price data from the Office of Federal Housing Oversight’s
284 M. LIU AND J. CLARK

repeat sales index, but find a positive or no relationship between housing price and state or Metro-
politan Statistical Area (MSA)-level total fertility.
The studies mentioned above test directly for the effect of housing price on fertility, but have not
recognised the influence of home ownership. We expect that the effect of changes in the price of
housing on fertility and family size differs depending on whether households are home owners or
renters. Renters may experience only substitution effects from price changes, whereas home owners
may experience substitution and wealth effects in complex interactions. Two more recent US-based
studies on house prices and fertility by Lovenheim and Mumford (2013) and Dettling and Kearney
(2014) do make the distinction between home owners and renters, and therefore are the most rele-
vant to our approach.
Lovenheim and Mumford (2013) follow individuals over time using restricted geo-coded data
from the Panel Study on Income Dynamics (PSID) from 1985 to 2007. They estimate the likelihood
of having a child in response to house price levels, and two- or four-year changes in house price. For
home owners, house price is measured from self-reported house value, while for renters, average
MSA level house prices are used. Regression specifications control for the number of other children
already in the household, and include year- and state (or MSA)-fixed effects. For home owners, the
authors find a statistically significant positive relationship between house price changes and fertility,
suggesting wealth effects dominate substitution effects. Surprisingly, no statistically significant nega-
tive relationship between house price changes and fertility was found for renters.
The second study recognising home ownership by Dettling and Kearney (2014) uses data on
births from the American vital statistics natality files from 1990 to 2007. Dettling and Kearney esti-
mate the effect on MSA-level fertility of lagged MSA-level house prices (constructed using the Fed-
eral Housing Finance Agency House Price Index) and MSA-level home ownership rates. To
distinguish between the effect on home owners and renters, they include an interaction term
between lagged house prices and home ownership rates. They interpret the coefficient on the house
price variable alone as the conditional main effect of house prices on fertility, net of home ownership
(i.e. the substitution effect). The interaction term then picks up the effect of house prices on the fer-
tility of home owners, net of any effects experienced by renters (i.e. the wealth effect). Including
MSA- and year-fixed effects, Dettling and Kearney find that high housing prices have a negative
effect on the fertility of renters but a positive effect for home owners.3
Our investigation is motivated by the same intuition presented in the two key studies by Love-
nheim and Mumford (2013), and Dettling and Kearney (2014). However, neither of these studies
provides a theoretical model to examine the conditions under which house prices could be expected
to increase or decrease fertility for home owners. Recognising this, we begin our investigation by
summarising a theoretical model of a household’s choice of family size. We demonstrate that the
quantity of housing already owned by the household (not just binary home owner/renter status)
plays an important role, as does elasticity of substitution between children and other ‘goods’.

3. Theoretical model
Consider a representative household that chooses consumption over three goods: children ðcÞ,
leisureðlÞ, and all other goods ðx2 Þ.4 We assume the household’s preferences over these three goods
can be represented by a constant elasticity of substitution (CES) utility function, represented as5

r 1
U D ½acr C blr C ð1 ¡ a ¡ bÞx2 r ; (1)

where 0 < a; b < 1. The parameter r is the household’s (uniform) willingness to substitute between
any pair of the three goods, and is bounded between negative infinity and one.6 As r approaches
one, the household becomes perfectly willing to substitute between children, leisure, and all other
goods. As r approaches minus infinity, the household becomes unwilling to substitute between the
 NEW ZEALAND ECONOMIC PAPERS 285

goods, preferring them in fixed proportions. At the intermediate case where r D 0, preferences
become Cobb
Douglas with an elasticity of substitution of one. We will focus on two cases of
r: 0 < r < 1, where the household is relatively willing to substitute between the goods, and r < 0,
where it is not.7
Preferences aside, we assume the household can have children according to a Cobb
Douglas
production function using two inputs: total quantity of housing either owned ðx1O
T
Þ or rented ðx1R Þ,
and time spent on children (t). If we assume the household views owned or rented housing as perfect
substitutes of some ratio for the purpose of raising children, the household’s production function
can be represented as

1¡d
c D t d ðax1O
T
C bx1R Þ : (2)

The parameter d measures the relative importance of time and housing in the production of chil-
dren, and we assume constant returns to scale, or 0 < d < 1.
We allow that the effect of (owned or rented) housing price on the household’s desired number of
children could be affected by the quantity of housing it already owns, x1O E
0. Because housing of
either tenure type enters as perfect substitutes in (2), a household will be best off choosing the tenure
type that is cheapest, adjusted for any difference it faces in effectiveness of rented vs. owned housing
for raising children, a to b. In what follows, we use the simplest case where a D b D 1, while recognis-
ing in practice that it might vary across families. When renting is cheaper than owning, or p1R < p1O ,
the household will sell off any endowment of owned housing and choose an optimal amount of
housing to rent, x1R . When owning is cheaper than renting, or p1O < p1R , the household will adjust
any initial endowment of housing x1O E
up or down by amount ~x1O , to arrive at an optimal total
amount of owned housing:

T
x1O D x1O
E
C ~x1O : (3)

Finally, the household faces constraints of time and budget. Assuming the household’s endow-
ment of time (T) is exhausted, it will allocate this between time spent raising children, leisure, and
labour (L), or

t C l C L D T: (4)

For its budget, the household has two potential sources of income: earnings from labour hours at a
E
wage rate w, and if x1O > 0, income from any sale of endowed housing.8,9 Assuming the household’s
time constraint is binding, it will fully allocate its endowment (T) between time spent raising chil-
dren, leisure, and labour (L), or making the price of the composite good the numeraire, the budget
constraint is

If p1R < p1O ; p1R x1R C x2 D wL C p1O x1O

E
: (5a)
If p1O < p1R ; p1O ~x1O C x2 D wL: (5b)

Substituting L from (4) into (5a) and (5b) gives the single constraint:

If p1R < p1O ; p1R x1R C x2 D wðT ¡ t ¡ lÞ C p1O x1O

E
: (6a)
If p1O < p1R ; p1O ~x1O C x2 D wðT ¡ t ¡ lÞ: (6b)

Performing a monotonic transformation on the utility function (1) to make calculations simpler,
we now have all the information required to set up the household’s optimisation problem. In the
286 M. LIU AND J. CLARK

case where renting is cheaper than owning (p1R < p1O ):

