![Sample our Built Environment journals, sign in here to start your access, Latest two full volumes FREE to you for 14 days]({"type":"image" , "src" : "/sda/5926/TOC-Subject-Sample-2021-Built-Environment.jpg"})
Abstract
This paper is concerned with assessing how valuations for mortgage purposes reflect market evidence. Differences between the value obtained through the analysis of comparables and the final assigned value are analysed. The study is undertaken for the Spanish housing market at the peak of the house price boom. High levels of accuracy are apparent but with a tendency to over- rather than undervalue properties. Physical housing variables are shown to have a relatively homogeneous effect, whereas factors relating to the environment and location lead to wider differences between valuations. The effect of characteristics is shown to vary substantially between cities. There is no evidence from the analysis that the property bubble in Spain was driven by inaccuracy in valuations.
Acknowledgments
The authors wish to thank TABIMED for access to data sources. Also we express our appreciation to the anonymous reviewers for their considerable advice on this paper.
Notes
1. The valuation database used in this study has evidence on list price and not sale price, meaning that the paper cannot analyse bias between list and sale price.
2. The Orden Eco/805/2003 of 27 March, BOE no. 85, published on 9 April 2003.
3. The source of the data is from TABIMED, one of the major valuation companies in Spain.
4. The time period covered by the sample was at the end of a highly dynamic and major growth period for the Spanish housing market.
5. A minimum of six and in some cases up to 12 for each subject property were recorded in the database.
6. These include parking places, stores and workshops.
7. The expression of alpha is where: cjk
is the characteristic k of the testigo j, cik
is characteristic k of the property i to be valued, ik , jk are the coefficients – weights for each characteristic k both of the property and of the testigo and ij random component. See Taltavull and McGreal (2006).
8. OLS regression between both variables; results available on request from the authors.
9. A weighted least-square model, weighted by province, is also employed.
10. OLS models were initially run and are available from the authors on request. The 2SLS models were found to provide robust estimators of the parameters and are hence presented in this paper.
11. All models use 2SLS apart from the model with deviations lower than −15% and higher than15% for Alicante city due to sample size constraints.
12. The authors wish to acknowledge helpful comments on endogeneity suggested by a reviewer.
13. The province omitted is Lérida, located in the interior-north of Cataluña. The city of Lérida has 140,000 inhabitants; the main economic sectors are industry and services. The city has a high degree of accessibility to Barcelona and to the Barcelona–Madrid motorway.
14. This reflects renewal processes during the 1980s and 1990s which has taken place in most Spanish cities and the older city centres.
15. Public houses in Spain have a maximum price for valuation and are transaction fixed yearly
16. A homogeneity test of each coefficient was run in order to clarify the differences between the valuation results for the three cities. A chi-square test as well as directional and symmetric measures were calculated including the Goodman and Kruskal tau test, Somers’ d symmetric test, Kendall’s tay b and c, and Cramer’s V. Results for the Pearson chi -square (ρ value < 0.01) suggests that relationships exist among the characteristics for the three cities, however tests of strength, significance and direction indicate that any relationship captured between the cities is weak and does not affect the general model results.
17. Homogeneity tests, the F-Snedecor test, of the parameters for the truncated samples at a city level show that a high level of homogeneity is achieved with the exception of the Madrid model for the <= −15% sub-group. For >= 15% sub-group models F = 1.0398 for Alicante, F = 0.7054 for Valencia and F = 1.0417 for Madrid. For the <=−15% sub-group for Alicante F = 0.2809, for Valencia F = 0.2443 and for Madrid F = 141.136. For all models, with the exception of the latter, values are below the respective critical values at the 0.05 level.
18. Alicante does not have an underground network.
