Académique Documents
Professionnel Documents
Culture Documents
To cite this article: Xiangling Liu & Glenn Otto (2017) Housing supply elasticity in
local government areas of Sydney, Applied Economics, 49:53, 5441-5461, DOI:
10.1080/00036846.2017.1307936
ABSTRACT KEYWORDS
We report supply elasticity estimates of residential property (houses and apartments) for local Housing supply; supply
government areas (LGAs) in metropolitan Sydney. Using annual data for 1991–2012, the average elasticity; development
supply elasticity estimate across all LGAs is 0.2 for houses and 0.8 for apartments. The supply of application; undevelopable
houses is inelastic in all 43 LGAs; in contrast, apartment supply is elastic – greater than unity – in land
about one-third of LGAs. We develop theoretical and empirical models to explain the cross-
JEL CLASSIFICATION
section variation in supply elasticity across LGAs. For houses, supply elasticity is negatively related R31; R52
to an LGA’s population density, the time taken by a local council to process a development
application and to different measures of the amount of land in an LGA that is unavailable for new
housing development. In contrast to houses, variation in supply elasticity for apartments across
LGAs is unrelated to any of the available regressors.
and apartments in each of the 43 LGAs in metropo- available for housing development due to geographi-
litan Sydney. cal factors or planning regulations. The total area of
Our estimates indicate that the supply elasticity developable land in the city is ΛL. The city has a
for houses is relatively inelastic in most areas of total number of dwellings H, where each individual
Sydney. In no LGA is the house supply elasticity dwelling has a constant floor space of γ. The average
greater than one, and the median estimate is only housing density for the city ~S is given by
0.15. Estimates of the supply elasticity for apart-
ments are considerably larger – than for houses – ~S ¼ γH (1)
with about one-third of LGAs having elasticities ΛL
above one. The median estimate for apartments where we assume that there is one person per dwell-
is 0.65. ing. In contrast to Saiz, we allow average housing
To understand the factors that might influence density ~S to be greater than one due to multistory
the supply elasticity within a city, we adapt Saiz’s (or high-rise) buildings. We can use the above infor-
(2010) theoretical model to analyse housing supply mation to calculate the city radius as
and supply elasticity for LGAs. Housing supply elas-
ticity in an LGA is shown to depend upon its total ΛL~S ¼ γH (2)
land area, the fraction of developable land and the or using L ¼ πΦ2 as
stringency of planning regulations imposed by rffiffiffiffiffiffiffiffiffi
the LGA. γH
Φ¼ (3)
Based on our theoretical framework, we develop Λπ~S
an empirical model to explain the variation in hous-
The city is assumed to comprise a number of sub-
ing supply elasticity across LGAs. For houses, the
regions or LGAs. An LGA is assumed to be located d
cross-city variation in supply elasticity is found to be
kilometres from the centre (central business district
related to three variables: a measure of the share of
[CBD]) of the city and has total land area Ld and
land in an LGA that currently undeveloped, the
developable land area Λd Ld . We also assume that
average time required by a local council to decide
each LGA has a different level of regulatory stringency
on a development application (DA) and the popula-
in terms of allowing development on its developable
tion density of an LGA. In the case of apartments,
land, indicated by σ d . Thus, housing density in an
supply elasticity across LGAs is found to be uncor-
LGA is given by Sd ¼ ~Sð1 þ σ d Þ, where Sd 0 so
related with any of the potential explanatory vari-
that σ d 1. As σ d approaches −1, there is a high
ables for which we have data.
level of regulatory stringency so that housing density
The reminder of the article is organized as fol-
in the LGA is less than the city-wide average.
lows. Section II contains a model of intra-city varia-
The total number of houses in an LGA is given by
tion in housing supply. In Section III, we estimate
the supply elasticity for houses and apartments in Λd Ld Sd
Hd ¼ (4)
each LGA. Estimates of our cross-section model for γ
supply elasticity in LGAs are reported in Section IV.
Section V concludes. and using the city-level result (2) to substitute for γ
ΛL~SHd
H¼ (5)
II. A model of intra-city supply elasticity Λd Ld Sd
Saiz (2010) derives a city-level model of housing we get using Equation (3)
supply, which is used to analyse the determinants rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
LHd γ
of supply elasticities across US metropolitan areas. Φ¼ (6)
Sd Λd Ld π
In this section, we extend Saiz’s model of housing
supply-to-supply elasticity within a given city. We can derive a formula for the price of a house in
Following Siaz, we consider a circular city with an LGA that is d kilometres from the CBD using
radius Φ and total land area of L ¼ πΦ2 . Of the total Saiz’s approach. Let R0 be the rent per dwelling at
land area L, only some fraction 0 Λ 1 is the CBD and assume that the rent on a dwelling in
APPLIED ECONOMICS 5443
an LGA that is d kilometres from the CBD is This is a generalized version of Equation 1 by Saiz
given by (2010) and it is what he uses to compute the inverse
(average) supply elasticity for a particular city.
Rd ¼ R0 ðt dÞ (7) Equation (16) can be rearranged to obtain
where t is the per kilometre cost of commuting to 2
πΛd Ld Sd PdH CC ði=tÞ þ d
the CBD. In the steady state, the following arbitrage Hd ¼ (17)
condition links prices and rents in an LGA: Lγ
which allows us to derive the direct supply elasticity
R 0 ðt d Þ
PdH ¼ (8) in each LGA. Let the supply elasticity in an LGA that
i is d kilometres from the CBD be
where i is the real discount rate. ηSd ;@Hd =@PdH PdH =Hd , then using (16), we can
The housing market is assumed to be competitive, derive the housing supply elasticity for an LGA as
with developers buying land at market prices LCd sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
and selling competed houses at PdH , which includes 2i πΛd Ld Sd
ηd ¼
S
(18)
the cost of land plus the (constant city-level) con- t Hd Lγ
struction cost CC of the house:
It is apparent from Equation (18) that the supply
PdH ¼ CC þ LCd (9) elasticity of housing in an LGA depends on (1) the
total land area in the LGA Ld , (2) the faction of land
Using the above equations, we can write Λd available for development and (3) the housing
Rd ¼ iCC þ iLCd (10) density in an LGA Sd , which we assume can differ
from the city-wide average density ~S depending on
and if we normalize that the price of land at the the degree of regulatory stringency σ d imposed in an
city’s boundary to zero is zero, so LCΦ ¼ 0,
LGA, that is, Sd ¼ ~Sð1 þ σ d Þ. It is easy to show that
then RΦ ¼ iCC.
supply elasticity in an LGA will be positively related
Using the equation for the rent gradient, we can
to its total land area and its share of developable land
write
but negatively related to the degree of regulation
Rd ¼ R0 ðt dÞ (11) imposed on development by a local council.