If 0 < r < 1; Max aðt d x1R 1 ¡ d Þr C blr C ð1 ¡ a ¡ bÞx2 r

x1R ;t;l;x2 (7a)
subject to p1R x1R C x2 D wðT ¡ t ¡ lÞ C p1O x1O
E

If r < 0; Max ¡ ½aðt d x1R 1 ¡ d Þr C blr C ð1 ¡ a ¡ bÞx2 r 

x1R ;t;l;x2 (7b)
subject to p1R x1R C x2 D wðT ¡ t ¡ lÞ C p1O x1O
E

The steps involved in solving (7a) and (7b) are outlined in Appendix 1. Assuming an interior
solution, the optimal values for the input and final goods variables are the same for both ranges of r:

w  p1O E 
x1R  D h i      TC x ; (8)

dr p dr ¡ 1 r ¡ 1
1 1
r¡1
1
r¡1 w 1O
a w ð1 ¡ dÞ
d 1R 1
1¡a¡b C bwr
C 1¡d
p1R

1  p1O E 
t D h ir ¡1 1  r ¡1 1  r ¡1 1  TC x ; (9)

dr w 1O
addr ¡ r C 1 1
w ð1p¡1RdÞrðd ¡ 1Þ 1
1¡a¡b C wr
b C 1d

1  p1O E 
l D h ir ¡1 1  h
irð1 ¡ dÞ r ¡ 1 T C x ; (10)
w 1O
1

1C b
ð1 ¡ a ¡ bÞwr C ðÞ
b 1
a d
dr p1R
ð1 ¡ dÞw

w  p1O E 
x2  D h ir ¡1 1 h ir ¡1 1 T C x : (11)
rð1 ¡ dÞ w 1O
1C ð1 ¡ a ¡ bÞwr
b C 1¡a¡b w
a d ð Þdr ð1p¡1RdÞ

Using optimal housing and time from (8) and (9), we obtain the number of children desired, c ,
via (2)10:

1  p1O E 
c D h i1   r ¡1 1  r ¡1 1  h i1 ¡ d T C x : (12)
d 1 ¡ r ¡ d p1R d ¡ 1 r ¡ 1 w 1O
ad w ð1 ¡ dÞ 1
1¡a¡b C b
wr
C wð1 ¡ dÞ
p1R
ð dÞ
1 d

Similarly, using (9) and (10), we can obtain optimal labour, L , from (4):
8 1 9
>
>   >
>
>
> h

 i 1  1   1 >
>
>
> dr ¡ r C 1 1 dr p1R rðd ¡ 1Þ r ¡ 1 r¡1 r r¡1 >
>
 >
< ad w 1¡d
1
1¡a¡b C w
b C d >
1
=
p1O E
L D T ¡ T C x 1 : (13)
w 1O >
> C >
>
>
> h i  h irð1 ¡ dÞ r1¡ 1 >
>
>
> 1 >
>
> 1C
: b r¡1
b 1 dr p1R
C a ð d Þ ð1 ¡ dÞw >
;
ð1 ¡ a ¡ bÞwr

A sharp-eyed reader may note from (13) that a household holding a sufficiently large endowment of
E
housing wealth (p1O x1O ) relative to wages might choose a negative value of labour if uncon-
11
strained. We shall focus here on the interior solution where the household chooses to supply non-
negative labour.12
In the analogous case where owning is cheaper than renting (p1O < p1R ), the household’s
problem becomes one of finding the optimal adjustment up or down from its initial endowment of
 NEW ZEALAND ECONOMIC PAPERS 287

owner-occupied housing, ~x1O :

If 0 < r < 1; Max a½t d ðx1O

E
C ~x1O Þ1 ¡ d r C blr C ð1 ¡ a ¡ bÞx2 r
~x1O ;t;l;x2 (14a)
subject to p1O ~x1O C x2 D wðT ¡ t ¡ lÞ
If r < 0; Max ¡ fa½t d ðx1O
E
C ~x1O Þ1 ¡ d r C blr C ð1 ¡ a ¡ bÞx2 r g
~x1O ;t;l;x2 (14b)
subject to p1O ~x1O C x2 D wðT ¡ t ¡ lÞ

Again assuming an interior solution, the optimal values for both ranges of r are

wT
~x1O  D h i  r1¡ 1  r r 1¡ 1 

dr
p dr ¡ 1 r ¡ 11
a wd 1O
1¡d
1
1¡a¡b C wb C 1p¡1Od
8 9
>
> >
> (15)
>
< >
=
1
¡ x1O 1 ¡ h
E
i1   r1¡ 1  r r 1¡ 1 
>
>
dr >
>
>
: ap1O rðd ¡ 1Þ wd ð1 ¡1 dÞ
dr ¡ 1 r ¡ 1 1 C wb C 1 ¡1 d>
;
1¡a¡b

1  p1O E 
t D h ir ¡1 1  r ¡1 1  r ¡1 1  TC x ; (16)

dr w 1O
addr ¡ r C 1 1
w ð1p¡1OdÞrðd ¡ 1Þ 1
1¡a¡b C wr
b C 1d

1  p1O E 
l D h ir ¡1 1  h
irð1 ¡ dÞ r ¡ 1 T C x ; (17)
w 1O
1

1C b
ð1 ¡ a ¡ bÞwr C ð Þdr
b 1
a d
p1O
ð1 ¡ dÞw

w  p1O E 
x2  D h ir ¡1 1 h ir ¡1 1 T C x ; (18)
rð1 ¡ dÞ w 1O
1C ð1 ¡ a ¡ bÞwr
b C 1¡a¡b w
a d ð Þdr ð1p¡1OdÞ
1  p1O E 
c D h ir ¡1 1  r ¡1 1  r ¡1 1  h i1 ¡ d TC x ; (19)
w 1O
add w1 ¡ r ¡ d ð1p¡1OdÞd ¡ 1 1
1¡a¡b C wr
b C p1O
wð1 ¡ dÞ ð 1 d
d Þ
8 1 9
>
>   >
>
>
> h

 ir1¡ 1  r1¡ 1  r r 1¡ 1 >
>
>
> ¡ >
Cd>
dr ¡ r C 1 1 dr p rðd 1Þ
 >
< ad w
1O
1¡d
1
1¡a¡b C b
w 1 >
=
p1O E
L D T ¡ T C x 1 : (20)
w 1O >
> C >
>
>
> h ir 1¡ 1  h irð1 ¡ dÞ r1¡ 1 >
>
>
> >
>
>
: 1 C ð1 ¡ ab¡ bÞwr C ba ð1d Þ ð1p¡1OdÞw
dr >
;

For either type of housing tenure, our main equation of interest is the household’s optimal num-
ber of children c , whether for renters (12) or owners (19). For renters, this is a function of the price
of rental housing ðp1R Þ, and if an endowment of housing is owned initially, its price ðp1O Þ: For own-
ers, this is only a function of p1O . Also note that (12) and (19) are functions of the household’s will-
ingness to substitute between children and other ‘goods’ ðrÞ. How might we expect c , and the
underlying inputs, to respond to an increase in the cost of rent or the price of housing, p1R or p1O ?
To answer this, we take the derivatives of equations ((8)/(15)), ((9)/(16)), and ((12)/(19)) with
respect to p1R and p1O , and sign them where possible. The working is outlined in Appendix 2, and
summarised in Table 1 for the case where renting is preferable (renters), and in Table 2 for the case
where owning is preferable (home owners). We consider comparative statics for ‘small’ changes in
the price of rental or owned housing that preserve the price ranking of the two tenure options.13
288 M. LIU AND J. CLARK

