6416.0 - House Price Indexes: Eight Capital Cities, Jun 2012 Quality Declaration 
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 01/08/2012   
   Page tools: Print Print Page Print all pages in this productPrint All

FEATURE ARTICLE: EXPERIMENTAL OTHER DWELLINGS PRICE INDEX


INTRODUCTION

1 This article provides information about the development of a new experimental price index which covers the prices of other dwellings (which includes flats, units, apartments and semi-detached houses) in the eight capital cities. This article outlines the index methodology and presents experimental price index results for other dwellings as well as a total price index for all dwellings. Subject to further analysis, available funding, and feedback from users, the ABS intends to incorporate the new indexes into its suite of price statistics in the future.


BACKGROUND

2 The coverage of the Established House Price Index (referred to from now on as the HPI) is currently limited to detached houses in the eight capital cities. A 2004 review of the HPI identified several possible improvements to the measure, one being the expansion in coverage to include dwellings other than detached houses (from now on referred to as other dwellings). In 2009 the ABS was provided funds for regular production of performance indicators related to the Council of Australian Governments (COAG) National Affordable Housing Agreement. One of these was a measure of housing affordability and this work has resulted in an extension of the collection and processing of house sales data from State and Territory Valuers-General (VGs) to cover all dwellings in the whole of each state. The collection of additional data and development of editing and processing tools for these data has provided the opportunity to do work on extending the scope of the HPI to include other dwellings.

3 According to 2011 Census data, other dwellings make up almost a quarter (24%) of all occupied private dwellings in Australia. The current coverage of the HPI may not be representative of all dwellings: segments of the property market may experience different demand and supply factors at any one time and over periods of time. Extending the coverage to all dwelling types in the capital city is a significant step to improving the usefulness of the HPI as an economic indicator. There is also a significant amount of user interest for the ABS to produce an index for other dwellings.

4 The ABS has adopted a two-stage approach to produce the HPI. The first stage is to compile the benchmark series based on the complete, or near complete, VGs dataset for each quarter. The second stage, referred to as the leading indicator series, involves compiling price indexes for the two most recent quarters based on a combination of mortgage lenders’ data and the VGs data available at that point in time. This article explains the proposed methodology (for both the benchmark, and leading indicator series) for an Experimental Other Dwellings Price Index (EODPI) to complement the current HPI, as well as an Experimental All Dwellings Price Index (EADPI) covering both detached houses and other dwellings in the eight capital cities. Experimental results of the two benchmark indexes are presented in a time series spreadsheet as part of this publication.


CONCEPTS AND METHODS

Approach Used for the Experimental Other Dwellings Price Index

5 The standard procedure for constructing price indexes is to select a sample of representative items and to re-price the identical items through time (a matched sample). This approach is not viable in the case of established houses or other dwellings as the observable prices in each period invariably relate to a different set of dwellings. The EODPI uses a stratification approach to control for compositional change in the sample of other dwellings used to compile the EODPI each quarter. This approach stratifies (clusters) other dwellings according to two characteristics: the long-term level of prices for the suburb in which the other dwelling is located, and the neighbourhood characteristics of the suburb, as represented by the ABS Socio-Economic Indexes for Areas (SEIFA). The cluster design process aims to seek a balance between maximising the homogeneity of dwellings within each cluster and ensuring that there will be sufficient price observations in each cluster to produce a reliable median price. The approach used to produce the EODPI clusters is identical to that used in the HPI, however the outcome for the EODPI is a different set of clusters to that produced for the HPI. The number of clusters in both indexes is presented below.

6 The EODPI is a fixed weighted index, where the weights are based on the value of the in-scope housing stock (other dwellings). The quantities underpinning the values are sourced from the most recent Population Census and are held constant until they can be updated with new Census data. Link period or base period values for each cluster are derived as the product of the number of other dwellings in the cluster and the average price of other dwellings in the cluster at the link or base period. Like the HPI, the EODPI is actually two indexes which have been chain linked at March quarter 2008. Series 1, based on the 2001 Census covers the September quarter 2003 to the March quarter 2008. Series 2, the current series is based on the 2006 Census and covers the March quarter 2008 to the present quarter. ABS intends to re-weight the index based on the results of the 2011 Census of Population and Housing. Further information about re-weighting and linking is available from House Price Indexes: Concepts, Sources and Methods, Australia (cat. no. 6416.0)

7 The VGs data provided to the ABS for use in construction of the EODPI will also allow for the compilation of a time series of the number of transfers and unstratified median prices for other dwellings in each of the eight capital cities. These figures are presented in a datacube as part of this publication.

