Capitalising the Value of Free Schools: The Impact of Land Supply Constraints

2 Abstract There has been a growing literature in both the US (for example Haurin and Brasington 1996, and Black 1999) and the UK (for example Gibbons & Machin, 2001), which estimates the way in which school quality is capitalised into house prices. In earlier work the present authors (Cheshire & Sheppard 1995; 1999) estimated a hedonic model for the housing markets of Darlington, Nottingham and Reading in the UK in which the quality of the secondary school to which a household was assigned was a significant variable. The more recent work measured secondary school quality using a continuous measure of exam success. This provided evidence that the value of secondary school quality was being capitalised into the price of houses. In contrast Gibbons and Machin concluded that primary schools had an identifiable and significant price associated with their quality but that secondary schools did not. Their study did not have data for individual houses but used post-code sector data and then various techniques to standardise for all but one variable: either the notional primary school catchment area or the notional secondary school catchment area. All of these analyses are predicated on the assumption that the value of local schools should be reflected in the value of houses. We expect variation in the capitalised price of a given school quality at either primary or secondary level according to the elasticity of supply of 'school quality' in the local market. This will vary systematically between and even within cities and this paper explores the sources and impact of such variations. Using an hedonic model and data from 1999-2000, we estimate values attached to both secondary school and primary school quality in the Reading housing market. The results support the conclusion that both secondary and primary school quality is capitalised into the market price of houses and that the capitalisation of school quality is discounted within the highly constrained city of Reading in areas where new construction is concentrated. We also find evidence that appropriate model specification is imperative since bias is evident both when key neighbourhood characteristics are omitted and if the actual allocation of addresses to schools is not included.