Table 1. The effects of an increase in the price of rental or owned housing on the fertility of renters (p1R < p1O ).
Increase p1R Increase p1O
x1R  E
x1O D0 r<0 ¡ 0
0<r<1 ¡ 0
E
x1O >0 r<0 ¡ C
0<r<1 ¡ C

t E
x1O D0 r<0 C 0
0<r<1 ¡ 0
E
x1O >0 r<0 C C
0<r<1 ¡ C

c E
x1O D0 r<0 ¡ 0
0<r<1 ¡ 0
E
x1O >0 r<0 ¡ C
0<r<1 ¡ C

Table 2. The effects of an increase in the price of owned or rental housing on the fertility of home owners (p1O < p1R ).
Increase p1O Increase p1R
~x1O  E
x1O D0 r<0 ¡ 0
0<r<1 ¡ 0
E
x1O >0 r<0 ¡a 0
0<r<1 ¡ 0

t E
x1O D0 r<0 C 0
0<r<1 ¡ 0
E
x1O >0 r<0 C 0
0<r<1 ¡ if x1O
E
small/moderatea 0
C if x1O
E
largea

c E
x1O D0 r<0 ¡ 0
0<r<1 ¡ 0
E
x1O >0 r<0 ¡ if x1O
E
small/moderatea 0
C if x1O
E
largea
0<r<1 ¡a 0
‘x1O
E E
large’ refers to values of x1O that are large, but not so large we leave the interior solution where L is non-negative. For
the parameter values used in the simulation, we found that relative to a wage of 1, only ‘extreme’ combinations of x1O E
(10)
and p1O (2.5) moved outside the interior solution (L D ¡0.5).
a
Found by simulation.

While comparative statics for both p1R and p1O could be signed for renters, we had to resort to simu-
lations in most cases for home owners for p1O if x1O E
> 0: For simulations, we set r D ¡ 0:5 for r < 0
and r D 0:5 for 0 < r < 1, a D b D 3, d D 0:5, w D 1, and T D 24; while varying our parameters of
1
E E
interest, p1O and x1O . We increased p1O from 1 to 2.5 in increments of 0.5, and x1O from 0.5 to 10 in
increments of 0.5, and then 1.
From Table 1, we see that for households best off renting, an increase in the price of rental hous-
ing unambiguously lowers desired number of children. This is true regardless of whether the house-
hold has an initial endowment of owned housing, or whether it has a high or low willingness to
substitute between goods. Underlying this, renters with a high willingness to substitute between
goods both rent less housing and devote less time to raising children. In contrast, renters with a low
willingness to substitute between goods rent less housing but partially offset this by devoting more
time to raising children. That is, the decline in children caused by a rise in rental costs for high-sub-
stitution renters exceeds that for low-substitution renters.
Still with renters, an increase in the price of owned housing is relevant only in so far as it raises
their wealth from selling off any initial housing endowment. For such renters, a rise in owned hous-
ing price increases their desired number of children, increasing both their optimal amount of rental
housing and time spent raising children. For renters lacking such an endowment, a rise in the price
of owned housing has no effect.
 NEW ZEALAND ECONOMIC PAPERS 289

From Table 2, we see that for those best off owning housing, the effect of an increase in the price
of owned housing is sensitive to the quantity already owned, and to the household’s willingness to
substitute between goods. For ‘new entrants’ who own zero housing to begin with, an increase in the
price of owner-occupied housing lowers their desired number of children. Underlying this, the
house price rise decreases the quantity of housing they purchase; it also decreases the time spent
raising children for high-substitution households, but increases it for low-substitution households.
That is, the decline in children caused by a rise in housing price for new entrant owners with high
substitution exceeds that for those with low substitution.
For those best off owning who already possess an initial endowment of housing, responses to
house prices changes are even more sensitive to elasticity of substitution between goods. For those
with a high willingness to substitute between goods, an increase in housing price appears to decrease
desired family size in all the simulations that we tried, regardless of wealth effects from initial hous-
ing owned. Underlying this, high-substitution households reduce their total holding of housing.
They also reduce their time spent raising children if their initial stock of housing is low, but increase
it if their initial stock of housing is sufficiently high.
In contrast, for those best off owning, with an initial endowment, and a low willingness to substi-
tute between goods, an increase in housing price appears to decrease desired family size only if a
low-to-moderate amount of initial housing is owned, as the substitution effect outweighs the offset-
ting wealth effect. With enough initial housing, however, an increase in housing price appears to
raise desired number of children. Underlying this, low-substitution households reduce their total
holding of housing, but increase their time spent raising children.
Still with home owners, an increase in the price of rental housing has no effect on the number of
children desired, changing neither their desired total holding of housing, nor their time spent raising
children.
The above theoretical results suggest that an increase in the price of rental housing is always anti-
natal for renters, but has no effect on home owners. On the other hand, an increase in the price of
owned housing has a more complex effect on home owners’ desired family size than might be casually
expected. While higher housing prices might be pro-natal for home owners because of wealth effects,
they may also be anti-natal if the household’s willingness to substitute away from children to other
‘goods’ is sufficiently great, or if their initial quantity of housing owned is sufficiently modest.
Exploring this latter finding more formally for home owners, we show in equation (B.4) in

Appendix 2 that @p@c1O is a function of x1O E
. How initial housing influences the effect of housing price
2 
on a home owner’s desired family size can be shown from the cross partial derivative @xE@ @p
c
1O
:
1O

@ 2 c 1
Dh   r r1¡ 1  h
@x1O
E @p
1O
p d ¡ 1 ir1¡ 1  r1¡ 1 i1 ¡ d
C wð1p1O
d
add w1 ¡ r ¡ d 1¡d
1O 1
1¡a¡b C w
b ¡ dÞ ð1d Þ
8 h i 1  r1¡ 1  r r1¡ 1   d h i1 ¡ d 9
>
> ðd ¡ 1Þ r ¡ 1 >
>
> 1 ¡ d
< r ¡ 1 ap1O
d ¡ r d
dw 1 ¡ r ¡ d
ð 1 ¡ dÞ
1 1
1¡a¡b C b w C ðd ¡ 1Þ p1O d wð1 ¡ dÞ
1 1
p  1 >=
1O
h   C :
>
>
 i1 
d 1 ¡ r ¡ d p1O d ¡ 1 r ¡ 1
r1¡ 1  r r1¡ 1 h i1 ¡ d w w>
>
>
: ad w 1 C bw p1O
C wð1 ¡ dÞ ðdÞ
1 d >
;
1¡d 1¡a¡b

(21)
With some effort, it can be shown that the sign of (21) varies as follows:

h ir ¡1 1
1 r ¡ dr ¡ 1 dr p1O rð1 ¡ dÞ
@c 2 
> r¡d > ¡ ad w ð 1 ¡ d Þ
0 if  r ¡1 1  r ¡1 1 : (22)
@x1O
E @p
1O < r¡1 < 1
1¡a¡b C wr
b
290 M. LIU AND J. CLARK

