Scores
Section scores are the unweighted mean of item scores.
Item scores are distance-to-frontier (DTF) transformations of underlying variables.
In some cases, the DTF calculation utilises a truncated range or a truncated maximum. This is typically set to 3.5 standard deviations below or above the mean. Exceptions are flood area, flood risk, walking, cycling and greenbelt, all at 7 standard deviations. Observations outside this range are given 0 or 1.
In some cases, item scores are weighted combinations of DTF transformations of underlying variables.
In some cases, DTFs are based on logarithmic transformation of variables. This applies to house price ratios, incinerator distance, airport distance, airplane traffic, antisocial behaviour, community centres (count), playgrounds (count), parks (count), nearest library (distance), nearest swimming pool (distance), cafes (count), mean distance to all cuisine types, tree cover, bus frequency, distance to nearest local rail station, rail journey time to nearest major city (minutes), cycling score, distance to nearest landscape and distance to nearest shoreline.
In all cases, a higher score indicates a ‘better’ performance. This means, in some cases, the item score is 1 minus the DTF. This applies to house price ratios, house price deviation from model valuation, nonresidential % of buildings, flood area, flood risk, airplane traffic, deprivation, antisocial behavior, school quality (higher Ofsted scores indicate lower quality), distance to nearest library, distance to nearest swimming pool, mean distance to all cuisine types, rail journey to nearest major city (minutes), distance to nearest local real station (miles), nearest landscape (miles) and nearest shoreline (miles).
Model Insight
Oxford Scientia’s cross-sectional model of postcode-sector-median house prices is based on a four-quarter rolling average of median house prices per postcode sector. House prices in question are non-newbuild, terraced houses.
The model’s independent variables include a range of information in the Neighbourhood Navigator and outside of the Neighborhood Navigator. The model explains 90% of variation in median house prices on a cross-sectional basis.
The Model Insight score is the difference between the actual median price of the postcode sector and the model estimate, expressed as a percentage of the model estimate. A positive difference indicates overvaluation relative to what the model predicts.
Sources
House prices are from HM Land registry. Other data come from UK government sources including Ofsted, Defra, CAA, Police UK, CQC, ONS, DfT, Network Rail, Natural England, the Met Office, MHCLG and Historic England.