Spatial Econometrics

Similar documents
Spatial Econometrics. Wykªad 6: Multi-source spatial models. Andrzej Torój. Institute of Econometrics Department of Applied Econometrics

Outline. Overview of Issues. Spatial Regression. Luc Anselin

Lecture 6: Hypothesis Testing

Lecture 4: Heteroskedasticity

Spatial Regression Models: Identification strategy using STATA TATIANE MENEZES PIMES/UFPE

Spatial Regression. 9. Specification Tests (1) Luc Anselin. Copyright 2017 by Luc Anselin, All Rights Reserved

Econometrics 2, Class 1

Econometrics. Week 4. Fall Institute of Economic Studies Faculty of Social Sciences Charles University in Prague

Simultaneous Equation Models Learning Objectives Introduction Introduction (2) Introduction (3) Solving the Model structural equations

Empirical Economic Research, Part II

Lecture 1: OLS derivations and inference

Spatial Regression. 11. Spatial Two Stage Least Squares. Luc Anselin. Copyright 2017 by Luc Anselin, All Rights Reserved

Proceedings of the 8th WSEAS International Conference on APPLIED MATHEMATICS, Tenerife, Spain, December 16-18, 2005 (pp )

splm: econometric analysis of spatial panel data

Repeated observations on the same cross-section of individual units. Important advantages relative to pure cross-section data

The Simple Regression Model. Part II. The Simple Regression Model

Econometrics I KS. Module 2: Multivariate Linear Regression. Alexander Ahammer. This version: April 16, 2018

Chapter 6. Panel Data. Joan Llull. Quantitative Statistical Methods II Barcelona GSE

Econometrics of Panel Data

Applied Econometrics (MSc.) Lecture 3 Instrumental Variables

Econometric Methods. Prediction / Violation of A-Assumptions. Burcu Erdogan. Universität Trier WS 2011/2012

Econometrics of Panel Data

Advanced Econometrics I

Statistics 910, #5 1. Regression Methods

Econometrics of Panel Data

ON THE NEGATION OF THE UNIFORMITY OF SPACE RESEARCH ANNOUNCEMENT

1. You have data on years of work experience, EXPER, its square, EXPER2, years of education, EDUC, and the log of hourly wages, LWAGE

Econometrics. Week 8. Fall Institute of Economic Studies Faculty of Social Sciences Charles University in Prague

Multivariate Regression Analysis

Testing Random Effects in Two-Way Spatial Panel Data Models

The regression model with one fixed regressor cont d

5. Erroneous Selection of Exogenous Variables (Violation of Assumption #A1)

The Linear Regression Model

Regression with time series

Økonomisk Kandidateksamen 2004 (I) Econometrics 2. Rettevejledning

In the bivariate regression model, the original parameterization is. Y i = β 1 + β 2 X2 + β 2 X2. + β 2 (X 2i X 2 ) + ε i (2)

F9 F10: Autocorrelation

Spatial Econometrics

ECON Introductory Econometrics. Lecture 16: Instrumental variables

Econometrics. 7) Endogeneity

Intermediate Econometrics

Reliability of inference (1 of 2 lectures)

Econometrics - 30C00200

Spatial Regression. 14. Spatial Panels (2) Luc Anselin. Copyright 2017 by Luc Anselin, All Rights Reserved

Robust Standard Errors to spatial and time dependence in. state-year panels. Lucciano Villacorta Gonzales. June 24, Abstract

The Multiple Regression Model Estimation

Linear models. Linear models are computationally convenient and remain widely used in. applied econometric research

Controlling for Time Invariant Heterogeneity

Short T Panels - Review

LECTURE 11. Introduction to Econometrics. Autocorrelation

Econometrics Honor s Exam Review Session. Spring 2012 Eunice Han

Homoskedasticity. Var (u X) = σ 2. (23)

Introduction to Estimation Methods for Time Series models. Lecture 1

Economics 308: Econometrics Professor Moody

Vector autoregressions, VAR

Outline. Nature of the Problem. Nature of the Problem. Basic Econometrics in Transportation. Autocorrelation

Linear Regression. Junhui Qian. October 27, 2014

GMM Estimation of Spatial Error Autocorrelation with and without Heteroskedasticity