Number of Clusters from the March quarter 2008 onwards

Other Dwellings
Established Houses

Sydney
10
22
Melbourne
10
20
Brisbane
10
20
Adelaide
5
11
Perth
4
10
Hobart
3
5
Darwin
3
6
Canberra
4
7




Benchmark Index Construction

8 The benchmark index number is produced for each quarter once data has been collected over three quarters - by this time the VGs dataset is considered complete or near complete (e.g. this Feature Article is being published in the June quarter 2012 which means the Benchmark series will go up to the December quarter 2011). In the regular production of the HPI the benchmark series is considered final and not subject to any further revision. Users should note, however, that as the EODPI and EADPI are experimental indexes, the data published in this Feature Article may be revised when the ABS first publishes the new series in the future.

9 The reliability of each index is largely dependant on the availability of sufficient pricing information each quarter. Users should note that the small cities of Hobart and Darwin have considerably fewer sales than other cities, with around 200 to 400 sales each quarter in each city, and there may be occasions when clusters have a low number of price observations leading to volatility in the index. Rather than suppress publication, the series are included here because it is believed that the long term trends are reliable. However, because of the limitations in the reliability of individual quarter-to-quarter movements, users are advised to exercise due care when analysing such movements.

10 To calculate the benchmark series, each quarter a price relative for each cluster is obtained by dividing the median other dwelling price in the current period by the median other dwelling price in the previous period. The dwelling stock is then re-valued based on the price relatives. This value is then the current estimated value of the other dwelling stock in each cluster. An aggregate capital city value is obtained by summing the value of each cluster, and then the price index for the city is then derived by dividing this current period value by a link or base period value and then multiplying by a link or base period index number. The manner of index construction is the same as is used in the HPI. Refer to House Price Indexes: Concepts, Sources and Methods, Australia (cat. no. 6464.0) for more information on this approach.

11 Once the current period value of the dwelling stock in all capital cities has been obtained, these values can be summed together to allow the derivation of a weighted average index of the eight capital cities. The weight of each capital city in the eight capitals index is presented below.

Experimental Other Dwellings Price Index: Weighting Pattern(a), percentage contribution to the weighted average of eight capital cities

Other Dwellings(b)

Sydney
45.6
Melbourne
26.9
Brisbane
9.2
Adelaide
5.8
Perth
9.2
Hobart
0.7
Darwin
0.7
Canberra
2.1

(a) As at the March quarter 2008.
(b) Percentages may not add due to rounding.



Proposed Methodology for the Leading Indicator Series

12 In the HPI, the two most recent quarters (the leading indicator series) are compiled by using a combination of mortgage lenders' data and VGs data. Both sets of data are merged and duplicates (where the same property appears in both data sets) are identified and removed (the VGs data is kept, whilst the mortgage lenders data is discarded where a duplicate is found). The first estimate (the most recent quarter) of the EODPI and HPI is referred to as the P1 estimate and the second estimate (the second most recent quarter) is referred to as the P2 estimate. Due to the fact that identification of duplicates is not as straightforward for other dwellings as it is for houses, a number of alternative methods were investigated for compiling the EODPI indicator series. These methods were assessed on how accurately the leading indicators predicted the benchmark result.

13 For the P2 estimate, it is proposed to compile the index in the same way as the benchmark series, using only VGs data. For the P1 estimate (except for Perth), the proposed method involves using VGs data for the first two months of the quarter, and mortgage lenders data for the third. For example, the data used to compile the June quarter 2012 P1 estimate would consist of VGs data for properties with exchange dates in April and May 2012 and mortgage lenders data for properties with exchange dates in June 2012. This merged data set is then used to calculate the median for each cluster and update the estimated value of the other dwelling stock accordingly. For Perth, it is proposed to compile the P1 estimate using VGs data only as this produces fewer revisions to the indexes. As in the HPI, P2 and P1 estimates for the EODPI (and consequently the EADPI) will be subject to revision, even if compiled only from VGs data.