We obtain a clear positive sign for (22) when r < 0; because the left-hand side of the condition is
then positive and the right-hand side negative. In contrast, when 0 < r < 1, the left-hand side of
(22) is non-negative if r  d and negative if r > d, while the right-hand side is always negative.
Therefore, we obtain a clear positive sign for (22) if r  d, but cannot clearly sign it if r > d.
This cross partial derivative sheds light on our simulation results in Table 2 regarding how house
prices affect fertility for home owners. Recall that for home owners with a low willingness to substi-

tute between goods ðr < 0Þ; simulations showed that @p@c1O < 0 for small-to-moderate values of x1O E
,
@c @2 c
and @p1O > 0 for large values of x1O
E
. The positive cross partial derivative @x1O
E @p
1O
> 0 implies that if
the household owns little housing initially, the effect of an increase in housing price on fertility is
negative, but less so, the more initial housing the household owns. Conversely, when the household
owns much housing initially, the effect of an increase in housing price on the demand for children is
E
positive, and this effect strengthens as x1O increases. This implies that there will be a critical value of
@c
x1O , say x1O , where @p1O crosses from negative to positive. At xE1O , a marginal increase in the price of
E E

housing has no effect on the home owner’s demand for children
substitution and wealth effects
exactly cancel out.
On the other hand, for home owners with a high willingness to substitute between goods
 2 
(0 < r < 1), @p@c1O < 0 always but @xE@ @p
c
1O
can be positive or negative. If we assume that rd (i.e. a
1O
household’s willingness to substitute between goods is no greater than the importance of time rela-
2 
tive to housing in the production of children), then @xE@ @p
c
1O
> 0. That is, for high-substitution house-
1O
holds, the effect of an increase in the price of housing on desired number of children is always
E
negative, but becomes weaker as x1O increases.
The key implications to take away from this discussion are that higher rental prices should lower
fertility for renters, while higher housing prices may lower or raise fertility for home owners depend-
ing both on the quantity of initial housing they own, and on their willingness to substitute between
children, leisure, and all other goods. Higher house prices are predicted to increase the number of
children desired by home owners only if they have sufficient initial housing and a low substitution
between children and other consumption. Under all other conditions, higher house prices are pre-
dicted to lower fertility for home owners.

4. Aggregate correlations
While static, the model above could provide a framework for testing empirically the effects of chang-
ing house prices and rental costs on household fertility, including in New Zealand. Such empirical
tests would ideally follow individual women and their families over time, to control for their usually
unobserved characteristics such as strength of preferences and elasticity of substitution between chil-
dren and other ‘goods’. While such an analysis is beyond the scope of this paper, we do provide
some simple correlations between housing price or rental cost and fertility in New Zealand at pub-
licly available aggregated census units. We describe our measures below, and then illustrate the cor-
relations we find.
For measures of fertility and rental cost, we use the New Zealand Census rounds of 2001, 2006,
and 2013, rebased to 2013 geographic boundaries.14 We use ‘area units’ as our unit of observation,
as opposed to coarser aggregations such as ‘territorial authority’.15 For fertility, the 2006 and 2013
Censuses report the number of children born alive to females aged 15 years or above. This is
reported publicly in seven categories: no children, one child, two, three, four, five, and six or more
children. Arbitrarily assigning a mean value for the outcome of ‘six or more children’ to seven chil-
dren, we obtain a weighted average of the total number of children born alive to females aged 15
years and above for each area unit. Unfortunately, this fertility measure does not indicate how
recently a woman’s children were born. Since children born many years previously cannot be
 NEW ZEALAND ECONOMIC PAPERS 291

Table 3. Variable descriptions and descriptive statistics.

Standard
Variable Description Year N Mean deviation Min Max
Total children Mean number of children born alive to 2006 377 1.568 0.551 0.069 2.790
females aged 15 years and above
2013 377 1.552 0.537 0.038 2.930

Rent Real median weekly rent paid by 2001 364 $208.83 $70.66 $56.43 $451.46
households
2006 364 $231.34 $77.97 $49.75 $457.71

HPI Constructed annual real house price 2001 66 867.76 138.96 567.64 1145.30
index
2006 66 1604.92 265.14 1034.28 2260.06

affected by recent movements in housing or rental price, we focus on those area units with a greater
proportion of women of child-bearing age. To do this, we filter our area units to include only those
where the proportion of people aged 15
44 years (of all those aged 15 and above) was greater than
0.60 in 2006.16 We label this fertility variable Total Children. This yields a sample of 377 area units
with observations in both 2006 and 2013.
For rental cost, the census provides the median weekly rent paid by renting households in each
area unit. We adjust this for inflation using the CPI ‘All Groups’ series, and label this measure
Rent.17 To examine the effect of rental cost on subsequent fertility, we consider rent in the lagged
census years, 2001 and 2006. Descriptive statistics for Rent as well as Total Children are provided in
Table 3.
Unfortunately, the census does not ask people to estimate the current market price or value of
their homes. We, therefore, source data from Quotable Value Limited (QV) (2014, n.d.), a private
company that provides sales price and housing valuation data for New Zealand. In order for our
house price and census data to be compatible, we use the QV quarterly house price index (HPI) for
detached houses only. Unlike our Rent measure, the HPI is publicly released only at the coarser ter-
ritorial authority level (defined in endnote 15). The QV quarterly HPI is a quality-adjusted house
price measure, constructed using the Sales Price Appraisal Ratio (SPAR) method by relating freehold
open market sales data to valuation data. See Appendix 3 for more detail on the construction of the
HPI. To obtain an annual HPI, we take a simple average of the quarterly HPI. Then, as with Rent,
we adjust for inflation using the CPI ‘All Groups’ series. We again consider our house price measure,
HPI, in the years that lag the 2006 and 2013 censuses, namely 2001 and 2006. See Table 3 for
descriptive statistics.
The lack of HPI at area unit level might beg the question of whether we could use rental costs as a
proxy for housing price. Under standard theory, equilibrium house prices should reflect the dis-
counted present value of future rents, so that one might assume that house prices and rental rates
would correlate perfectly, and only one or the other measure would be necessary. However, interna-
tional researchers who have studied the relationship between house prices and rental rates have dif-
fering findings as to the extent to which one tracks the other. Hott and Monnin (2008) argue that,
in the short term, the rental market may be disconnected from the housing market due to institu-
tional constraints. Lin (1993) makes the distinction that house prices are determined by both hous-
ing consumption demand, and investment and speculative demand, while rental rates are
determined only by housing consumption demand, suggesting that rental rates may not be in line
with house prices. Empirically, Cheung, Tsang, and Mak (1995) find in Hong Kong that demand in
the housing and rental markets are not strongly correlated, likely due to market segmentation, sug-
gesting that renters and home owners are two different groups of people. In contrast, Hui and Zheng
(2012) find evidence also for Hong Kong to support the hypothesis that house prices and rental rates
are correlated. Thus, it is not clear ex ante whether movements in rental rates appropriately capture
movements in housing prices in New Zealand.
292 M. LIU AND J. CLARK

Figure 1. CPI rent component vs. quotable value quarterly house price index, 2001–2013. Data sources: With permission from
Statistics New Zealand and Quotable Value Limited.

