Workshop on Residential Property Price Indices RPPI Handbook Chapter 6: Hedonic Regression Methods
Contents 6.1 Introduction 6.2 Time dummy variable method 6.3 Characteristics prices or imputation methods 6.4 Stratified hedonic indexes 6.5 Main advantages and disadvantages 6.6 – 6.8 Applications to Dutch town of “A” Chapter 6
6.1 Introduction Assumption: house is a bundle of (price determining) characteristics Different functional forms possible Typically estimated by least squares regression More flexible if characteristics parameters are allowed to change Chapter 6
6.2 Time dummy variable method Time dummies capture price change Usually: logarithmic-linear model Resulting price index has a geometric structure Revisions when new data is added Chapter 6
6.3 Characteristics prices / imputation methods Uses traditional index number formulae for a ‘standardized’ house Are imputation indexes (with quantities equal to 1) Linear and log-linear model specifications possible As with other methods: data cleaning may be necessary Chapter 6
6.4 Stratified hedonic indexes Stratification highly recommended … as different sets of characteristics will be needed for different market segments, and parameter values may differ across segments Sales RPPI: stratified Fisher hedonic imputation index probably the best solution Chapter 6
6.5 Main advantages and disadvantages Can adjust for both sample mix and quality changes (data permitting) Probably the most efficient method Imputation variant is analogous to matched-model methodology Data intensive General idea easily understood but some technicalities not so easy to explain Chapter 6
6.6 – 6.8 Application to Dutch town of “A” Three variables: land size, structure size and age Models with quality adjusted structures (net depreciation rate, depending on age) Including linear time dummy model Fisher hedonic imputation index is “best” - very similar to (non-hedonic) stratified Fisher index Chapter 6