In the following two sections of the article, we
while for an LGA at the city boundary first present estimates of supply elasticity for both
RΦ ¼ R0 ðt ΦÞ (12) houses and apartments across the LGAs in Sydney
city and then provide empirical evidence on the
Talking the difference between the above two equa- importance of the land availability, regulatory strin-
tions gives gency and the housing density in explain the distri-
bution of housing supply elasticities.
Rd RΦ ¼ tðd ΦÞ (13)
and using RΦ ¼ iCC, we obtain
III. Estimating supply elasticity
Rd ¼ iCC þ tðΦ dÞ (14)
The focus of this section of the article is on obtaining
Dividing both sides by i gives estimates of supply elasticity for houses and apart-
ments in each of Sydney’s 43 LGAs. Given the lim-
t
PdH ¼ CC þ ðΦ dÞ (15) ited number of annual observations for each LGA,
i we use a relatively simple model for the housing
Finally, substituting for Φ gives the supply curve for supply curve and obtain supply elasticity estimates
housing in an LGA that is d kilometres from using an instrumental variables (IV) procedure and
the CBD: the Autoregressive Distributive Lag (ARDL) bounds
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi estimator (Pesaran, Shin, and Smith 2001). The later
t Hd Lγ estimator is valid even if the data are non-stationary
Pd ¼ CC þ
H
d (16)
i πΛd Ld Sd due to stochastic trends.
5444 X. LIU AND G. OTTO
Statistics Profiles for New South Wales, 1992–2003, Table 1. Sydney local government areas.
and the Australian National Regional Profile for the 1. Ashfield 16. Holroyd 30. Penrith
2. Auburn 17. Hornsby 31. Pittwater
remaining years, 2004–2012). In effect, the housing 3. Bankstown 18. Hunters Hill 32. Randwick
4. Baulkham Hills 19. Hurstville 33. Rockdale
approval numbers are used to interpolate housing 5. Blacktown 20. Kogarah 34. Ryde
stocks for the inter-censual years.3 6. Blue Mountains 21. Ku-ring-gai 35. Strathfield
7. Botany 22. Lane Cove 36. Sutherland
Estimates of income for LGAs are based on data 8. Burwood 23. Leichhardt 37. Sydney
from the Australian Taxation Office’s Taxation 9. Camden 24. Liverpool 38. Warringah
10. Campbelltown 25. Manly 39. Waverley
Statistics. For the period 1990–91 to 2005–06, the 11. Canada Bay 26. Marrickville 40. Willoughby
Bureau of Infrastructure, Transport and Regional 12. Canterbury 27. Mosman 41. Woollahra
13. Fairfield 28. North Sydney 42. Wollondilly
Economics reports real income per taxpayer in 14. Gosford 29. Parramatta 43. Wyong
2007–08 prices by LGA (using figures derived from 15. Hawkesbury
350.00
300.00
250.00
F-statistic
200.00
150.00
100.00
50.00
0.00
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43
Houses Apartments
1.2
0.8
Elasticity
0.6
0.4
0.2
0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43
3
Elasticity
0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43
-1
Sydney with the lowest estimated supply elasticity for under one-third of LGAs are found to have supply
houses are Lane Cove (0.05), Mosman (0.05), Wollahra elasticities for apartments greater than unity. The largest
(0.05) and Kogarah (0.06). estimates in Figure 3 correspond to Baulkham Hills (4.3)
Figure 3 reports the estimated supply elasticity for and Bankstown (2.3). The average supply elasticity
apartments in each LGA. In all but three LGAs (Blue across all LGAs is 0.80 (with the median being 0.56).
Mountains, Camden and Liverpool), the estimated sup- On average, the supply elasticity for apartments is about
ply elasticity for apartments is larger than for houses. Just 3.5 times that for houses.5
5
This ratio is of similar magnitude to Gitelman and Otto’s (2012) finding using panel data for Sydney.
APPLIED ECONOMICS 5447
One anomalous result in the supply elasticity esti- We can test for the presence of a relationship in (log)
mates for apartments is the negative figure for the levels by testing the null hypothesis H0 : δQP ¼ δQQ ¼ 0
Blue Mountains. There appears to be evidence of a in Equation (22a), or H0 : δPP ¼ δPQ ¼ 0 in Equation
structural break in the supply relationship for apart- (22b), using an F-statistic. Since we do not know the
ments over the period 1991–2012. During the period order of integration of the two variables, this hypothesis
2003–2012, both the number and price of apart- test takes the form of a bounds test. The asymptotic
ments in the Blue Mountains declined. critical values for the ADRL bounds F-test are [4.94,
5.73] at the 5% level of significance, see Pesaran, Shin
ARDL bounds procedure and Smith (2001, Table CI[iii]). If the F-statistic is less
As a check on the reliability and robustness of the IV than 4.94 or greater than 5.73, we can make a decision
estimates for supply elasticity, we also employ the about the null hypothesis, without knowing if the series
ARDL bounds procedure due to Pesaran, Shin and are I(1) or I(0). Provided we can reject the null hypoth-
Smith (2001). The ARDL approach can be used to esis of no-levels relationship using either Equation (22a)
test for the presence of a relationship in levels or (22b), an estimate of the long-run supply elasticity βi
regardless of whether the variables in Equation (19) in Equation (19) can be computed using the relevant
are trend or first-difference stationary. Provided the ratio δQP =δQQ or δPP =δPQ . These ratios provide an alter-
null hypothesis of no-levels relationship is rejected, native set of estimates of the supply elasticity to the IV
we can also use the estimates from the ARDL model estimator.
to estimate βi . Figure 4 reports the F-statistic for houses for the
One issue that does arise with using the ARDL two specifications Equations (22a) and (22b), while
procedure is the need to assume that one of the Figure 5 reports the same test statistic for
variables in the supply curve can be treated as apartments.6 In the case of houses, there are only
being weakly exogenous. In using IV to estimate 10 LGAs where the hypothesis of no-levels relation-
Equation (19), we have treated ln Qit and ln Pti as ship cannot be rejected using both Equations (22a)
being simultaneously determined. However, for the and (22b). For the majority of LGAs, at least one of
ARDL analysis, we assume that it is valid to condi- the ARDL models allows for rejection of the null
tion on either ln Qit or ln Pti , and since it is not hypothesis of no relationship in levels between ln Qit
evident which is the more reasonable assumption, and ln Pti . However, in the case of apartments
we estimate both possible conditional error correc- (Figure 5), the evidence against the null of no-levels
tion models. Since there is only a relatively small relationship is considerably weaker than for houses.
sample of observations for each LGA, we use the There are 24 LGAs in which the hypothesis of no-
following parsimonious models: levels relationship is not rejected by both the ARDL
models.