As a first rough indicator, in Table 3, we see that the inflation-adjusted mean HPI (over territorial
authority units) increased by 84.9% between 2001 and 2006, a far greater increase than that occurred
for real median rent (10.8%) in our filtered area units.18 As a second rough indicator, in Figure 1, we
compare the annual percentage changes in the rent component of the consumer price index (CPI)
with the quotable value (QV) quarterly HPI for all residential housing from 2001 to 2013.19,20 From
Figure 1, we see that annual rental rate inflation has been low and relatively stable, remaining
between zero and five percent from the end of 2001 through to the end of 2013. In contrast, there
has been significant movements in house price inflation, with a period of especially high inflation
between 2003 and 2007, and significant deflation between 2008 and 2009 (global economic reces-
sion). Therefore, it seems that rental rates and house prices are not that well aligned in New Zealand,
and thus we retain both measures.
With this by way of background, Figures 2 and 3 provide separate scatterplots of changes in
lagged rental cost or lagged housing price and the change in average total number of children born

Figure 2. Change in total number of children vs. change in lagged real median weekly rent.
 NEW ZEALAND ECONOMIC PAPERS 293

Figure 3. Change in total number of children vs. change in lagged real HPI.

to women in New Zealand. Recall these are for those area units where adults of child-bearing age
make up at least 60% of the adult population in 2006. The correlation between lagged change in
Rent or HPI and subsequent change in Total Children is also provided. For Figure 3, (change in) fer-
tility measures for area units are assigned the HPI (change) of the broader territorial authority,
which can lead to multiple area units being assigned the same housing price change value.
While Figure 2 finds a negative correlation between real rental costs and total number of children,
it is weak (r D ¡0.08). A weak negative correlation occurs also in Figure 3 between real house prices
and total number of children (r D ¡0.06). While the negative sign of the correlation between rental
costs and fertility is consistent with our theoretical prediction for renters, and possibly consistent
for homeowners if they have a high willingness to substitute between family size and other ‘goods’
or low initial holdings of housing, the weakness of the magnitude of the correlation is suggestive. In
particular, correlations between changes in housing price or rental cost and fertility at aggregated
level cannot say anything causal, since they do not control for other household characteristics
known to affect fertility decisions, nor for households’ location self-selection between regions. None-
theless, the weak correlations do not provide strong prima facie evidence that fluctuations in rental
costs or housing price are having large effects on fertility and family size in New Zealand.

5. Conclusion
The extent to which fertility responds to economic incentives and price signals has been studied
extensively in the demography and economics literature. However, only limited attention has been
paid to the effect of the price of housing on fertility and family size. To the extent this has been
examined, few studies have recognised that the quantity of housing owned (Dettling & Kearney,
2014; Lovenheim & Mumford, 2013) and parents’ elasticity of substitution between children and
other ‘goods’ play an important role. This paper investigates the theoretical effect of an increase in
the price of housing or cost of renting on family size, with an illustration using aggregate data from
New Zealand.
Our theoretical model uses a CES household utility function defined over children, leisure, and all
other goods. This is combined with a Cobb
Douglas household production function in which chil-
dren are ‘produced’ using parental time and housing. Households choose to rent or own housing
based on the suitability of either tenure type for raising children, and the relative cost of each. Our
model demonstrates that the effect of an increase in house price or rental cost on fertility depends
on the quantity of housing already owned and on the household’s willingness to substitute between
294 M. LIU AND J. CLARK

children, leisure, and all other goods. While an increase in the cost of renting is predicted to decrease
the number of children desired by renting households, the effect of higher housing prices is ambigu-
ous for home owners. Higher housing prices would be pro-natal for those with substantial initial
holdings of housing and limited willingness to substitute between children and other ‘goods’, but
they would be anti-natal for those without much initial housing, or who are relatively willing to sub-
stitute away from larger families. A limitation of our predictions is that, for the case of home owners,
some of the comparative statics had to be obtained by simulation. We therefore cannot be sure that
they hold in general. For future research, our static, unitary household model could be extended to a
two-parent, two-period model by incorporating the wages and time constraints of the mother and
father separately and by modelling consumption over two periods. Doing so, we could gain insight
into the predicted amounts of time mothers spend with their children compared to fathers, and how
this affects the optimal number of children desired by a two-person household. A two-period model
could also provide information on fertility timing.
Using aggregate census data on the number of children born to women and rental costs, and
Quotable Value data on housing prices, we also examined correlations between changes in either
lagged rental cost or housing price and subsequent changes in fertility in New Zealand. Using rental
costs and number of children at the census ‘area unit’ level, filtered to focus on areas with more
adults of child-bearing age, and housing price at ‘territorial authority’ level, we find negative but
weak correlations between either lagged changes in rental costs (r D ¡0.08) or housing price (r D
¡0.06), and subsequent change in average cumulative number of children born to women. The signs
of these correlations at aggregate level are consistent with our individual-level model, though the
magnitudes might suggest the negative effects of rental cost or housing price on fertility could turn
out to be slight. To establish whether higher housing prices and rental costs do indeed have a minor
negative causal effect on fertility, empirical analysis is needed that follows individual households
over time and addresses location choice.

Acknowledgments
We wish to thank Arthur Grimes and Andrea Menclova for their helpful feedback on this project, as
well as three anonymous referees. We also thank Quotable Value for providing us with house price
index data.

Notes
1. Delayed household formation may imply delayed plans to have children.
2. The total fertility rate is defined as the number of children born to a woman, by the end of her reproductive
period, if she experienced the current age-specific fertility rates throughout her child-bearing years.
3. The coefficient on the house price variable is negative, while the coefficient on the interaction term is positive and
of a greater magnitude than the former.
4. x2 can be thought of as a composite good that excludes children and leisure.
5. The CES utility function assumes that all goods, here including children, are normal goods. It may be objected
that there is historical evidence that people have fewer children as incomes increase. However, a driving factor
behind the increase in household incomes has been the increase in female labour force participation and wom-
en’s wages. For simplicity, we have adopted a unitary household wage in our model and do not address this issue.
One way to incorporate the possibility that people have fewer children as incomes increase would be to allow for
a choice between quantity and (investment-related) ‘quality’ of children (Becker, 1960).
6. The household’s elasticity of substitution between any pair of the three goods, s, is defined as s D 1 ¡1 r.
7. While we do not calculate Cobb
Douglas outcomes, we cover the cases that approach those outcomes from
above and below.
8. We recognise that actual household decisions regarding how many children to try for is driven both by women’s
and men’s opportunity costs of time. However, since our focus is on the effects of housing price or rental costs,
we use a unitary wage/household model.
9. When renting is cheaper than owning, the household will sell off all endowed housing. When owning is cheaper
than renting, the household may choose to sell off some endowed housing, in which case ~x1O < 0.
 NEW ZEALAND ECONOMIC PAPERS 295

10. In this case where renting is optimal, x1O T

D 0 since any endowed housing is sold.
11. This possibility also holds for the case where owning is optimal, as seen in (20).
12. The corner solution does not add appreciable insights, but is available from the authors upon request.
13. This rules out ‘big changes’ that reverse the ranking of the two types of housing prices that would result in the
household switching from renting to owning, or vice versa.
14. The New Zealand Census is normally held every five years, but the 2011 round was delayed until 2013 due to the
Canterbury earthquake of February 2011.
15. At the 2013 Census, there were 2012 ‘area units’ in New Zealand, with an average population of roughly 2000
each. A ‘territorial authority’ is a city council or district council in New Zealand, and is an aggregate of area units.
At the 2013 Census, there were 67 territorial authorities in New Zealand, with an average population of roughly
62,600 each.
16. The publicly released census does not report the age groups for females and males separately, and so we control
for the proportion of both sexes in the child-bearing age bracket for women.
17. Note that Rent does not control for the quantity (size) or type of residential housing.
18. Goodyear and Fabian (2014) similarly document a divergence in growth of housing prices and rental costs for the
Auckland region in particular.
19. The rent component of the CPI (actual rentals for housing) is a subgroup of the ‘housing and household utilities’
group in the CPI. It measures the rent paid by tenants to landlords and comprises three price indices: private
and local authority rentals, state rentals, and education accommodation. As of the June 2006 quarter, the rent
component had an expenditure weight of 6.87% of the CPI, for which 90% was attributable to private and local
authority rentals (Statistics New Zealand, 2015, n.d.).
20. For this comparison, we use the QV HPI for ‘all residential’ housing (rather than detached houses only) due to
the nature of the rent component of the CPI (the private and local authority rentals index includes apartments,
flats, town houses, and houses). The HPI for ‘all residential’ housing takes into account detached houses, flats,
and apartments. More about the HPI can be found in Appendix 3.

