Efficient parametric estimation of the ...
Document type :
Pré-publication ou Document de travail
Permalink :
Title :
Efficient parametric estimation of the spatial autoregressive model
Author(s) :
Debarsy, Nicolas [Auteur]
Lille économie management - UMR 9221 [LEM]
Verardi, Vincenzo [Auteur]
Vermandele, Catherine [Auteur]

Lille économie management - UMR 9221 [LEM]
Verardi, Vincenzo [Auteur]
Vermandele, Catherine [Auteur]
Publication date :
2025-03-11
English keyword(s) :
spillovers
Local Asymptotic Normality
Flexible distributions
efficiency
Local Asymptotic Normality
Flexible distributions
efficiency
HAL domain(s) :
Économie et finance quantitative [q-fin]
English abstract : [en]
This paper introduces a new one-step parametric estimation method for spatial autoregressive (SAR) models, providing an efficient estimator for any error distribution with a defined quantile function. Based on Le Cam's ...
Show more >This paper introduces a new one-step parametric estimation method for spatial autoregressive (SAR) models, providing an efficient estimator for any error distribution with a defined quantile function. Based on Le Cam's Local Asymptotic Normality (LAN) theory, it extends the maximum likelihood approach to cases like the Laplace distribution, which lacks a globally defined first derivative.\We further develop this estimator for two highly flexible distributions: Tukey's gand-h and Pewsey and Jones's sinh-arcsinh (SAS), designed to capture skewness and non-normal tail weight. These flexible distributions mitigate the risks of distributional misspecification by approximating a wide range of parametric distributions. Monte Carlo simulations assess finite-sample performance, showing that our estimator outperforms traditional parametric spatial methods when the error distribution deviates from normality and is well-approximated by these flexible alternatives.Show less >
Show more >This paper introduces a new one-step parametric estimation method for spatial autoregressive (SAR) models, providing an efficient estimator for any error distribution with a defined quantile function. Based on Le Cam's Local Asymptotic Normality (LAN) theory, it extends the maximum likelihood approach to cases like the Laplace distribution, which lacks a globally defined first derivative.\We further develop this estimator for two highly flexible distributions: Tukey's gand-h and Pewsey and Jones's sinh-arcsinh (SAS), designed to capture skewness and non-normal tail weight. These flexible distributions mitigate the risks of distributional misspecification by approximating a wide range of parametric distributions. Monte Carlo simulations assess finite-sample performance, showing that our estimator outperforms traditional parametric spatial methods when the error distribution deviates from normality and is well-approximated by these flexible alternatives.Show less >
Language :
Anglais
Collections :
Source :
Submission date :
2025-03-15T03:00:41Z
Files
- document
- Open access
- Access the document
- DVV_Para.pdf
- Open access
- Access the document