Abstract
In this paper, we empirically test the stability of valuation estimates of road traffic noise based on house prices. A hedonic valuation model is constructed and estimated for two separate housing markets in Belgium. We estimate noise depreciation sensitivity index (NDSIs) for different modelling choices for both markets separately, jointly and taking into account spatial dependency. The results confirm that housing markets are region-specific and several housing characteristics are valued differently across regions. The effect of road noise, however, appears to be rather robust. Thus, the use of NDSI estimates from one region to value traffic noise in another region seems to be acceptable.
Acknowledgements
We would like to thank the Environment, Nature and Energy department (Department voor Leefmilieu, Natuur en Energie LNE) for providing us with the necessary data. Finally, valuable comments by Stef Proost (KU Leuven) on an earlier version are gratefully acknowledged.
Notes
1. Flanders (also called the Flemish Region) is one of the three regions that jointly constitute the federal state of Belgium. It has about six million inhabitants.
2. Nelson (Citation2008) states that changes of 3–5 dB are noticeable and that working with changes of 10 dB is excessive. However, when we work with 4-dB changes, there are too few observations within each class. Therefore, we work with noise classes of 9 dB.
4. We have estimated several models in addition to what has been reported in this paper to investigate the impact of this secondary road. The results indicate that most of the highest levels of noise in Aalter arrive from this road. The results can be obtained from the authors upon request.
5. The ambient noise level did not severely change in any manner between 2004 and 2009.
6. This NDSI was calculated by using noise60_70 and noise70_80_aalter
7. Assuming that prices of nearby houses have a stronger impact on the price of a particular house than prices of houses further away, we use the inverse of the distance between houses as spatial as input for our spatial weight matrices. As we only want to measure the price impact of house i on house j, if house i is sold before house j, the element wij of the weight matrix is set to zero when the selling date of i is after the selling date of j when testing for spatial lag dependency. When testing for spatial error dependency, on the other hand, the selling date is less relevant, so a symmetrical matrix containing the inverse distances is used.
8. These results of the spatial error models for Aalter will be made available by the authors upon simple request.