Δ ln Qt ¼ δQ0 þ δQD D0204 þ δQP ln Pt1 For our analysis, the primary question is whether –
þ δQQ ln Qt1 þ θQP0 Δ ln Pt when the bounds test points to evidence of a levels
þ εQt (22a) relationship – the ARDL model yields an estimated
elasticity that is similar in magnitude to the IV esti-
Δ ln Pt ¼ δP0 þ δPD D0204 þ δPP ln Pt1
mate. Figures 6 and 7 present a scatter plot of the IV
þ δPQ ln Qt1 þ θPQ0 Δ ln Qt
estimate and an ARDL estimate for those LGAs where
þ εPt (22b) the hypothesis of no-levels relationship is rejected or
Finally, given the sharp increase in residential prop- where the bounds test is inconclusive. It is apparent
erty prices around 2002 – which is not well that there is a reasonably strong positive correlation
explained by the standard ARDL regressors – we between the supply elasticity estimates obtained from
augment these specifications with a dummy variable the two estimators. On balance, the ARDL estimates –
D0204 that takes the value 1 for the years 2002, 2003 particularly for houses – are broadly consistent with
and 2004 and zero elsewhere. those obtained from the IV estimator; and this sug-
6
A full set of estimates are reported in Tables A2 and A3 in the Appendix.
5448 X. LIU AND G. OTTO
20.0
15.0
F-stat
10.0
5.0
0.0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43
Q_dep P_dep
14.0
12.0
10.0
8.0
F-stat
6.0
4.0
2.0
0.0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43
Q_dep P_dep
1.2
1.0
0.8
IV Estimate
0.6
0.4
y = 0.7603x + 0.0155
0.2
R² = 0.8567
0.0
-0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2
-0.2
ARDL Estimate
5.0
4.5
4.0
3.5
IV Estimate
3.0
2.5
2.0
1.5 y = 0.7743x + 0.0629
1.0 R² = 0.8977
0.5
0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0
ARDL Estimate
gests that the IV estimates are robust to possible number of housing units in an LGA Hd . We can re-
stochastic non-stationarity in the data. write Equation (18) as
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2i πΛd ~Sð1 þ σ d Þ
IV. A cross-section model of supply elasticity ηSd ¼ (24)
t ðHd =Ld ÞLγ
In this section of the analysis, we develop an empiri-
and given the available data, we proxy these deter-
cal model to explain the cross-section distribution of
minants of supply elasticity as follows. We measure
the supply elasticity estimates across LGAs. Figure 8
Hd =Ld by the population density of an LGA. The
shows estimated density functions (using an
average time taken by a local council to decide on a
Epanechnikov kernel) for the IV estimates of the
DA is used as an indicator of regulatory stringency
supply elasticities for houses and apartments.
σ d . We use four geographic variables as measures of
the proportion of land in an LGA that is unavailable
Modelling framework for residential development ð1 Λd Þ. Finally, we
also include a measure of the distance of an LGA
The cross-section model has the following general from the CBD of Sydney.7 The cross-section regres-
form: sions use the supply elasticity estimates obtained by
^ ¼ c0 þ X0 θi þ i
β (23) IV as the dependent variable and separate models
i i
are estimated for houses and for apartments.
where β^ is an estimate of supply elasticity in the ith
i
LGA and Xi is a vector of cross-section explanatory
variables. Similar models are estimated by Green,
Data
Malpezzi and Mayo (2005) and Saiz (2010) for
major US cities. These studies point to cross-city The distance of an LGA from the Sydney CBD is
differences in regulatory constraints and in natural measured as the average of the distances to the CBD
geography as being the primary sources of regional of the suburbs that make-up the LGA.8 The popula-
variations in housing supply elasticity. tion density of an LGA equals the ratio of its resident
Equation (18) indicates the factors that affect population to its area and is measured as the number
supply elasticity for an LGA within a given city. of people per square kilometre. Median and mean
The four variables that vary across LGAs are total times required by Local Councils to decide on a DA
land area of the LGA Ld , the faction of land Λd are reported in Comparative Information on NSW
available for development, the degree of regulatory Local Government and are available from 1994–95 to
stringency σ d imposed by a Local Council and the 2011–12.
0
-1 0 1 2 3 4 5
-1 Supply Elasticity
Apartments Houses
7
Gitelman and Otto (2012) find that supply elasticity for houses increases for clusters of LGAs that are further away from the CBD.
8
The distance from a suburb to the CBD is measured from the most significant point in a suburb, such as the largest intersection in the centre of a suburb.
Distance is calculated as great circle distance, which is the shortest distance between two points on a sphere, measured along the surface of the sphere.
5450 X. LIU AND G. OTTO
We construct four geographic variables (called Our third geographic variable (Geo3) measures
Geo1, Geo2, Geo3 and Geo4) that are designed to the proportion of land within each LGA that can
measure proportion of land in an LGA that is poten- currently be characterized as built-up area. The forth
tially unavailable for residential development. This is geographic variable (Geo4) measures unavailable
done by seeking to identify those areas of land within land in an LGA as the aggregate of undevelopable,
the total area of an LGA that are likely to be unavail- reserved and built-up areas (i.e. Geo2 + Geo3).
able (or relatively costly) to use for new housing.9
The initial indicator of unavailable land (Geo1) is
Scatter plots
based on the approach of Saiz (2010) who considers
US metropolitan statistical areas (MSAs) with popu- To examine the possible relationships in the data, we
lations of at least half a million. Focusing on a circle begin our analysis by examining scatter plots between
with a 50-km radius from the centre of the MSA, supply elasticity and the explanatory variables.