Spatial stochastic frontier model

Answers to Problem Set #4

Linear Models in Econometrics

Section 2 NABE ASTEF 65

Econometrics II - EXAM Answer each question in separate sheets in three hours

ECON 4160: Econometrics-Modelling and Systems Estimation Lecture 9: Multiple equation models II

Introductory Econometrics

Econometrics Master in Business and Quantitative Methods

Review of Classical Least Squares. James L. Powell Department of Economics University of California, Berkeley

Model Mis-specification

INTRODUCTION TO BASIC LINEAR REGRESSION MODEL

Economics 536 Lecture 7. Introduction to Specification Testing in Dynamic Econometric Models

Econometrics Summary Algebraic and Statistical Preliminaries

Lecture 3: Spatial Analysis with Stata

Lecture: Simultaneous Equation Model (Wooldridge s Book Chapter 16)

Lecture 2: Spatial Models

Dealing With Endogeneity

Advanced Econometrics

Finite Sample Performance of A Minimum Distance Estimator Under Weak Instruments

Rewrap ECON November 18, () Rewrap ECON 4135 November 18, / 35

Robust Standard Errors to spatial and time dependence when. neither N nor T are very large. Lucciano Villacorta Gonzales. February 15, 2014.

Test of hypotheses with panel data

Topic 10: Panel Data Analysis

Statistics: A review. Why statistics?

Heteroskedasticity. Part VII. Heteroskedasticity

Business Statistics. Tommaso Proietti. Linear Regression. DEF - Università di Roma 'Tor Vergata'

Applied Econometrics (QEM)

Paddy Availability Modeling in Indonesia Using Spatial Regression

Dynamic Regression Models (Lect 15)

ECON 4160, Autumn term Lecture 1

A Guide to Modern Econometric:

Instrumental Variables, Simultaneous and Systems of Equations

PhD/MA Econometrics Examination January 2012 PART A

ECON The Simple Regression Model

Topic 7: Heteroskedasticity

ECONOMETRICS HONOR S EXAM REVIEW SESSION

The regression model with one stochastic regressor (part II)

Applied Statistics and Econometrics

Reading Assignment. Distributed Lag and Autoregressive Models. Chapter 17. Kennedy: Chapters 10 and 13. AREC-ECON 535 Lec G 1

Introductory Econometrics

The exact bias of S 2 in linear panel regressions with spatial autocorrelation SFB 823. Discussion Paper. Christoph Hanck, Walter Krämer

Problem Set #6: OLS. Economics 835: Econometrics. Fall 2012

Transcription:

Spatial Econometrics Lecture 5: Single-source model of spatial regression. Combining GIS and regional analysis (5) Spatial Econometrics 1 / 47

Outline 1 Linear model vs SAR/SLM (Spatial Lag) Linear model SAR (Spatial Lag, SLM) 2 Model SEM (Spatial Error) SEM model with global error dependence SEM model with local error dependence 3 SLX model 4 Combining point GIS data with regional statistics Example: location of Biedronka markets Homework (5) Spatial Econometrics 2 / 47

Plan prezentacji 1 Linear model vs SAR/SLM (Spatial Lag) 2 Model SEM (Spatial Error) 3 SLX model 4 Combining point GIS data with regional statistics (5) Spatial Econometrics 3 / 47

Linear model Linear regression model specication The well-known linear regression model: y = Xβ + ε Its parameters can be estimated in an unbiased, consistent and ecient way via Ordinary Least Squares (OLS) method. Appropriate, when spatial links in y are fully (implicitly) captured through the spatial autocorrelation of regressors included in X (spatial clustering of X). (5) Spatial Econometrics 4 / 47

Linear model Linear regression model specication The well-known linear regression model: y = Xβ + ε Its parameters can be estimated in an unbiased, consistent and ecient way via Ordinary Least Squares (OLS) method. Appropriate, when spatial links in y are fully (implicitly) captured through the spatial autocorrelation of regressors included in X (spatial clustering of X). (5) Spatial Econometrics 4 / 47

Linear model Linear regression model specication The well-known linear regression model: y = Xβ + ε Its parameters can be estimated in an unbiased, consistent and ecient way via Ordinary Least Squares (OLS) method. Appropriate, when spatial links in y are fully (implicitly) captured through the spatial autocorrelation of regressors included in X (spatial clustering of X). (5) Spatial Econometrics 4 / 47

