Seminar on organization of a statistical survey on house price changes and the calculation of the HICP Kiev, 27 – 28 February 2013 Conceptual issues relating to house price index calculations and the practical methods available David Fenwick National accounts Conceptual & measurement issues Stock versus sales weighted Decomposition of building & land Constant quality – mix-adjusted, hedonics, repeat sales

Handbook on Residential Property Price Indices – represents current state-of-the-art Makes recommendations on best practice – Including how to improve international comparability Recommendations take into account the different situations confronted by countries – Most particularly data availability Cannot be too prescriptive but Increasing international comparability can be achieved by the introduction and adoption of standards on conceptual, methodological and computational issues Conceptual basis of the index – Type of index to be compiled, will depend on its purpose The System of National Accounts (SNA) 1993 should be used as conceptual framework Chapter 4 describes conceptual framework Chapters 5 to 8 describe the main statistical methods Chapter 12 gives recommendations on best practice.

Recommendations and Guidelines – the Laspeyres principle & the issue of weighting Broadly speaking two separate Residential Property Price Indices can be distinguished – A constant quality price index for the stock of residential housing at a particular moment in time – A constant quality price index for residential property sales that took place during a particular period of time. Also the scope of the index will depend on its use – Whether all properties should be covered or only new ones

Recommendations and Guidelines – weighting & decomposition between building & land components A stock-weighted index is appropriate when measuring the wealth associated with the ownership of residential property should be stock-weighted – A stock-weighted index is also appropriate for a financial stability indicator, in particularly to identify property price bubbles. A sales-weighted index is appropriate for measuring the real output of the residential real estate industry – This is consistent in treatment to the acquisition or purchase of goods and services in a consumer price index. A decomposition between the building & the land is an added complication – should be made where a country’s balance sheet estimates of national wealth make this distinction – may also be required when a residential property price index is an input into the CPI for the measurement of owner-occupier housing costs using the net-acquisition approach

Recommendations and Guidelines – scope A price index covering all residential property is appropriate for measuring the wealth associated with the ownership of residential property – The index should cover existing properties and properties which have been recently built – It should include conversions of existing property, for example where a warehouse has been converted into flats or an existing property has been sub- divided. An index covering all residential properties is also appropriate when used as a financial stability indicator A price index covering new property only is appropriate for measuring the real output of the residential real estate industry The value of new housing is part of gross investment – The cost of the land, apart from the value of any improvements made to this element, should be excluded A price index restricted to new property is also appropriate for the inclusion of owner-occupier housing costs on a net-acquisition cost basis in a consumer price index – I.e. where the consumer price index covers the cost of acquiring properties which are new to the owner-occupier housing market

Recommendations and Guidelines – constant quality A residential property price index (stock or sales-weighted) compares the values of the stock or of the sales of residential property between two time periods – after allowing for changes in the attributes of the properties involved – Price changes need to be decomposed into those associated with changes in attributes and the residual which relates to the underlying “pure price” change Challenging due to the following three factors – Residential properties are heterogeneous No two properties are identical – Prices are often negotiated The (asking) price of a property is not fixed & can change throughout the transaction process until the price is finalised A property's market value can only be known with certainty after it has been sold – Property sales are infrequent For example, typically less than ten per cent of housing stock changes hands every year A given house is likely to have a confirmed value not more than every ten years

Recommendations and Guidelines - Statistical methods for compiling constant quality indices Stratification (“mix-adjustment”) – The most straightforward method for controlling for changes in the composition or ‘quality mix’ of properties sold – Also addresses any user need for sub-indices relating to different housing market segments Stratification or “mix-adjustment” is an appropriate method where – An appropriate level of detail is chosen for the number of cells and can be applied in practice – A decomposition of the index into structure and land components is not required Stratification/mix-adjustment is recommended where the volume of sales is large enough to support a detailed classification of properties