14 As in the HPI, in order to minimise bias in compiling the index, price relatives are determined only by comparing current benchmark medians with previous benchmark medians and current leading indicator medians with previous leading indicator medians. Thus, in the leading indicator series, medians from the current second preliminary estimates (P2) quarter are compared with medians from the previous second preliminary estimates quarter, and medians from the current first preliminary estimates (P1) quarter are compared with medians from the previous first preliminary estimates quarter. Using this approach, price relatives are derived from datasets which have a similar composition of VGs and mortgage lenders' data.

15 The methods proposed above have, to date, produced the most reliable leading indicator series. However, the methodology proposed is subject to change as the leading indicator series is still being evaluated, and for this reason results of the leading indicator series are not presented in this Feature Article. The ABS welcomes user feedback on the proposed methodology.


Approach used for the Experimental All Dwellings Price Index

16 A benefit of using the total value of the dwelling stock to construct price indexes allows for aggregation to higher levels (as values are expressed in dollars, they can be added together without further transformation). This property is the basis for the construction of the weighted average of the eight capital cities in both the HPI and EODPI. The value of the dwelling stock for each city in the current period is added together for a value of the dwelling stock in the eight capital cities. This additivity means that with information on the total value of the established house dwelling stock and other dwelling stock, prices indexes can be constructed for all dwellings in each of the capital cities, as well as a weighted average of the eight capital cities. The Experimental All Dwellings Price Index (EADPI) is an experimental price index that measures the price change for the stock of all dwellings in each capital (as well as a weighted average of the eight capital cities).

17 Users should note however, that prior to the March quarter 2008, the Experimental All Dwellings Price indexes are not constructed in the manner described above. Methodological differences between the HPI and EODPI prevent the use of the standard value aggregate approach. Prior to the March quarter 2008, the Experimental All Dwellings Price Indexes are weighted averages of the HPI and EODPI. The weights are based on the value of the capital city housing stock from the 2001 Census at the September quarter 2003 prices as this methodology is similar to that used for the construction of link period value aggregates.

18 For example, if the weight for other dwellings was 0.3 (and the index was 120.0), and the weight for established houses was 0.7 (and the index was 110.0), the EADPI would be calculated as follows:

EADPI = (0.3 x 120.0) + (0.7 x 110.0)

EADPI = 113.0

19 The weighting pattern for the EADPI is presented below.

Experimental All Dwellings Price Index: Weighting Pattern(a)(b)

Percentage contribution to capital city index
Percentage contribution to weighted average of eight capital cities
Established Houses
Other Dwellings
All Dwellings
All Dwellings

Sydney
69.3
30.7
100.0
36.4
Melbourne
75.7
24.3
100.0
27.2
Brisbane
81.4
18.6
100.0
12.2
Adelaide
80.0
20.0
100.0
7.1
Perth
82.6
17.4
100.0
13.0
Hobart
84.5
15.5
100.0
1.1
Darwin
73.5
26.5
100.0
0.6
Canberra
78.8
21.2
100.0
2.4

(a) As at the March quarter 2008.
(b) Percentages may not add due to rounding.



THE EXPERIMENTAL OTHER DWELLINGS PRICE INDEX

Scope and Coverage

20 The HPI measures price change of the stock of established houses over time. Established houses are defined as detached residential dwellings on their own block of land regardless of age. The EODPI is designed to measure the price change of the stock of other dwellings in the capital cities whose primary purpose is residential. For the purposes of the EODPI, other dwellings are defined consistent with ABS standards, the Functional Classification of Buildings (cat. no. 1268.0.55.001) and the Dwelling Structure Classification (STRD) used in the Census of Population and Housing (refer to Census Dictionary, 2006 (cat. no. 2901.0)).

21 Examples of dwellings in scope of the EODPI are:
  • Semi-detached, row or terrace houses
  • Townhouses
  • Flats, units and apartments.