Disclosure statement
No potential conflict of interest was reported by the authors.

References
Becker, G.S. (1960). An economic analysis of fertility. In National Bureau of Economic Research Series (Ed.), Demo-
graphic and economic change in developed countries (pp. 209
240). New York, NY: Columbia University Press.
B€orsch-Supan, A.H. (1986). Household formation, housing prices, and public policy impacts. Journal of Public Eco-
nomics, 30, 145
164.
Cain, G.G., & Dooley, M.D. (1976). Estimation of a model of labor supply, fertility, and wages of married women.
Journal of Political Economy, 84, S179
S199.
Cheung, Y.-L., Tsang, S.-K., & Mak, S.-C. (1995). The causal relationships between residential property prices and
rentals in Hong Kong: 1982
1991. The Journal of Real Estate Finance and Economics, 10, 23
35.
Clark, W.A.V. (2012). Do women delay family formation in expensive housing markets? Demographic Research, 27,
1
24.
Curtis, M.A., & Waldfogel, J. (2009). Fertility timing of unmarried and married mothers: Evidence on variation across
U.S. cities from the Fragile Families and Child Wellbeing Study. Population Research and Policy Review, 28,
569
588.
DeFronzo, J. (1980). Female labor force participation and fertility in 48 states: Cross-sectional and change analyses for
the 1960
70 decade. Sociology and Social Research, 64, 263
278.
Dettling, L.J., & Kearney, M.S. (2014). House prices and birth rates: The impact of the real estate market on the deci-
sion to have a baby. Journal of Public Economics, 110, 82
100.
Feyrer, J., Sacerdote, B., & Stern, A. D. (2008). Will the stork return to Europe and Japan? Understanding fertility
within developed nations. The Journal of Economic Perspectives, 22, 3
22.
Giannelli, G.C., & Monfardini, C. (2003). Joint decisions on household membership and human capital accumulation
of youths. The role of expected earnings and local markets. Journal of Population Economics, 16, 265
285.
Goodyear, R., & Fabian, A. (2014). Housing in Auckland: Trends in housing from the census of population and dwell-
ings 1991 to 2013. Statistics New Zealand. Retrieved from www.stats.govt.nz
Hott, C., & Monnin, P. (2008). Fundamental real estate prices: An empirical estimation with international data. The
Journal of Real Estate Finance and Economics, 36, 427
450.
Hughes, M.E. (2003). Home economics: Metropolitan labor and housing markets and domestic arrangements in
young adulthood. Social Forces, 81, 1399
1429.
296 M. LIU AND J. CLARK

Hui, E.C.-M., & Zheng, X. (2012). The dynamic correlation and volatility of real estate price and rental: An application
of MSV model. Applied Economics, 44, 2985
2995.
Kim, H.S. (2014). Female labour force participation and fertility in South Korea. Asian Population Studies, 10,
252
273.
Kostelecky, T., & Vobecka, J. (2009). Housing affordability in Czech regions and demographic behaviour
does hous-

ing affordability impact fertility? Sociologicky Casopis/Czech Sociological Review, 45, 1191
1213.
Lalive, R., & Zweim€ uller, J. (2009). How does parental leave affect fertility and return to work? Evidence from two nat-
ural experiments. Quarterly Journal of Economics, 124, 1363
1402.
Lin, C.-C.S. (1993). The relationship between rents and prices of owner-occupied housing in Taiwan. The Journal of
Real Estate Finance and Economics, 6, 25
54.
Lovenheim, M.F., & Mumford, K.J. (2013). Do family wealth shocks affect fertility choices? Evidence from the housing
market. Review of Economics & Statistics, 95, 464
475.
Matysiak, A., & Vignoli, D. (2008). Fertility and women’s employment: A meta-analysis. European Journal of Popula-
tion, 24, 363
384.
Pan, L., & Xu, J. (2012). Housing price and fertility rate. China Economic Journal, 5, 97
111.
Quotable Value Limited (2014). Quotable value quarterly house price index. Obtained Upon Request.
Quotable Value Limited (n.d.). How is the HPI calculated? Retrieved from http://www.qv.co.nz/help/FAQ
Ridao-Cano, C., & McNown, R. (2005). The effect of tax-benefit policies on fertility and female labor force participa-
tion in the United States. Journal of Policy Modeling, 27, 1083
1096.
Simon, C. J., & Tamura, R. (2009). Do higher rents discourage fertility? Evidence from U.S. cities, 1940
2000.
Regional Science and Urban Economics, 39, 33
42.
Statistics New Zealand (2014). 2013 Census meshblock dataset. Retrieved from http://www.stats.govt.nz/Census/2013-
census/data-tables/meshblock-dataset.aspx#excelfiles
Statistics New Zealand (2015). CPI Level 3 classes for New Zealand: Actual rentals for housing. Retrieved from Statis-
tics New Zealand Infoshare Database. Retrieved from http://www.stats.govt.nz/infoshare/ViewTable.aspx?pxI
D=e1cfd3b0-b06d-4d2d-bc08-bdb1920f6692
Statistics New Zealand (n.d.). Rented dwellings in the CPI. Retrieved from http://www.stats.govt.nz/browse_for_stats/
economic_indicators/CPI_inflation/rented-dwellings-in-the-cpi.aspx
Vignoli, D., Rinesi, F., & Mussino, E. (2013). A home to plan the first child? Fertility intentions and housing condi-
tions in Italy. Population, Space and Place, 19, 60
71.
Walker, J.R. (1995). The effect of public policies on recent Swedish fertility behavior. Journal of Population Economics,
8, 223
251.
Whittington, L.A. (1992). Taxes and the family: The impact of the tax exemption for dependents on marital fertility.
Demography, 29, 215
226.
Whittington, L.A., Alm, J., & Peters, H.E. (1990). Fertility and the personal exemption: Implicit pronatalist policy in
the United States. The American Economic Review, 80, 545
556.
Yi, J., & Zhang, J. (2010). The effect of house price on fertility: Evidence from Hong Kong. Economic Inquiry, 48,
635
650.
 NEW ZEALAND ECONOMIC PAPERS 297

Appendices

Appendix 1. Solving the model

Case I: Renting is cheaper than owning (p1R < p1O ):
For 0 < r < 1, the household’s optimisation problem is given by

Max aðt d x1R 1 ¡ d Þr C blr C ð1 ¡ a ¡ bÞx2 r

x1R ;t;l;x2 (7a)
subject to p1R x1R C x2 D wðT ¡ t ¡ lÞ C p1O x1O
E

The Lagrangian for this problem is

L D aðt d x1R 1 ¡ d Þr C blr C ð1 ¡ a ¡ bÞx2 r C λ½wðT ¡ t ¡ lÞ C p1O x1O

E
¡ p1R x1R ¡ x2 : (A:1)