Linear model Flow of impacts in the linear model (5) Spatial Econometrics 5 / 47

SAR (Spatial Lag, SLM) Flow of impacts in SAR model (5) Spatial Econometrics 6 / 47

SAR (Spatial Lag, SLM) SAR model relation to other models (5) Spatial Econometrics 7 / 47

SAR (Spatial Lag, SLM) SAR model relation to other models (5) Spatial Econometrics 8 / 47

SAR (Spatial Lag, SLM) SAR model specication Spatial autoregression with additional regressors. y = ρwy + Xβ + ε Without any explanatory variables X in the model, it would be identical with pure SAR. In this model, we do not assume any spatial clustering of the causes, but spatial interactions in outcomes (spatial global spillovers, spatial spillovers). Problem with OLS estimation: endogeneity (like in pure SAR). (5) Spatial Econometrics 9 / 47

SAR (Spatial Lag, SLM) SAR model specication Spatial autoregression with additional regressors. y = ρwy + Xβ + ε Without any explanatory variables X in the model, it would be identical with pure SAR. In this model, we do not assume any spatial clustering of the causes, but spatial interactions in outcomes (spatial global spillovers, spatial spillovers). Problem with OLS estimation: endogeneity (like in pure SAR). (5) Spatial Econometrics 9 / 47

SAR (Spatial Lag, SLM) SAR model specication Spatial autoregression with additional regressors. y = ρwy + Xβ + ε Without any explanatory variables X in the model, it would be identical with pure SAR. In this model, we do not assume any spatial clustering of the causes, but spatial interactions in outcomes (spatial global spillovers, spatial spillovers). Problem with OLS estimation: endogeneity (like in pure SAR). (5) Spatial Econometrics 9 / 47

SAR (Spatial Lag, SLM) SAR model specication Spatial autoregression with additional regressors. y = ρwy + Xβ + ε Without any explanatory variables X in the model, it would be identical with pure SAR. In this model, we do not assume any spatial clustering of the causes, but spatial interactions in outcomes (spatial global spillovers, spatial spillovers). Problem with OLS estimation: endogeneity (like in pure SAR). (5) Spatial Econometrics 9 / 47

SAR (Spatial Lag, SLM) Consequences of omitting spatial structure SAR (1) True data generating process: y = ρwy + Xβ + ε Estimated linear model omitting Wy (method OLS): y = Xβ KMNK + ε According to the general principles of econometrics, omitting a variable results in the estimation bias of β, that converges to the product of: (true) parameter of the skipped variable slope of the regression of the skipped variable on the included variables In our case: plimˆβ KMNK = β + ρ Cov(Wy,X) Var(X) (5) Spatial Econometrics 10 / 47

SAR (Spatial Lag, SLM) Consequences of omitting spatial structure SAR (1) True data generating process: y = ρwy + Xβ + ε Estimated linear model omitting Wy (method OLS): y = Xβ KMNK + ε According to the general principles of econometrics, omitting a variable results in the estimation bias of β, that converges to the product of: (true) parameter of the skipped variable slope of the regression of the skipped variable on the included variables In our case: plimˆβ KMNK = β + ρ Cov(Wy,X) Var(X) (5) Spatial Econometrics 10 / 47

SAR (Spatial Lag, SLM) Consequences of omitting spatial structure SAR (1) True data generating process: y = ρwy + Xβ + ε Estimated linear model omitting Wy (method OLS): y = Xβ KMNK + ε According to the general principles of econometrics, omitting a variable results in the estimation bias of β, that converges to the product of: (true) parameter of the skipped variable slope of the regression of the skipped variable on the included variables In our case: plimˆβ KMNK = β + ρ Cov(Wy,X) Var(X) (5) Spatial Econometrics 10 / 47