Saiz estimates the area of undevelopable land, which
he defines as the area corresponding to steep slopes Distance to CBD
(>50%), oceans, lakes, wetlands and other water fea- Figure 9 plots the estimated supply elasticity for
tures. He views this measure of undevelopable land houses for an LGA against the distance of the LGA
as reflecting original constraints on development from the CBD. It is evident that there is a positive
and as being a fundamentally exogenous variable. relationship between the supply elasticity for houses
Because we are using data for a single metropolitan and the distance from the CBD.10 Supply elasticity
area, Saiz’s approach does not map perfectly into our for houses increases with distance from the CBD.
analysis. For most LGAs in Sydney, a relatively low For each 10 km increase in distance from the CBD,
proportion of land is undevelopable according to the the supply elasticity for houses increases by 0.08.
measure proposed by Saiz. One notable feature of Figure 9 is the three LGAs
A more inclusive – but possibly less exogenous – for which the estimated supply elasticity is well
measure of unavailable land can be obtained by above what is predicted by their respective distances
adding to the Saiz measure, land within an LGA from the CBD – Baulkham Hills, Liverpool and
that is currently used for airports, cemeteries, Camden.
reserves, recreation areas and prohibited areas. This Figure 10 shows the supply elasticity for apart-
yields a second variable Geo2. ments against distance from the CBD. What is
1.2
y = 0.0076x + 0.0589
1 R² = 0.48
Camden
0.8 Baulkham Hills
Elasticty
Liverpool
0.6
0.4
0.2
0
0 10 20 30 40 50 60 70 80 90 100
Distance to CBD (kms)
9
The area of an LGA is based on its current boundaries. Where boundaries have changed in the period 1991–2012, we have tried to splice the data to obtain
a consistent set of data for each LGA. The LGA with the largest area is Hawkesbury (2783.2 sq. km) and the smallest is Hunters Hill (5.6 sq. km). The
estimated supply elasticities for houses and apartments in Hawkesbury are 0.32 and 0.33, respectively, while the comparable estimates for Hunters Hill are
0.19 and 0.57.
10
Since population density declines with distance from the CBD, the supply elasticity for houses shows a negative relationship with population density. We
have not presented a scatter plot for this relationship.
APPLIED ECONOMICS 5451
5
y = -0.0006x + 0.8151
Baulkham Hills
R² = 0.0002
4
Elasticity
2
0
0 20 40 60 80 100
-1 Distance to CBD (kms)
apparent is that – unlike houses – the supply elasti- that supply elasticity to be negatively related with the
city for apartments is unrelated to the distance of an time required in processing a DA. Figures 12 and 13
LGA from the CBD. plot supply elasticity for houses and apartments
(respectively) against the mean number of days
Processing time for a DA taken to process a DA. The scatter plots do point
Residential construction within LGAs is partially to a negative relationship for both houses and apart-
regulated though a DA process. As illustrated by ments – higher supply elasticity is associated with
Figure 11, the time taken by a local council to lower mean processing times – although the
process (i.e. to make an initial decision on) a DA correlation is much stronger for houses than for
varies considerably across LGAs. We might expect apartments.
16
14
12
Frequency
10
8
6
4
2
0
45 60 75 90 105 More
No. of Days
1.2
y = -0.539ln(x) + 2.5073
R² = 0.4208
1
0.8
Supply Elasticity
0.6
0.4
0.2
0
0 20 40 60 80 100 120
-0.2 Mean No. Days for DA
5 y = -0.432ln(x) + 2.6354
R² = 0.0168
4
Supply Elasticity
3
0
0 20 40 60 80 100 120
-1 Mean No. of Days for DA
Geographic factors low (less than 10%) and the scatter plot indicates a
One factor that is expected to influence supply elas- weak positive (rather than the expected negative)
ticity is the amount of land available for housing relationship. An attractive feature of Saiz’s measure
development within a particular LGA. Figure 9 of undevelopable land is its exogeneity, in that steep
already suggests this is the case, since the supply slopes and natural water features are likely to be
elasticity for houses increases with distance from fundamental influences on the cost of development.
the CBD and it is generally the case that the size The remaining three measures of undevelopable
(area) of LGAs increases with distance from the land all include land, which, while it may not be
CBD. However, Figure 10 indicates that land avail- currently available for housing development, is not
ability is more likely to affect the supply elasticity for undevelopable due to its physical characteristics, but
houses rather than for apartments. Apartments use because it is currently being used for some other
less land per square metre of housing space and new purpose (possibly including housing). Geo2 adds
apartments can be supplied through urban redeve- reserved land to Saiz’s measure. However, despite
lopment (i.e. replacing existing houses [or old apart- its more comprehensive status, Figure 15 indicates
ments] by new one or by replacing industrial areas that there is also a positive correlation between sup-
with new apartments). In inner-city LGAs, it is likely ply elasticity and Geo2.
to be relatively costly to significantly increase the Figure 16 plots supply elasticity for houses against
supply of new houses, due to a lack of undeveloped the share of built-up areas within each LGA (Geo3).
land available for new construction. In this case, we do see a negative relationship, with
Figure 14 presents a scatter plot of supply elasti- lower supply elasticity in LGAs which are relatively
city for houses against Saiz’s measure of undevelop- more built-up. This negative correlation is also evi-
able land in each LGA. Note that the share of dent in Figure 17 where potentially unavailable land
undevelopable land in Sydney’s LGAs is relatively is measured by the sum (Geo2 + Geo3).
0.8
0.6
0.4
0.2
0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Share of Undevelopable Land (Geo1)
Figure 14. Supply elasticity for houses and Saiz’s measure of undevelopable land.
APPLIED ECONOMICS 5453
Supply Elasticity
0.8
0.6
0.4
0.2
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
Reserved and Undevelopable Land (Geo2)
Figure 15. Supply elasticity for houses and reserved and undevelopable land.
0.8
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1 1.2
Built-up Area as a Share of Total Land (Geo3)
0.8
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1 1.2
Share of Land Potentially Unavailable for New Housing (Geo4)
Figure 17. Supply elasticity for houses and potentially unavailable land.
Determinants of supply elasticity the specifications, the three variables that are most
consistently correlated with estimated supply elasti-
Estimates of the cross-section model for supply elas-
city for houses, in an LGA, are the mean-time taken
ticity are reported in Tables 2 and 3, for houses and
to approve a DA, population density and the quan-
apartments, respectively. The results in Table 2 indi-
tity of undevelopable land. The distance of an LGA
cate that around 60–80% of the variation in the
from the CBD is not found to be statistically signifi-
supply elasticity of houses across Sydney LGAs is
cant influence on supply elasticity, when it is
explained by the available regressors. Across all of
included with the other regressors.