22 Examples of dwellings out of scope of the EODPI are:
  • Detached houses
  • House with office
  • House with flat.

23 When data are received by the ABS, processing classifies residential buildings into either established houses or other dwellings. However, given that the VGs in each State and Territory have different methods of classifying residential buildings, some dwellings (in Sydney, Brisbane, Hobart and Darwin) in-scope for EODPI may instead be included in the HPI established houses price sample.

24 The scope of the EODPI includes other dwellings within capital city statistical divisions (SD), as defined in the Australian Standard Geographic Classification (cat. no. 1216.0), however coverage is limited to those suburbs which existed within SD boundaries at the time of the 2006 Census. As the ABS has moved to a new geography classification, the Australian Statistical Geography Standard (ASGS) (Vol 1, cat. no. 1270.0.55.001), it is planned that the EODPI, together with the HPI will move to the ASGS during the next re-weight (to be implemented for the HPI in the December quarter 2013), and capital cities will be defined by the ASGS Greater Capital Cities Statistical Areas.

25 The ABS uses the date of exchange of contracts to determine when the price of the property was agreed upon, even though the property may not legally change owners until several months after that. Exchange dates are not captured in South Australia or the Northern Territory. The exchange dates for Adelaide and Darwin are modelled based on the settlement date (the date the property legally changes hands).


Results and Analysis: December quarter 2011

26 The experimental price index for other dwellings for the weighted average of the eight capital cities fell 1.4% in the December quarter 2011, compared to the final index for established houses which fell 0.6% in the quarter. Overall, the experimental price index for all dwellings in the eight capitals fell 0.8% in the December quarter 2011.

27 For other dwellings, the negative quarterly movement was a result of falls in Sydney (-2.3%), Melbourne (-1.7%) and Perth (-0.7%). This was partially offset by rises in Brisbane (+1.4%), Adelaide (+1.6%), Canberra (+2.7%), Darwin (+0.4%) and Hobart (+0.2%).

28 In Brisbane in the December quarter 2011, all of the positive contribution to growth was in clusters where median prices were below $400 000. Price increases in clusters with other dwellings median prices in the range of $200 000 to $400 000 were also seen in Sydney, Darwin, Hobart and Adelaide. Clusters with median prices above $600 000 experienced moderate falls in prices in Sydney and Melbourne in December quarter 2011.

29 For all dwellings, the negative quarterly movement was a result of falls in Sydney (-1.6%) and Melbourne (-1.3%). This was partially offset by rises in Canberra (+2.3%), Brisbane (+0.3%), Perth (+0.3%), Adelaide (+0.3%), Hobart (+1.3%) and Darwin (+1.4%).

30 In the year to the December quarter 2011, the experimental price index for other dwellings for the weighted average of the eight capital cities fell 3.0%, whilst for established houses it fell 4.4%. Overall, the experimental price index for all dwellings in the eight capital cities fell 4.0%.

Percentage Change from Corresponding Quarter of Previous Year: December quarter 2011
Graph: Percentage Change from Corresponding Quarter of Previous Year: December quarter 2011


31 Over the past 5 years the experimental price index for other dwellings for the weighted average of the eight capital cities has risen 23.9%, with all the capital cities having a positive movement. Over the 5 year period, Perth has risen the least (+7.4%). All other cities have rises in their indexes of above 10%, with the largest rise in the index being in Darwin (+47.2%).

Experimental Other Dwellings Price Index (a), Weighted average of eight capital cities
Graph: Experimental Other Dwellings Price Index (a), Weighted average of eight capital cities


32 A full time series of both indexes is presented in two tables at the end of this Feature Article. Further information on the two experimental indexes, which includes a full time series of quarterly and annual percentage changes of both index series, as well as a time series of unstratified medians and transfers for other dwellings, are available as a datacube from the ABS Website.


FUTURE DEVELOPMENT

33 The ABS believes the experimental benchmark indexes presented in this paper are robust and fit for purpose. In the absence of any compelling arguments to the contrary, it is proposed that the experimental label be lifted and the index continue to be compiled and published in some form in the future. The leading indicator series is still undergoing evaluation and the publication of these indexes will be subject to them being assessed as fit for purpose.

