Font Size: a A A

Spatial Panel Models: Identification, Estimation And Empirical Analysis

Posted on:2017-12-05Degree:MasterType:Thesis
Country:ChinaCandidate:Y BaiFull Text:PDF
GTID:2359330503490269Subject:Quantitative Economics
Abstract/Summary:PDF Full Text Request
Spatial econometrics is a subfield of econometrics that deals with interactions among cross-sectional units. Traditional econometrics assume that units are independently distributed, which is usually wed to Gaussian thinking, making them of limited value in description accuracy. There is a growing literature on spatial econometrics since Anselin(1988) and the models are now widely used in many areas of economics including reginal economics, labor economics, educational economics and finance. The thesis tries to extend the existing literature on three main subjects: identification, estimation and empirical analysis.In the literature review part, I first focus on the question that why spatial econometrics is better than traditional modeling techniques. I give a brief overview of spatial weight matrix and the model specifications of spatial panels. I then introduce both moment based estimation procedures and quasi-maximum likelihood estimation procedures of spatial models. I also introduce the methods of testing for spatial effects.For the identification, I focus on the Wald test of COMFAC restrictions on spatial durbin panel models. I find that Wald test has good power but large size distortion in finite samples. Then I introduce a residual based bootstrap procedure to correct the size distortion and Monte Carlo results show that test based on bootstrap works much better.For empirical works, I analyze the spatial effects on carbon dioxide emissions based on spatial panel models. The empirical results show that spatial error panel models is best to fit the model and the structure of energy consumptions plays the key role in CO2 emissions and regional cooperation is also important in CO2 emissions reduction.For the estimation, I consider a short dynamic panel data model with spatial errors. Following Su and Yang(2015), I introduce a three-step system GMM estimators for the model and design Monte Carlo experiments to verify the small sample performance. I find that SGMM and QMLE works better than the other in different situations, however, QMLE works much better in more general cases due to the use of initial observations.
Keywords/Search Tags:Spatial Panel Models, Residual based Bootstrap Method, Carbon Dioxide Emissions per Capita, Generalized Method of Moments Estimation, Quasi-Maximum Likelihood Estimation
PDF Full Text Request
Related items