5454 X. LIU AND G. OTTO
Table 2. Models of supply elasticity for houses. In model (iii), we use the proportion of an LGA
Dependent variable: supply elasticity for houses that is classified as build-up, as a proxy for land that
(i) (ii) (iii) (iv) (v)
is difficult or costly to use for houses. The share of
Geo1 −3.2682 – – – –
(1.6824) build-up land has a significantly negative effect on
Geo2 – −0.5592 – – – the supply elasticity for houses, although including
(0.1893)
Geo3 – – −0.3378 – – built-up area in the model results in population
(0.1303) density becoming insignificant. Model (iv) uses our
Geo4 – – – −0.6324 −0.6681
(0.1192) (0.1249) broadest measure of undevelopable land (Geo4) and
CBD distance 0.0179 0.0042 0.0329 −0.0195 −0.0340
(0.0356) (0.0272) (0.0285) (0.0181) (0.0256)
implies that the supply elasticity for houses is largest
Pop density −0.0892 −0.1351 0.0194 −0.0293 −0.0336 in LGAs with low population densities, fast proces-
(0.0388) (0.0389) (0.0283) (0.0131) (0.0135)
Mean DA −0.3179 −0.1785 −0.3302 −0.1252 – sing times for DAs and relatively large areas of
(0.0967) (0.0647) (0.1048) (0.0391) potentially developable land. Model (v) is similar to
Med DA – – – – −0.0712
(0.0486) (iv) except that mean DA processing time is replaced
Constant 2.2096 2.0632 1.5842 1.5197 1.3691 by the median.
(0.6076) (0.4210) (0.4592) (0.2351) (0.3074)
R2 0.608 0.677 0.619 0.809 0.798 We can visualize the fit of the cross-section model
The numbers in parentheses are heteroscedasticity-robust SEs (White for houses by comparing the fitted values for model
1980). Significant coefficients at the 5% level are highlighted in bold face. (iv) with the estimated supply elasticities, see
Figure 18. The overall cross-section variation in the
Table 3. Models of supply elasticity for apartments. supply elasticity estimates is well explained by the
Dependent variable: supply elasticity for apartments variables in the model. The two LGAs with the
(i) (ii) (iii) (iv) (v) largest (absolute) residuals are Camden where the
Geo1 −10.3299 – – – –
(10.5649)
model under-predicts the supply elasticity (0.74 vs.
Geo2 – −1.8742 – – – 0.96) and Penrith where the model over-predicts
(1.7832)
Geo3 – – −0.0517 – – (0.63 vs. 0.36). Table 4 presents a comparison of
(0.9353) the supply elasticity estimates for Camden and
Geo4 – – – −1.1886 −1.1384
(1.6360) (1.4999) Penrith, along with their main characteristics.
CBD distance 0.0210 −0.0298 0.0981 −0.0174 −0.0407 These two LGAs have broadly similar characteristics,
(0.2290) (0.2839) (0.2189) (0.2996) (0.3499)
Pop density −0.0753 −0.2379 0.0787 0.0958 0.0954 although Camden has a lower population density
(0.2123) (0.3410) (0.1562) (0.1073) (0.1093) than Penrith (8 vs. 12 people per square kilometre)
Mean DA −0.3630 0.1097 −0.4521 −0.0379 –
(0.4572) (0.7263) (0.4723) (0.6771) and also a somewhat larger proportion of develop-
Med DA – – – – −0.0036
(0.6228)
able land (87 vs. 74%). What is interesting is that
Constant 2.9835 2.5546 1.9173 1.2734 1.1449 while Camden has the larger supply elasticity for
(3.2747) (3.5942) (2.5189) (2.9356) (3.4544)
R2 −0.049 0.003 −0.080 −0.025 −0.025 houses (0.96 vs. 0.36), Penrith has the larger supply
The numbers in parentheses are heteroscedasticity-robust SEs (White 1980). elasticity for apartments (0.53 vs. 0.18). In fact,
Camden has one of the lowest apartments supply
elasticity estimates across all of Sydney’s LGAs.
Model (i) includes Saiz’s measure of undevelop- Turing to the elasticity of supply for apartments,
able land, and while this variable does have a nega- it is evident from Table 3 that none of the potential
tive effect on supply elasticity, the effect is not explanatory variables have any significant relation-
estimated very precisely. Nevertheless, it is interest- ship with elasticity. Variations in supply elasticity for
ing that geographic features such as natural water- apartments across LGAs in Sydney appear to be
bodies and land gradient do have some influence on essentially random. Compared with separate houses,
the supply elasticity for houses across Sydney. In apartments embody a much lower ratio of land to
model (ii), use of a broader measure of undevelop- capital; so, this might explain why supply elasticity
able land (which includes certain types of reserved has no clear association with land availability.
land) yields a negative and highly significant coeffi- Finally, we examine whether there is a correlation
cient estimate. In effect, the higher the proportion of between the long-run price of housing in an LGA and
undevelopable or reserved land in an LGA, the lower the estimated supply elasticity. For houses, there is a
is the supply elasticity for houses. strong correlation between the estimated supply
APPLIED ECONOMICS 5455
1.20
1.00
0.80
Elasticity
0.60
0.40
0.20
0.00
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43
8
ln(Average Real House Price)
7
6
5
4
3
y = 0.0643x + 5.3582
2
R² = 0.4801
1
0
0 5 10 15 20 25
Inverse Supply Elasticity
elasticity and the average real house price in an LGA. elasticity of apartments is larger than for houses. On
Figure 19 plots the log of average real house prices (for average, a uniform 5% increase in real residential
the period 1991–2012) against the inverse of the esti- property prices in Sydney would – other things equal
mated supply elasticity for each LGA. In general, LGAs – increase the stock of houses by 1% and the stock of
with more inelastic supply responses have higher apartments by 4%.
priced houses. No such relationship holds for We develop a theoretical model of intra-city
apartments. housing supply and supply elasticity as a function
of housing density, physical land availability and
regulated land availability. Results from the theore-
V. Conclusion tical model are used to guide the development of an
empirical model of the variation in supply elasticity
This article presents estimates of the supply elasticity across LGAs. In modelling the variation in supply
for houses and apartments across various regions elasticity across LGAs, we find no useful explanatory
(LGAs) of Australia’s largest city, Sydney. Using data variables for differences in the estimated supply
for the period 1991–2012, we generally obtain sensible elasticity for apartments. In the case of houses, we
(non-negative) estimates for supply elasticities. For the find that variation in supply elasticity at the LGA-
majority of LGAs, we find that the estimated supply level is associated with population densities,
5456 X. LIU AND G. OTTO
processing speed for development approvals and the Supply of Housing and Their Sources.” American
availability of developable land. Economic Review 95 (2): 334–339. doi:10.1257/
000282805774670077.
Kennedy, S. 2010, “Housing Supply and Affordability.”