Recommendations and Guidelines - Statistical methods for compiling constant quality indices Hedonic regression Simple concept - measures the empirical relationship between the observable attributes of a good or service e.g. a house & the price But no uniformity in practical application or agreement on what is best practice – Although the ‘best’ form of the hedonic function may be linear rather than log-linear Reflects fact that the value of a property is generally equal to the sum of the price of the structure and the price of the land In practice, many explanatory variables will be categorical rather than continuous and represented by a set of dummy variables. Two alternative methods of application of hedonics – Time dummy variables (revisions issue) versus predicted prices – Hedonic regression is generally the best technique for constructing a constant quality residential property price index using the predicted prices (imputations) approach (plus stratification)

Recommendations and Guidelines - Statistical methods for compiling constant quality indices (repeat sales) Repeat sales – Observes price development of a specific house over time by reference to selling price each time it is sold – The price development of a “representative” selection of houses during overlapping time periods can then be used to measure trend in property prices – Measuring the average price changes in repeat sales on same properties gives a like for like comparison Recommendation - repeat sales method is not preferred above the hedonic method for constructing a constant quality residential property price index – Problem of home improvements – But methodology is satisfactory where There is limited or no information on housing characteristics There are a relatively large number of repeat transactions – To provide enough data points to populate the required types of residences and where sample selection bias is not a problem – Repeat sales method is not recommended when distinction needed between price of structure & price of land

Recommendations and Guidelines - Statistical methods for compiling constant quality indices (appraisal-based methods) Appraisal-based methods (applied to matched-models) – Use “assessed” values e.g. valuations for taxation purposes or from specially commissioned surveys using estate agents Often done by reference to similar properties that have been sold – Overcomes two problems associated with repeat sales methodology The relatively small number of price observations which are generated The susceptibility to sample selection bias e.g. High turnover amongst starter homes or houses that have been renovated – But cannot deal adequately with quality changes to individual houses (same as repeat sales) & relies on expert judgement

Recommendations and Guidelines – Other statistical methods for compiling constant quality indices: appraisal-based methods SPAR INDEX In some countries information on assessed values or appraisals of properties is available, which might be useful as proxies for selling prices or, more generally, market values In countries where assessed values have been collected for tax purposes, appraisals will typically be available for all properties at a particular reference period – Assessed values can be used in addition to sale prices in a repeat sales framework to reduce the problem of inefficiency and the potential problem of sample selection bias – For each property sold in some comparison period there is a sale price & a base period “price” – the assessed value. Price relatives with a common base period – the valuation period – can then be constructed The value-weighted arithmetic Sale Price Appraisal Ratio (or SPAR) index – Re-scales appraisal-based indices by dividing by the base-period values – Corrects for the potential bias which may result from inaccurate valuations – Problems (bias) can arise from frequent re-assessments and reduced precision over time can arise from new appraisals

Recommendations and Guidelines – Other statistical methods for compiling constant quality indices: appraisal-based methods SPAR INDEX The main advantages – It is consistent with index number theory & straight forward to compute Being based on standard matched model methodology – Can benefit from many more observations than the repeat sales method In which case less susceptible to problems arising from small number of price observations – It is not susceptible to sample selection bias – No revisions (because no modelling) The main disadvantages – It cannot deal with major repairs or renovations (or depreciation) of the dwelling units, lack of data to decompose land from building – When the properties are reassessed and new appraisal data becomes available, the SPAR index can, and probably should be, rebased

Recommendations and Guidelines – Other statistical methods for compiling constant quality indices: appraisal-based methods SPAR INDEX Recommendation: preferred to the repeat sales methodology – if assessment data of sufficient quality is available The SPAR methodology addresses some of the weaknesses of the repeat sales methodology – E.g. Selection bias – Free from revisions, no modelling The SPAR methodology does have its drawbacks but is a recommended when hedonics (particularly in combination with stratification) is not possible Used in several European countries, notably in Denmark, the Netherlands and Sweden

