Publication :
Decomposing and interpreting spatial effects in spatio-temporal analysis : evidences for spatial data pooled over time

En cours de chargement...
Vignette d'image
Date
2018-01-01
Auteurs
Legros, Diégo
Direction de publication
Direction de recherche
Titre de la revue
ISSN de la revue
Titre du volume
Éditeur
Springer
Projets de recherche
Structures organisationnelles
Numéro de revue
Résumé
Empirical applications using individual spatial data pooled over time usually neglect the fact that such data are not only spatially localized: they are also collected over time, i.e. temporally localized. So far, little effort has been devoted to proposing a global way for dealing with spatial data (cross-section) pooled over time, such as real estate transactions, business start-up, crime and so on. However, the spatial effect, in such a context, can be decomposed in two different components: a multidirectional spatial effect (same time period) and a unidirectional spatial effect (previous time period). Based on real estate literature, this chapter presents different spatio-temporal autoregressive (STAR) models and shows how spatial econometrics models can be extended for empirical investigation. Using a Monte Carlo experiment, we underline the effect of neglecting the decomposition of the spatial effect on the bias of the autoregressive coefficients as well as on the interpretation of the marginal effect. An empirical experiment using apartment sales in Paris between 1990 and 2003 supports the global results obtained through the Monte Carlo experiment.
Description
Revue
DOI
10.1007/978-3-319-59511-5_19
URL vers la version publiée
Mots-clés
Spatio-temporal data , Weights matrix , Spatial econometric , Real estate , Data generating process (DGP)
Citation
Type de document
chapitre d'ouvrage