Disclosure statement Speech to the Council of Capital City LordMayors’
Towards a National Urban Policy Summit, Canberra,
No potential conflict of interest was reported by the authors. May 27.
Newey, W., and K. D. West. 1987. “A Simple, Positive Semi-
Definite, Heteroskedasticity and Autocorrelation-
Funding
Consistent Covariance Matrix.” Econometrica 55 (3):
Otto wishes to acknowledge financial support from the 703–708. doi:10.2307/1913610.
Australian Research Council (LG 0884095). Pesaran, M. H., Y. Shin, and R. J. Smith. 2001. “Bounds
Testing Approaches to the Analysis of Levels
Relationships.” Journal of Applied Econometrics 16 (3):
References 289–326. doi:10.1002/jae.616.
Richards, T. 2009, “Housing Market Developments.” Talk to
Ball, M., G. Meen, and C. Nygaard. 2010. “Housing CEDA Housing Forum: A National Round-up, Sydney,
Supply Price Elasticities Revisited: Evidence from September 29.
International, National, Local and Company Data.” Saiz, A. 2010. “The Geographic Determinants of Housing
Journal of Housing Economics 19: 255–268. doi:10.1016/j. Supply.” Quarterly Journal of Economics 125 (3): 1253–
jhe.2010.09.004. 1296. doi:10.1162/qjec.2010.125.3.1253.
Caldera, A. C., and Å. Johansson. 2013. “The Price Stock, J. H., and M. Yogo. 2002, “Testing for Weak
Responsiveness of Housing Supply in OECD Countries.” Instruments in Linear IV Regression.” NBERTechnical
Journal of Housing Economics 22: 231–249. doi:10.1016/j. Working Paper 284. Cambridge, MA: National Bureau of
jhe.2013.05.002. Economic Research.
Gitelman, E., and G. Otto. 2012. “Supply Elasticity Estimates The Economist. 2013. “Mixed Messages: America Surges,
for the Sydney Housing Market.” Australian Economic Much of Europe Sinks.” The Economist, August 31.
Review 45 (2): 176–190. doi:10.1111/j.1467- http://www.economist.com/news/finance-andeconomics/
8462.2012.00679.x. 21584361-americasurges-much-europe-sinks-mixed-
Glaeser, E. L., J. Gyourko, and R. E. Saks. 2006. “Urban messages.
Growth and Housing Supply.” Journal of Economic White, H. 1980. “A Heteroskedasticity-Consistent
Geography 6 (1): 71–89. doi:10.1093/jeg/lbi003. Covariance Matrix Estimator and a Direct Test for
Green, R. K., S. Malpezzi, and S. K. Mayo. 2005. Heteroskedasticity.” Econometrica 48: 817–838.
“Metropolitan-Specific Estimates of the Price Elasticity of doi:10.2307/1912934.
APPLIED ECONOMICS 5457
Table A1. Supply elasticity estimates and SEs – instrumental Table A2. ARDL bounds test and long-run elasticity – houses.
variables. Model (22a) Model (22b) ARDL
Houses Apartments LGA F-stat. ^δQP =^δQQ F-stat. ^δPP =^δPQ LR elasticity
LGA Elasticity SE Elasticity SE Ashfield 1.17 −0.16 1.13 0.00 na
Ashfield 0.09 0.015 0.16 0.031 Auburn 9.02 0.26 8.51 0.26 0.26
Auburn 0.23 0.032 1.02 0.151 Bankstown 6.25 0.17 10.92 0.18 0.18
Bankstown 0.15 0.017 2.31 0.252 Baulkham Hills 6.19 0.76 4.70 0.71 0.76
Baulkham Hills 0.71 0.049 4.34 0.367 Blacktown 10.42 0.49 7.17 0.32 0.49
Blacktown 0.45 0.063 1.17 0.268 Blue Mountains 6.26 0.32 8.17 0.11 0.11
Blue Mountains 0.29 0.038 −0.57 0.247 Botany 5.39 0.16 12.13 0.15 0.15
Botany 0.15 0.008 0.34 0.075 Burwood 6.84 0.15 8.41 0.15 0.15
Burwood 0.09 0.019 0.56 0.037 Camden 9.48 1.06 5.69 1.10 1.06
Camden 0.96 0.079 0.18 0.208 Campbelltown 7.45 0.38 2.14 0.16 0.38
Campbelltown 0.26 0.031 0.51 0.272 Canada Bay 5.02 0.19 14.18 0.18 0.18
Canada Bay 0.14 0.016 0.55 0.091 Canterbury 6.23 0.16 10.57 0.15 0.15
Canterbury 0.09 0.016 0.20 0.027 Fairfield 8.33 0.21 8.86 0.02 0.02
Fairfield 0.19 0.026 0.24 0.029 Gosford 4.81 0.51 3.54 0.20 na
Gosford 0.36 0.040 1.11 0.317 Hawkesbury 8.79 0.36 8.68 0.22 0.36
Hawkesbury 0.32 0.044 0.33 0.076 Holroyd 4.35 0.25 5.43 0.42 0.42
Holroyd 0.20 0.039 1.60 0.211 Hornsby 8.14 0.29 7.02 0.26 0.29
Hornsby 0.28 0.033 1.50 0.136 Hunters Hill 10.35 0.19 8.61 0.19 0.19
Hunters Hill 0.19 0.016 0.57 0.089 Hurstville 4.27 0.15 4.56 0.14 na
Hurstville 0.12 0.013 1.13 0.061 Kogarah 7.21 0.10 11.68 0.12 0.12
Kogarah 0.06 0.013 0.97 0.071 Ku-ring-gai 2.10 0.10 1.76 0.18 na
Ku-ring-gai 0.09 0.007 1.21 0.308 Lane Cove 1.99 0.13 9.35 0.11 0.11
Lane Cove 0.05 0.010 0.34 0.041 Leichhardt 1.97 0.18 8.01 0.16 0.16
Leichhardt 0.13 0.011 0.66 0.056 Liverpool 19.04 0.95 15.02 1.08 0.95
Liverpool 0.66 0.052 0.35 0.106 Manly 3.01 −2.67 12.74 0.43 0.43
Manly 0.12 0.021 0.15 0.037 Marrickville 5.11 0.17 7.86 0.20 0.20
Marrickville 0.06 0.016 0.25 0.031 Mosman 10.23 0.11 6.90 0.11 0.11
Mosman 0.05 0.010 0.27 0.022 North Sydney 3.18 0.25 2.94 0.27 na
North Sydney 0.19 0.018 0.47 0.043 Parramatta 5.24 0.16 9.34 0.20 0.20
Parramatta 0.15 0.017 1.01 0.164 Penrith 7.36 0.40 13.03 0.11 0.40