The first-order conditions are

@L
D arð1 ¡ dÞt dr x1R r ¡ dr ¡ 1 ¡ p1R λ D 0 (A:2)
@x1R
@L
D adrt dr ¡ 1 x1R rð1 ¡ dÞ ¡ wλ D 0; (A:3)
@t
@L
D brlr ¡ 1 ¡ wλ D 0; (A:4)
@l
@L
D ð1 ¡ a ¡ bÞrx2 r ¡ 1 ¡ λ D 0; (A:5)
@x2
@L
D wðT ¡ t ¡ lÞ C p1O x1O
E
¡ p1R x1R ¡ x2 D 0; (A:6)
@λ

Dividing (A.2) by (A.3) and simplifying yields

dp1R x1R
tD : (A:7)
ð1 ¡ dÞw

Dividing (A.4) by (A.3) and simplifying yields

h ir ¡1 1
a dr ¡ 1
lD dt x1R rð1 ¡ dÞ : (A:8)
b

Substituting (A.7) into (A.8) and simplifying yields

(  dr ¡ 1 )r ¡1 1
a dr p1R
lD d x1R : (A:9)
b ð1 ¡ dÞw

Dividing (A.5) by (A.4) and simplifying yields

 r ¡1 1
b
x2 D l: (A:10)
ð1 ¡ a ¡ bÞw
298 M. LIU AND J. CLARK

Substituting (A.9) into (A.10) and simplifying yields

  dr   r ¡1 1
a d p1R dr ¡ 1
x2 D x1R : (A:11)
1¡a¡b w 1¡d

Equations (A.7), (A.9), and (A.11) describe t; l; and x2 , respectively, as functions of x1R . Substituting
each into (A.6) and simplifying, we solve for x1R  :

w  p1O E 
x1R  D h i       T C x : (A:12)

dr 1 1 1
w 1O
a wd ð1p¡1RdÞdr ¡ 1
r¡1 r¡1 r¡1
1
1¡a¡b C wr
b C p1R
1¡d

We next substitute (A.12) into (A.7), (A.9), and (A.11), respectively, to obtain the optimal values oft,
l, and x2 . Substituting (A.12) into (A.7) and simplifying yields

1  p1O E 
t D h ir ¡1 1  r ¡1 1  r ¡1 1  TC x : (A:13)

dr w 1O
addr ¡ r C 1 1
w ð1p¡1RdÞrðd ¡ 1Þ 1
1¡a¡b C wr
b Cd 1

Substituting (A.12) into (A.9) and simplifying yields

1  p1O E 
l D h ir ¡1 1  h
irð1 ¡ dÞ r ¡ 1 T C x : (A:14)
w 1O
1

1C b
ð1 ¡ a ¡ bÞwr C ðÞ
b 1
a d
dr p1R
ð1 ¡ dÞw

Substituting (A.12) into (A.11) and simplifying yields

w  p1O E 
x2  D h ir ¡1 1 h ir ¡1 1 T C x : (A:15)
rð1 ¡ dÞ w 1O
1C ð1 ¡ a ¡ bÞwr
b C 1¡a¡b w
a d ð Þdr ð1p¡1RdÞ

We have now obtained the solutions to x1R , t, l, and x2 in the optimisation problem described by (7a).
For households where r < 0, the analogous optimisation problem is

Max ¡ ½aðt d x1R 1 ¡ d Þr C blr C ð1 ¡ a ¡ bÞx2 r 

x1R ;t;l;x2 (7b)
subject to p1R x1R C x2 D wðT ¡ t ¡ lÞ C p1O x1O
E

The Lagrangian for this problem is

L D ¡ aðt d x1R 1 ¡ d Þr ¡ blr ¡ ð1 ¡ a ¡ bÞx2 r C λ½wðT ¡ t ¡ lÞ C p1O x1O

E
¡ p1R x1R ¡ x2 : (A:16)

Once combined, the first-order conditions and solutions are identical to the case where 0 < r < 1.
Case II: Owning is cheaper than renting (p1O < p1R ):
The set-up for this problem is already given in (14a) and (14b), and it is solved analogously to the
case above where renting is cheaper than owning.
 NEW ZEALAND ECONOMIC PAPERS 299

Appendix 2. Computing comparative statics

Case I: Renting is cheaper than owning (p1R < p1O ):
A. Investigating the effects of an increase in p1R on c , x1R  , and t 

To find @p@c1R , we differentiate (12) with respect to p1R and simplify:

h i 1  r ¡1 1  r ¡1 1   d h i1 ¡ d
d ¡ r d 1 ¡ r ¡ d 1 d ¡ 1 r¡1
@c
1¡d
r¡1 ap1R dw ð1 ¡ d Þ 1
1¡a¡b C wbr C ðd ¡ 1Þ p1R1 d wð11¡ dÞ  p1O E 
D h ir ¡1 1  r ¡1 1  r ¡1 1  h i1 ¡ d 2 TC x :
@p1R w 1O
add w1 ¡ r ¡ d ð1p¡1RdÞd ¡ 1 1
1¡a¡b C wr
b C p1R
wð1 ¡ dÞ ð 1 d
d Þ

(B:1)

Amongst the terms in the first part of (B.1), we know that

½ap1R d ¡ r dd w1 ¡ r ¡ d ð1 ¡1 dÞd ¡ 1 r ¡ 1 ½ð1 ¡ a1 ¡ bÞr ¡ 1 C ðwbr Þr ¡ 1  > 0, ðd ¡ 1Þðp1R1 dÞd ½wð11¡ dÞ1 ¡ d < 0, and that the
1 1 1

denominator is positive. Since 0 < d < 1, r1 ¡¡ 1d < 0 for both 0 < r < 1 and r < 0. Combining the above
information, we can conclude that the first part of (B.1) is always negative. The second part of (B.1),

T C pw1O x1O
E
, is always positive. Therefore, @p@c1R < 0, regardless of r and x1O E
.
We use the same reasoning and intuition as above to compute and sign the remaining derivatives.

To find @x @p1R , we differentiate (8) with respect to p1R and simplify:
1R

h
dr ir ¡1 1  r ¡1 1  r ¡1 1 
1 ¡ dr
ap1R rðd ¡ 1Þ wd ð1 ¡1 dÞdr ¡ 1 1
C wbr C d ¡1 1
@x1R  r¡1 1¡a¡b
D h  r ¡1 1  r ¡1 1  2 ðwT C p1O x1O
E
Þ: (B:2)
@p1R
dr p dr ¡ 1 ir ¡1 1 
a wd ð1 ¡1RdÞ 1
1¡a¡b C w
b
r
C p
1¡d
1R


The first part of (B.2) is always negative, and the second part is always positive. Therefore, @x@p1R1R < 0,
E
regardless of r and x1O .
@t 
To find @p1R , we differentiate (9) with respect to p1R and simplify:

h
dr ir ¡1 1  r ¡1 1  r ¡1 1 
dr ¡ 2r C 1 dr ¡ r C 1 rðd ¡ 1Þ
rð1 ¡ dÞ
ð Þ C wbr  p1O E 
1 1 1
ap1R d
@t  r¡1 w 1¡d 1¡a¡b
D h i 1  r ¡1 1  r ¡1 1  2 TC x :
@p1R

dr ¡ r C 1 1 dr p1R rðd ¡ 1Þ r ¡ 1
w 1O
ad w ð1 ¡ dÞ 1
1¡a¡b C wbr C 1d

(B:3)