34 Development of indexes covering areas outside the Capital City Statistical Divisions is continuing. The ABS expects to begin to produce experimental indexes this year, which, subject to data quality, will be the subject of a Feature Article in the future.

35 The ABS welcomes any comments on the developments described in this paper and on any aspect of the work undertaken to date. Of particular value would be any suggestions as to how best to publish the results in future. The ABS will be accepting feedback until 9 November 2012.

36 Queries or comments can be addressed to:

Mr. Mark Dubner
Assistant Director
Prices Branch
GPO Box 796
Sydney NSW 2001

Telephone: (02) 9268 4448

E-mail: house.prices@abs.gov.au

Table 1 - Experimental Other Dwellings Price Index Numbers(a)(b)

Period
Sydney
Melbourne
Brisbane
Adelaide
Perth
Hobart
Darwin
Canberra
Weighted average of eight capital cities

2003
September
100.0
100.4
92.0
96.3
95.2
91.1
90.2
98.1
98.9
December
102.1
100.1
100.9
99.3
98.2
98.9
101.7
101.2
101.1
2004
March
99.6
99.1
102.8
100.3
102.2
106.1
102.5
99.8
100.0
June
98.3
100.4
104.3
104.1
104.5
103.9
105.6
100.9
100.0
September
98.7
99.2
106.3
104.6
105.4
104.5
103.6
96.4
100.1
December
100.0
100.9
106.8
108.1
109.2
110.2
118.2
98.1
101.8
2005
March
98.8
103.1
107.4
106.0
113.6
115.5
133.6
101.3
102.0
June
99.6
103.2
110.6
108.7
119.8
115.9
136.4
100.6
103.2
September
98.5
102.9
111.9
112.8
124.9
112.6
129.5
101.8
103.2
December
98.4
102.7
113.8
114.7
132.5
115.0
148.8
102.5
103.9
2006
March
96.7
107.9
114.5
113.8
143.5
120.9
159.0
103.6
105.0
June
98.1
109.4
119.7
116.2
163.9
123.6
167.4
103.9
107.9
September
96.9
109.8
119.9
117.0
172.0
121.9
161.9
104.1
107.8
December
98.6
110.4
122.7
119.2
178.2
124.9
168.9
105.2
109.7
2007
March
95.9
114.3
125.9
122.6
180.1
128.4
169.6
109.1
109.8
June
100.3
121.7
133.9
130.7
183.0
125.5
183.1
113.4
115.2
September
102.4
125.2
140.9
138.3
186.6
131.3
190.0
120.1
118.5
December
104.0
130.1
147.1
141.1
189.4
134.9
187.9
121.6
121.3
2008
March
102.8
129.0
148.2
146.2
188.0
131.4
190.5
123.5
120.7
June
101.4
128.4
150.4
142.2
187.6
137.0
188.7
121.4
119.7
September
100.6
126.4
145.2
145.4
181.8
137.0
206.9
121.1
118.3
December
99.7
126.2
143.4
142.4
175.0
135.1
210.1
119.5
117.1
2009
March
101.0
129.6
143.3
145.2
174.9
136.8
224.7
125.1
118.9
June
104.1
134.6
148.3
149.6
185.6
137.9
226.0
128.8
123.1
September
107.7
142.7
154.1
149.4
187.2
143.5
236.6
131.6
127.7
December
113.2
152.6
156.7
156.6
201.2
148.7
252.9
136.9
134.7
2010
March
115.9
158.1
156.5
161.2
203.0
147.6
253.5
139.8
137.9
June
118.5
161.6
159.1
163.4
205.6
146.6
257.1
141.2
140.6
September
117.8
159.1
156.1
159.6
200.4
144.5
263.3
140.0
139.0
December
119.1
160.1
154.6
163.0
201.8
145.8
266.4
142.4
140.1
2011
March
118.8
160.7
150.4
161.6
201.3
144.5
254.6
141.7
139.6
June
120.0
159.4
152.8
157.6
199.2
145.5
248.4
139.2
139.7
September
119.4
155.4
150.7
157.6
192.6
142.7
247.5
137.8
137.8
December
116.6
152.8
152.8
160.2
191.3
143.0
248.6
141.5
135.9