Penrith 0.36 0.053 0.53 0.106 Pittwater 1.99 0.45 2.11 0.36 na
Pittwater 0.20 0.030 0.30 0.074 Randwick 3.40 0.15 4.76 0.13 na
Randwick 0.11 0.012 0.33 0.041 Rockdale 2.92 0.14 7.72 0.13 0.13
Rockdale 0.08 0.015 0.71 0.065 Ryde 8.93 0.12 6.88 0.12 0.12
Ryde 0.10 0.006 0.48 0.063 Strathfield 3.28 0.12 3.36 0.14 na
Strathfield 0.07 0.022 2.00 0.456 Sutherland 7.33 0.23 4.71 0.19 0.23
Sutherland 0.20 0.017 0.96 0.077 Sydney 3.21 0.32 3.43 0.41 na
Sydney 0.18 0.042 1.70 0.129 Warringah 7.85 0.13 4.41 0.12 0.13
Warringah 0.12 0.006 0.57 0.079 Waverley 1.69 −1.67 2.51 0.21 na
Waverley 0.08 0.019 0.23 0.045 Willoughby 7.32 0.09 7.52 0.09 0.09
Willoughby 0.08 0.003 1.75 0.114 Woollahra 11.11 0.56 8.59 0.41 0.56
Woollahra 0.50 0.052 0.85 0.316 Wollondilly 2.72 0.02 5.34 −0.04 −0.04
Wollondilly 0.05 0.020 0.25 0.044 Wyong 6.16 0.80 1.08 0.34 0.80
Wyong 0.55 0.070 0.88 0.277 Critical values at a 5% level of significance for the bounds test are [4.94,
Instruments used are real per-capita income, population and the real 5.73]. F-statistics below 4.94 imply that we cannot reject the null hypoth-
interest rate. SEs are robust to serial correlation (lags = 5) and hetero- esis of no-levels relationship between price and quantity. ARDL elasticity
scedasticity (Newey and West 1987). estimate corresponds to model with the maximum F-statistic.
5458 X. LIU AND G. OTTO
Table A3. ARDL bounds test and long-run elasticity – Dwelling stocks
apartments.
Model (22a) Model (22b) ARDL The quantity of residential property is measured by
LGA F-stat. ^δQP =^δQQ F-stat. ^δPP =^δPQ LR elasticity the number of private residential properties in each
Ashfield 5.05 0.18 4.71 0.18 0.18 of the 43 LGAs in metropolitan Sydney. For each
Auburn 5.48 1.14 4.29 1.09 1.14
Bankstown 5.36 2.68 3.68 2.26 2.68
LGA, we construct an annual series for the stock of
Baulkham Hills 9.48 5.03 10.46 4.94 4.94 residential properties at end June from 1991 to 2012.
Blacktown 3.68 1.60 4.03 1.50 na
Blue Mountains 1.74 −2.27 0.23 −1.66 na For the years 1991, 1996, 2001, 2006 and 2011, data
Botany 2.22 0.39 2.08 0.33 na from the relevant Census are used. To interpolate
Burwood 3.37 0.59 4.91 0.59 na
Camden 2.70 −1.23 1.64 10.87 na estimates for the inter-censual years, we use data on
Campbelltown 3.14 0.76 2.53 0.76 na the number of housing permits. Source: Regional
Canada Bay 2.26 2.20 9.10 1.84 1.84
Canterbury 8.51 0.22 7.69 0.23 0.22 Statistics Profiles for New South Wales, 1992–2003
Fairfield 1.64 0.25 3.86 0.25 na and Australian National Regional Profile,
Gosford 2.94 2.07 3.33 1.17 na
Hawkesbury 1.54 −0.18 1.06 1.18 na 2004–2012.
Holroyd 10.29 1.72 7.74 1.67 1.72
Hornsby 4.43 1.76 3.30 1.58 na
In the Census, private residential dwellings are
Hunters Hill 3.49 0.77 3.88 0.68 na classified into three main categories:
Hurstville 8.59 1.20 6.16 1.20 1.20
Kogarah 1.25 1.10 5.12 1.05 0.97
Ku-ring-gai 2.32 2.05 2.76 2.44 na (1) separate house;
Lane Cove 8.46 0.35 8.67 0.35 0.35
Leichhardt 2.00 1.20 11.75 0.72 0.72 (2) semi-detached, row or terrace house or town-
Liverpool 2.43 0.94 1.67 1.10 na house and
Manly 2.62 0.39 2.49 0.39 na
Marrickville 3.20 0.31 3.53 0.34 na (3) flat, unit or apartment.
Mosman 6.48 0.31 8.48 0.29 0.29
North Sydney 4.15 0.55 4.66 0.54 na
Parramatta 4.12 1.34 4.20 1.42 na In the Regional Profile data, there are two classi-
Penrith 4.28 1.02 4.37 1.17 na
Pittwater 3.43 0.48 3.16 0.50 na
fications for permits data:
Randwick 5.52 0.39 5.81 0.39 0.39
Rockdale 3.33 0.77 8.20 0.76 0.76
Ryde 7.73 0.47 8.96 0.43 0.43 (a) a house is defined as a detached or a semi-
Strathfield 7.91 3.49 5.00 3.52 3.49 detached building, consisting of one residen-
Sutherland 9.11 1.09 5.37 0.99 1.09
Sydney 3.60 1.97 5.10 1.83 1.83 tial unit and primarily used for long-term
Warringah 1.20 1.67 3.35 1.01 0.74 residential purposes and
Waverley 2.70 0.41 3.76 0.52 na
Willoughby 4.78 1.93 4.97 1.88 1.88 (b) a dwelling unit is defined as a self-contained
Woollahra 1.30 1.09 2.09 0.77 na suite of rooms, including cooking and bathing
Wollondilly 1.79 0.32 4.43 0.30 na
Wyong 4.19 −3.02 4.10 −1.56 na facilities and intended for long-term residential
Critical values at a 5% level of significance for the bounds test are [4.94, use.11
5.73]. F-statistics below 4.94 imply that we cannot reject the null hypoth-
esis of no-levels relationship between price and quantity. ARDL elasticity
estimate corresponds to model with the maximum F-statistic. To ensure consistency between the two data sources,
housing stock data in categories (1) and (2) of the
Data Appendix Census are combined and are counted as houses,
while dwelling units belonging to (3) or (b) are
counted as apartments.
