Renovations and the repeat-sales house price index

Submission note: A thesis submitted in total fulfilment of the requirements for the degree of Doctor of Philosophy to the School of Economics, Faculty of Business, Economics and Law, La Trobe University, Bundoora.

This thesis consists of three chapters related to repeat-sales house price indices. In the standard repeat-sales method, the maintained assumptions are that the quality and the characteristic coefficients do not change between a house’s two sale dates. Both maintained assumptions have been called into question. For quality changes, renovations between a house’s two sales are not accounted for in the repeat-sales method. In the first chapter, I develop a repeat-sales index that is able to control for renovations in a way that is consistent with the underlying hedonic theory, yet maintains the parsimony and simplicity of the repeat-sales method. Moreover, I am able to specify the exact way that renovations enter the repeat-sales regression thus avoiding econometric problems related to mis-specification of the functional form. I also consider how to control for changes in neighbourhood quality. In the second chapter, I use a large dataset of Melbourne house prices from 1992Q1 – 2002Q1 to compare the renovations-adjusted repeat-sales index I developed in my first chapter to the standard version. I find that the standard version over-estimated house price appreciation by approximately 21 percent over the decade (sample period). I also find that that, controlling for neighbourhoods, houses with long intervals between sales appreciate more than quicker re-sale houses. In the third chapter, I test the second maintained assumption that characteristic coefficients from the underlying hedonic equations remain unchanged over time and reject it. Moreover, using the flexible functional form Box-Cox method, I find that the assumption that prices in the underlying hedonic equations should be entered in a log fashion is a reasonable approximation to the true functional form. I also find little difference in hedonic house price indices using linear or non-linear functional form for the independent variables.