The first part of (B.3) is negative if 0 < r < 1 and positive if r < 0. The second part of (B.3) is always
 
positive. Therefore, @p@t1R < 0 if 0 < r < 1, and @p@t1R > 0 if r < 0. This is true for both x1O
E
D 0 and
E
x1O > 0.
B. Investigating the effects of an increase in p1O on c , x1R  , and t 
By inspection of (12), (8), and (9), we see that the household’s optimal amounts of c, x1R , and t are
: Therefore, if x1O D 0, we can conclude that @p@c1O D 0, @x@p1R1O D 0, and
E E 
linear in p1O , and dependent on x1O
@p1O D 0. However, if x1O > 0, then @p1O > 0, @p1O > 0, and @p1O > 0. These results hold for either range
   
@t E @c @x1R @t

of r.
Case II: Owning is cheaper than renting (p1O < p1R ):
A. Investigating the effects of an increase in p1O on c , ~x1O  , and t 
300 M. LIU AND J. CLARK


To find @p@c1O , we differentiate (19) with respect to p1O and simplify:

@c 1
Dh   r r 1¡ 1 
@p1O
p d ¡ 1 ir 1¡ 1  r1¡ 1 h
p1O
i1 ¡ d
d
add w1 ¡ r ¡ d 1O
1¡d
1
1¡a¡b C w
b C wð1 ¡ dÞ ð1d Þ
8 h i 1  r 1¡ 1  r r1¡ 1   d h i1 ¡ d 9
>
> 1 ¡ d ap d ¡ r dd w1 ¡ r ¡ d ð 1 Þd ¡ 1 r ¡ 1 >
>
>
< r¡1 1 C w C ðd ¡ 1 1 >
 p1O E  1 E =
1O 1¡d 1¡a¡b b 1Þ p1O d wð1 ¡ dÞ

h   T C x C x :
>
>
d ¡ 1 ir1¡ 1  r 1¡ 1  r r 1¡ 1 h i1 ¡ d w 1O w 1O >
>
>
: add w1 ¡ r ¡ d 1p¡1Od 1 C wb C wð1p1O ð1d Þ
d >
;
1¡a¡b ¡ dÞ

(B:4)

The first term of (B.4) is always negative. For the second term of (B.4), w1 x1O E
D 0 if x1O
E
D 0, and
@c
1 E
x
w 1O > 0 if x E
1O > 0. Thus, @p1O can be signed negative if x E
1O D 0 for either range of r, but cannot be
E
signed if x1O > 0.
x1O 
To find @~ @p1O , we differentiate (15) with respect to p1O and simplify:

h
 ir1¡ 1  r1¡ 1  r r 1¡ 1 
1 ¡ dr rðd ¡ 1Þ d dr 1 dr ¡ 1
ap1O ð Þ 1 C wb C d ¡1 1
@~x1O  r¡1 w 1¡d 1¡a¡b
D h  r1¡ 1  r r 1¡ 1  2 wT
@p1O
dr
p1O dr ¡ 1 ir1¡ 1  p1O
a wd 1¡d
1
1¡a¡b C w
b C 1¡d

h i      1 (B:5)
rð1 ¡ dÞ

dr ¡ 2r C 1 d dr 1 dr ¡ 1 r ¡ 1
1 1
r¡1
ap ð Þ 1 C wr r ¡ 1
r¡1 1O w 1¡d 1¡a¡b b
C  
E
2 x1O
h
dr ir1¡ 1  r 1¡ 1  r r1¡ 1
dr ¡ 1
ap1O rðd ¡ 1Þ wd ð1 ¡1 dÞ 1
1¡a¡b C wb C 1 ¡1 d

The first term of (B.5) is always negative. The second term of (B.5) is zero if x1O
E
D 0, but if x1O
E
> 0, it
@~x1O 
is negative for 0 < r < 1 and positive for r < 0: Therefore, @p1O < 0 if x1O D 0, or if x1O > 0 and
E E

0 < r < 1. @~
@p1O cannot be signed if x1O > 0 and r < 0.
x1O E

To find @p@t1O , we differentiate (16) with respect to p1O and simplify:

@t  1
D ir1¡ 1   r r 1¡ 1 
@p1O h
dr
p rðd ¡ 1Þ r1¡ 1
addr ¡ r C 1 1 w
1O
1¡d
1
1¡a¡b C wb C 1d
8 h ir 1¡ 1 
r 1¡ 1  r r1¡ 1  9
>
> rð1 ¡ dÞ

dr ¡ 2r C 1 dr ¡ r C 1 1 dr 1 rðd ¡ 1Þ >
>
>
< r ¡ 1 ap1O ð1 ¡ d Þ 1 C b w >
 p1O E  1 E =
d w 1¡a¡b

h i 1  r1¡ 1  r r 1¡ 1  TC x C x1O :
>
>


dr ¡ r C 1 1 dr p1O rðd ¡ 1Þ r ¡ 1
w 1O w >
>
>
: ad 1 C bw Cd
1 >
;
w 1¡d 1¡a¡b

(B:6)

The first term of (B.6) is negative if 0 < r < 1 and positive if r < 0, regardless of x1O
E
. The second
@t 
term of (B.6) is equal to zero if x1O D 0 and positive if x1O > 0. Therefore, @p1O > 0 if r < 0, and
E E

@t  
@p1O < 0 if 0 < r < 1 and x1O
E
D 0. @p@t1O cannot be signed if 0 < r < 1 and x1O
E
> 0.

B. Investigating the effects of an increase in p1R on c , ~x1O  , and t 

By inspection of (19), (15), and (16), we see that the household’s optimal amounts of c, ~x1O , and t are
not functions of p1R . Therefore, @p@c1R D 0, @~@px1O1R D 0, and @p@t1R D 0, regardless of r and x1O
 E
.
 NEW ZEALAND ECONOMIC PAPERS 301

Appendix 3. Quotable value quarterly house price index

The quotable value quarterly house price index (HPI) is produced for detached houses only, or more
broadly as an ‘all residential’ index which includes flats and apartments. Both indices are constructed
using the following methodology:
1. For each territorial authority, all sale prices are summed, then divided by the sum of all cur-
rent capital values of these properties. This gives a ‘price-to-value’ ratio.
2. The price-to-value ratio is multiplied by the total capital values of all properties within a terri-
torial authority. This gives an estimated current total valuation for all properties within a ter-
ritorial authority.
3. The estimated current total valuation is divided by the number of properties within the terri-
torial authority. This gives the average current valuation for the territorial authority.
4. The percentage change between the current average valuation and the previous quarter’s aver-
age valuation is calculated. This gives the percentage change in the price index movement.
5. This percentage change is used to calculate the current period’s index.
Notes: The HPI is based on the December 2003 quarter. For Figure 1, the index for the whole of
New Zealand (i.e. all territorial authorities combined) was used. This is obtained in the same way as
described above, replacing ‘territorial authority’ with ‘New Zealand’.
Source: Quotable Value Limited.