Property prices
There are some potential difficulties with use of the
Median prices for LGAs are sourced from the quar- permits data to interpolate the Census data on housing
terly Rent and Sales Reports published by Housing stocks. Not all housing permits that are issued are
NSW. http://www.housing.nsw.gov.au/About+Us/ necessarily converted into completions and result in
Reports+Plans+and+Papers/Rent+and+Sales the construction of a new property. In some cases, an
+Reports/ existing property is demolished and replaced with one
11
Regardless of whether they are self-contained or not, units within buildings offering institutional care (e.g. hospital) or temporary accommodation (e.g.
motels, hostels and holiday apartments) are not defined as dwelling units. Such units are included in non-residential building approvals.
APPLIED ECONOMICS 5459
12000
y = 1.1102x + 91.547
10000
6000
y = 1.1319x + 154.54
4000
2000
0
0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
-2000 Aggregate approvals
(or more) new residences. Also, housing approvals will linked (over the 5-year intervals), then all points
include approvals for renovations to existing residen- should lie on the 45 degree line. For houses, we see
tial buildings (and conversions of a non-residential that the approvals data tend to overestimate the actual
building to a residential building.) Finally, there is a increase in housing stock as measured by the census.
timing issue in that a housing approval recorded in In contrast for the apartments, data approvals tend to
1 year may not result in the completion of a new underestimate the actual increase.
residence until a later year. Despite the apparent limitations with the approvals
We can gauge an indication of the extent to which data, in the absence of anything better, we use the
the approvals figures are a good indication of longer annual approvals data to interpolate the Census fig-
term new constructions by comparing the change in ures. To illustrate our methodology, the following
the stock of houses (apartments) between two cen- table, Table A4, shows the calculation of the annual
suses (say 1996 and 1991) with the sum of approvals stock of houses in the Baulkham Hills LGA.
between the two censuses. We summarize these
results in Figures A1 and A2.
Population
The x-axis in Figure A1 measures the actual change
in the number of houses between censuses in a LGA, The following sources provide data for Estimated
while the y-axis measures the sum of building Resident Population for LGAs on end of financial
approvals for houses in an LGA over an equivalent year basis (end-June) for the period 1995–96 to
period. Figure A2 reports the same data for apart- 2011–12. Source: Regional Population Growth,
ments. If approvals and outcomes were perfectly Australia, 1996–2006, 2008–09, 2011 (cat. no. 3218.0).
20000
18000 y = 0.7697x - 261.23
16000
14000
Change in census
Table A4. Interpolated stock of houses for Baulkham Hills. the growth rate of real income per taxpayer in both
Sum Error 2010–11 and 2011–12 is equal to the growth rate for
Year Census Change Approvals App stock Adj Annual
1990–91 34,028 na na na 34,028
2009–10.
1991–92 559 76.4 34,663
1992–93 512 76.4 35,252
1993–94 597 76.4 35,925 Real interest rate
1994–95 681 76.4 36,683
1995–96 37,292 3264 533 2882 382 76.4 37,292
1996–97 909 −89.6 38,111 Data on the real interest rate are obtained from the
1997–98 1641 −98.6 39,663 RBA spreadsheet Capital Market Yields (F2). It is the
1998–99 1775 −98.6 41,348
1999–00 1932 −98.6 43,191 yield on the Australian Government inflation-
2000–01 44,541 7249 1440 7697 −448 −98.6 44,541 indexed bond with the longest maturity.
2001–02 2127 275 46,943
2002–03 1481 275 48,699
2003–04 948 275 49,922
2004–05 632 275 50,829 Consumer price index
2005–06 51,556 7015 452 5640 1375 275 51,556
2006–07 465 201.4 52,222
2007–08 411 201.4 52,835 The consumer price index for Sydney is obtained
2008–09 298 201.4 53,334 from ABS release 6401.0 – Consumer Price Index,
2009–10 350 201.4 53,886
2010–11 54,574 3018 487 2011 1007 201.4 54,574 Australia.
2011–12 363 54,937
Inter-census stocks of houses are calculated by taking the previous census
figure and adding approvals plus an adjustment equal to one-fifth of the Development approvals
difference between the change in the house stock between censuses and
the comparable sum of approvals. For example, the figure for 1996–97 is
38,111 = 37,292 + 909 + (7249 − 7697)/5. Median and mean times required by Local Councils to
decide on a DA are reported in Comparative
Information on NSW Local Government and are avail-
For data for the period 1990–91 to 1994–95, we able from 1994–95 to 2011–12.
use data from tables produced by the UNSW Local
Grants Commission. This reports (preliminary)
population estimates over the period 1990–91 to Geographic variables
2010–11. No data are reported for 1991–92; so, we The geographic variables are obtained from three
simply use the average of 1990–91 and 1992–93 for satellite-based geographic databases:
this year. Source: ABS Estimated Resident
Population of Statistical Local Areas New South (1) GEODATA TOPO 250K Series 3
Wales at 30 June 1990 Preliminary (Cat. No. 3210.1) (2) SRTM Digital Elevation Data, Australia and
(3) Local Government Area Digital Boundaries
(Australian Standard Geographical
Income
Classification 2006) in ESRI Shapefile format
Estimates of income for LGAs are based on data from
the Australian Taxation Office’s Taxation Statistics. using GDAL and GIS software. The first data set,
The Bureau of Infrastructure, Transport and Regional sourced from Geoscience Australia, provides a vector
Economics (BITRE) has a database derived from the representation of the major topographic features
ATO data that report by LGA real income per taxpayer appearing on the 1:250,000 scale NATMAPs. This
(in 2007–08 prices) for the period 1990–91 to 2005–06. allows us to calculate the area of each LGA that is
The original figures for nominal taxable income are taken-up with hydrology (e.g. lakes, reservoirs, water-
deflated using the CPI for Australia. courses and flats); habitation (e.g. build-up and recrea-
For the period 2006–07 to 2009–10, the ABS reports tion areas) and prohibited areas and reserves. The
data by LGA for nominal income per taxpayer in second data set (DEM) is a digital SRTM map with a
their National Regional Profile 2007–2011. We con- resolution of 90 m at the equator, sourced from
vert these figures to constant 2007–08 prices using CGIAR-CSI. The third data set, sourced from the
the CPI for Australia. Australian Bureau of Statistics, is LGA digital bound-
As we have no income data by LGA for the financial aries across Australia, which guarantee the LGA-spe-
years 2010–11 and 2011–12, we simply assume that cific level of the variables derived.
APPLIED ECONOMICS 5461