CHAPTER 9 PRICES AND PRICE COLLECTION


INTRODUCTION

9.1 The HPI is compiled using median prices of observed transactions for each cluster. This chapter describes some issues associated with collecting transaction prices and determining median prices.


TRANSACTION PRICES

9.2 Price is defined as the value placed on a product at the point of transaction.

9.3 The HPI measures the transaction prices agreed upon by vendor and purchaser. That is, it does not take account of or include any taxes and subsidies.


PRICING POINTS

9.4 In the Australian context, there are four significant dates related to the purchase of a residential dwelling. A general timeline of the stages of the sale of residential dwellings is as follows:

• verbal agreement to purchase at a negotiated price;
• approval of mortgage financing;
• exchange of contract; and
• settlement of the property sale.

The date of exchange of contract is the preferred date

9.5 For the purposes of measuring price changes for houses, it is desirable to select the earliest date at which the final purchase price is set. The point in time at which the price is first determined is when verbal agreement is reached. However, there is no effective way to capture this information and it is possible for the originally agreed sale price to be renegotiated before the exchange of contracts. Approval of mortgage finance data is limited to those sales that involve mortgages. A house price index constructed on a settlement date basis incorporates a lag in identifying the turning points in housing prices as the settlement date can occur several weeks or months after the exchange of contract. It is for these reasons that, in compiling the HPI, the date of exchange of contract is the preferred date.


Estimating the date of exchange for Adelaide and Darwin

9.6 For most States, the VGs data include information on the date of exchange of contracts. However, the contract exchange date is not captured in either South Australia or the Northern Territory. For Adelaide and Darwin, the ABS estimates the contract exchange date from the settlement date. The estimates are modelled on the relationship between the settlement and exchange dates of price quintiles in Brisbane, where similar administrative arrangements exist. These models are reviewed from time to time to ensure their continuing effectiveness and relevance.


MEDIAN PRICES

9.7 Median prices are used to represent the prices of all houses sold in a cluster in a given period. Therefore, price movements at the cluster level are determined by comparing the median price of the cluster in one period to the median price of the cluster in another period.

9.8 The median price is the value of the middle observation from among an ordered ranking of house prices. Medians are preferred as they are not affected by extreme or outlier values like arithmetic mean or geometric mean calculations, and give the most robust and consistent measure of central tendency. The price relative calculated from two medians is the most reliable measure of price change.

9.9 The weighted movements of the median prices of each cluster contribute to the capital city index movement, as described in Chapter 7.

9.10 It is important to note that due to the compositional differences of the benchmark and leading indicator series, the price relatives from quarter to quarter are determined by comparing current benchmark medians with previous benchmark medians, and current leading indicator series medians with previous leading indicator series medians. For more information refer to Chapter 8.15.


TIMING AND FREQUENCY

9.11 The unstratified dataset of prices is obtained from VGs and a sample of mortgage lenders. Transaction data are provided on a monthly or weekly basis. Files provided may contain details of transactions which exchanged in the current quarter, as well as those which exchanged in earlier quarters and have recently been settled and/or been registered.

9.12 Note that due to provision arrangements and the requirement to publish the indexes in a timely manner, the transaction data for the most recent quarter will be biased towards the first two months of the quarter. A smaller proportion of transactions which exchanged in the third month will have been registered by the relevant authority and subsequently provided to the ABS.


COLLECTION PROCEDURES

9.13 Providers selected in the HPI sample submit a file of new or updated transaction or mortgage records to the ABS' Secure Deposit Box. The format of the files is not standardised, in order to minimise provider burden. Data items provided also vary across providers. Therefore, ABS systems and processes are required to transform the individual files into a standardised dataset for subsequent processing.


PRICE OBSERVATION

9.14 While price observations in the CPI and PPI suites of indexes are carefully selected and monitored to ensure that the same 'specification' is being priced each quarter, this is not possible in the compilation of the HPI. The sample of the housing stock transacted contains houses with different characteristics from quarter to quarter.

9.15 As described in previous chapters, the mechanism which is used to substitute for pricing the same item each period is the choice of index methodology: stratifying the range of price observations collected into clusters of housing stock with similar characteristics and then calculating a representative price for these clusters.


NON-MARKET TRANSACTIONS

9.16 A typical dataset will contain records of transactions which are not representative of the market. These could be categorised as transactions between related parties, such as family members or in divorce settlements, or could be the result of data entry error. Where these transactions can be identified, they are removed from the datasets and therefore do not contribute to cluster median calculations.


MISSING OBSERVATIONS

9.17 As outlined in Chapter 5, the stratification methodology chosen aimed to produce clusters which maximised both the homogeneity of the price-determining characteristics of houses in a cluster, and the number of price observations that could be expected each period, in order to calculate a reliable median.

9.18 However, in practice, some suburbs (the building blocks of the index) are so unique that to include them with other suburbs would produce clusters which do not meet homogeneity requirements. Therefore, some clusters are formed from one or very few suburbs, which means that the number of price observations available each quarter is extremely low, or sometimes nil.

9.19 The price movements of clusters 'missing' observations in a particular quarter are therefore derived using appropriate imputation methodologies during the editing process.


EDITING

9.20 The two processes described above (removing non-market transactions and imputation) are examples of the types of procedures involved in the editing stage of the index compilation process. The former is an example of input editing, while the latter is an example of output editing. Duplicate records within and between data provided by VGs and mortgage lenders are also identified in the input editing stage; these records do not contribute to median price calculations.

9.21 The function of editing is to ensure that the datasets used to compile the index are sound, and that price changes which aggregate to produce the index number are realistic. The price index analyst makes important decisions on the quality of the data and the results produced and takes steps to rectify any issues identified.


QUALITY ADJUSTMENT

9.22 The term 'quality' in the context of a price index refers to the characteristics of a product being priced in each period. Differing products will have different qualities. Also, the qualities of one particular product may not be fixed between time periods.

9.23 Where these changing characteristics are considered to be price-determinant (dimensions, materials, extra features, for example), an assessment is made of the change in quality and the magnitude of the corresponding effect on price. This may be a simple matter where the quality which changes is, say, the volume of liquid in a bottle. However, more subtle differences in qualities are more problematic, such as when a car model is upgraded.

9.24 In these situations, in order to 'price to constant quality' to measure pure price change, an adjustment is made to ensure the correct price relative is derived. There are a number of techniques which can be applied, and these are detailed in the CPI and PPI Concepts, Sources and Methods publications (cats. no. 6461.0 and 6429.0 respectively).

9.25 In the HPI, the stratification method is used to address the fact that each quarter there is a change in the composition of houses sold. Therefore each cluster is assumed to have a fairly homogeneous quality in terms of certain characteristics of the housing stock it includes (these characteristics are described in greater detail in Chapters 5 and 11). It is also desirable that the sets of price observations for each cluster should have a constant or consistent quality in factors like data sources, data provision and processing practices. There may be instances where a perceived change in quality does occur in a cluster from one period to the next. This may be due to data provision changes (such as introduction of a new provider into the sample), or systems processing issues. In these instances, a perceived change in quality of a cluster may prompt a quality adjustment to the cluster median.


CHECKING

9.26 During the collection and processing cycle, checks are made at numerous points to ensure:

• the integrity of the data collected;
• correct transformation in ABS systems;
• accurate median calculation;
• appropriate application of estimation, imputation and quality adjustment techniques; and
• acceptable confrontation of results with market intelligence and expectations.