House price index calculations and the practical methods available: UK experience No single definitive official house price index – A number of UK government agencies produce/publish information on UK house prices Including Office for National Statistics (ONS), Land Registry & Valuation Office ONS responsible for official House Price Index (transferred from DCLG in April 2012) – There are other statistics on house prices which are produced by a range of private sector bodies using differing sources and measures

House price index calculations and the practical methods available: UK experience

House price index calculations and the practical methods available: UK experience – similar in Canada

House price index calculations and the practical methods available: UK experience - similar in Japan

House price index calculations and the practical methods available: UK experience – information flow similar in Japan

House price index calculations and the practical methods available: UK experience – methodology of official house price index produced by ONS The ONS House Price Index is. – Published monthly. – Based on completed mortgage transactions. Excludes cash purchases – Measures the price actually paid for a property in the UK. At the end of the sale/purchase phase. Some ambiguity - price paid versus valuation by mortgage lender. – Source is the Regulated Mortgage Survey (RMS) as collected by the Council of Mortgage Lenders (CML).

House price index calculations and the practical methods available: UK experience - methodology of official house price index produced by ONS The official index is a mix-adjusted chained Laspeyres-type price index. Variables collected for mix-adjustment. – Locality (postcode). – Type of dwelling (detached, terraced etc.) – Whether or not the property is new or old. – Number of habitable rooms. – Number of bedrooms. Chain linking of the index is carried out annually. Also records whether the buyer was a first time buyer or not.

House price index calculations and the practical methods available: UK experience – official House Price Index Weights are transaction weights (not stock), calculated using two data sources Based on the preceding three years’ transactions – For England and Wales, data as recorded by HM Land Registry (HMLR) are used. All house sales in England and Wales are recorded with the HMLR and therefore this represents a complete data set for these periods – For Scotland and Northern Ireland the same level of data is unavailable The numbers of transactions in the RMS dataset are used as a proxy

House price index calculations and the practical methods available: UK experience – official House Price Index If a record has a missing value, it is imputed from the base data using the set of records with the same values for the non-missing variables. For example If a record has missing values for the number of rooms and the old/new dwelling indicator – The number of records (N) with the same values for the other analysis variables (region, acorn group, cluster group, dwelling type, first time buyer marker, old/new property) plus price band is counted – A random number (n) in the range 0 to N is generated – The missing values are imputed by setting them to be the same as those of the nth record with the same values for the other analysis variables – If no matching records can be found, the least influential variable is dropped and the matching process repeated. This process is repeated until only region and price band are left to be matched against

House price index calculations and the practical methods available: UK experience – official House Price Index The average price for each cell is calculated each month by fitting a regression model to the set of data covering the mortgage transactions reported during the most recent month. – The dependent variable is the natural log of the price. – The independent variables are ONS’s Cluster group – a socio-economic and demographic area indicator Acorn class – a commercial geo-demographic and lifestyle classification First-time buyer or former owner occupier Dwelling type – e.g. detached house, flat etc Old /new property Number of rooms

House price index calculations and the practical methods available: UK experience – official House Price Index Limitations of the official house price index – ONS’s index only covers those properties purchased using a mortgage - cash sales are explicitly excluded – The index is the least timely one, being published on the second Tuesday in the second month after the reference period. – No indices published lower than regional level But not usually subject to revision Technically the most rigorous of the available indices

Recommendations and Guidelines Which method is best depends on local circumstances – the “best” may not be the “ideal” Meta-data on residential property price indices published by different countries is available from the website of the Bank for International Settlements (see andbook Different uses - different concepts? - Yes: addressed in Handbook Can international best practice be identified or is a flexible approach appropriate to reflect local circumstances? Yes: in principle, but cannot be too prescriptive Which techniques are most widely used? Relates to mix-adjustment – a number of options Can existing data sources be improved? Possibly: worth trying (new data collection expensive) What criteria should guide the production of Residential Property Price Indices? User driven? How far should and can a handbook go in answering these questions? Cannot be too prescriptive The End