(a) Series are experimental and subject to revision
(b) Reference base of each index: 2003-04 = 100.0

Table 2 - Experimental All Dwellings Price Index Numbers(a)(b)

Period
Sydney
Melbourne
Brisbane
Adelaide
Perth
Hobart
Darwin
Canberra
Weighted average of eight capital cities

2003
September
98.9
99.3
93.1
96.9
94.2
91.3
92.9
97.8
97.7
December
102.3
101.6
100.6
99.5
98.7
99.2
99.2
101.6
101.4
2004
March
100.9
99.5
102.4
100.9
102.3
102.4
104.3
100.2
100.9
June
97.9
99.6
103.9
102.7
104.8
107.1
103.5
100.4
100.0
September
97.6
99.6
103.2
104.5
106.6
107.9
107.2
98.3
100.0
December
98.3
102.1
104.4
106.8
111.3
111.2
113.9
99.9
101.7
2005
March
96.5
102.3
105.2
106.8
115.8
113.0
122.9
100.7
101.5
June
95.8
103.4
106.3
108.0
122.0
114.7
125.5
100.1
102.2
September
94.4
103.3
106.6
109.1
127.2
114.4
128.7
100.5
102.1
December
94.9
104.8
108.7
111.5
136.5
117.9
138.1
102.7
104.0
2006
March
93.7
107.1
109.5
112.5
147.4
120.8
144.9
103.9
105.2
June
95.4
109.9
112.3
114.3
168.6
124.4
153.9
106.4
109.0
September
95.1
111.5
113.8
115.0
185.5
126.2
154.3
108.9
111.0
December
95.9
113.8
116.7
117.9
191.9
128.4
161.6
110.5
113.0
2007
March
95.0
116.2
121.2
120.3
193.1
132.0
165.6
112.6
114.0
June
98.8
124.3
129.0
127.6
190.5
133.7
169.8
117.5
119.1
September
101.2
130.1
135.3
135.3
193.9
137.6
174.9
123.6
123.3
December
103.4
138.7
142.2
143.2
196.2
143.1
179.5
126.9
128.0
2008
March
102.6
140.3
145.8
147.8
194.0
140.0
178.1
128.0
128.5
June
101.2
139.9
146.8
146.1
190.3
142.0
179.8
125.7
127.4
September
99.3
135.8
141.3
146.6
185.8
139.2
187.5
122.1
124.5
December
98.0
134.7
138.9
145.8
181.2
140.0
193.3
121.5
122.9
2009
March
97.2
135.1
139.1
145.1
180.5
139.3
199.9
123.0
122.7
June
101.4
142.3
143.2
149.2
185.5
143.7
203.9
127.0
127.7
September
105.7
151.3
148.0
151.3
190.4
147.1
211.6
132.0
133.1
December
111.4
161.4
152.7
157.5
201.9
155.4
226.3
139.9
140.4
2010
March
114.7
169.1
154.1
160.0
207.7
158.0
227.7
145.7
144.9
June
117.6
173.8
155.0
163.0
207.9
154.5
231.2
145.5
147.6
September
117.2
170.7
152.6
161.8
202.5
154.4
231.8
145.6
145.9
December
117.4
172.9
152.2
163.3
202.6
159.0
233.6
146.9
146.7
2011
March
117.1
171.3
149.2
160.9
200.8
157.9
228.0
146.6
145.5
June
117.6
169.5
148.5
157.2
195.4
153.4
221.2
146.0
144.4
September
115.9
165.8
144.7
156.4
193.3
150.0
222.1
141.4
141.9
December
114.1
163.6
145.2
156.9
193.8
151.9
225.1
144.6
140.8

(a) Series are experimental and subject to revision
(b) Reference base of each index: 2003-04 = 100.0



FURTHER READING

House Price Indexes: Concepts, Sources and Methods, Australia (cat. no. 6464.0)

Information Paper: Renovating the Established House Price Index, Nov 2005 (cat. no. 6417.0)

Research Paper: Refining the Stratification for the Established House Price Index (Methodology Advisory Committee), Jun 2008 (cat. no. 1352.0.55.093)

Consumer Price Index: Concepts, Sources and Methods, 2011 (cat. no. 6461.0)