SAR (Spatial Lag, SLM) Consequences of omitting spatial structure SAR (2) Can Cov (Wy, X) possibly be 0? If the true data generating process is SAR, then... y = (I ρw) 1 Xβ + (I ρw) 1 ε y = Xβ + ρwxβ + ρ 2 W 2 Xβ +... + ε + ρwε + ρ 2 W 2 ε +... Wy = WXβ + ρw 2 Xβ + ρ 2 W 3 Xβ +... + Wε + ρw 2 ε + ρ 2 W 3 ε +... Thus (skipping the components related to ε, which as we know are uncorrelated to X): plimˆβ KMNK β = ρ Cov(WXβ,X) + ρ Cov(ρW2 Xβ,X) + ρ Cov(ρ2 W 3 Xβ,X) +... = Var(X) Var(X) Var(X) ρ [ ( = Var(X) Cov (WX, X) + ρ Cov W 2 X, X ) + ρ 2 Cov ( W 3 X, X ) +... ] β Even if X is not spatially autocorrelated and Cov (WX, X) = 0, further components cannot be equal to zero. W 2 and further powers of W are not any more matrices with zero diagonal elements. Interpretation: W 2 is the matrix of connections to neighbours of the neighbours. But the neighbour of your neighbour is i.a. You! (And You're always correlated with yourself.) (5) Spatial Econometrics 11 / 47

SAR (Spatial Lag, SLM) Consequences of omitting spatial structure SAR (2) Can Cov (Wy, X) possibly be 0? If the true data generating process is SAR, then... y = (I ρw) 1 Xβ + (I ρw) 1 ε y = Xβ + ρwxβ + ρ 2 W 2 Xβ +... + ε + ρwε + ρ 2 W 2 ε +... Wy = WXβ + ρw 2 Xβ + ρ 2 W 3 Xβ +... + Wε + ρw 2 ε + ρ 2 W 3 ε +... Thus (skipping the components related to ε, which as we know are uncorrelated to X): plimˆβ KMNK β = ρ Cov(WXβ,X) + ρ Cov(ρW2 Xβ,X) + ρ Cov(ρ2 W 3 Xβ,X) +... = Var(X) Var(X) Var(X) ρ [ ( = Var(X) Cov (WX, X) + ρ Cov W 2 X, X ) + ρ 2 Cov ( W 3 X, X ) +... ] β Even if X is not spatially autocorrelated and Cov (WX, X) = 0, further components cannot be equal to zero. W 2 and further powers of W are not any more matrices with zero diagonal elements. Interpretation: W 2 is the matrix of connections to neighbours of the neighbours. But the neighbour of your neighbour is i.a. You! (And You're always correlated with yourself.) (5) Spatial Econometrics 11 / 47

SAR (Spatial Lag, SLM) Consequences of omitting spatial structure SAR (2) Can Cov (Wy, X) possibly be 0? If the true data generating process is SAR, then... y = (I ρw) 1 Xβ + (I ρw) 1 ε y = Xβ + ρwxβ + ρ 2 W 2 Xβ +... + ε + ρwε + ρ 2 W 2 ε +... Wy = WXβ + ρw 2 Xβ + ρ 2 W 3 Xβ +... + Wε + ρw 2 ε + ρ 2 W 3 ε +... Thus (skipping the components related to ε, which as we know are uncorrelated to X): plimˆβ KMNK β = ρ Cov(WXβ,X) + ρ Cov(ρW2 Xβ,X) + ρ Cov(ρ2 W 3 Xβ,X) +... = Var(X) Var(X) Var(X) ρ [ ( = Var(X) Cov (WX, X) + ρ Cov W 2 X, X ) + ρ 2 Cov ( W 3 X, X ) +... ] β Even if X is not spatially autocorrelated and Cov (WX, X) = 0, further components cannot be equal to zero. W 2 and further powers of W are not any more matrices with zero diagonal elements. Interpretation: W 2 is the matrix of connections to neighbours of the neighbours. But the neighbour of your neighbour is i.a. You! (And You're always correlated with yourself.) (5) Spatial Econometrics 11 / 47

SAR (Spatial Lag, SLM) Spatial OLS (1) If the omission of spatial lag makes the OLS estimator biased, we should include it. Potentially easy to do: if W is predetermined, one can construct the spatial lag variable Wy upfront and estimate the SAR model y = ρwy + Xβ + ε with OLS (this method is referred to as Spatial OLS): y = [ Wy X ] [ ρ β ] + ε From OLS properties, we know that: ([ ]) E = ˆρˆβ [ ] ρ ( [ ] T [ ] ) 1 ( [ ] ) T + Wy X Wy X E Wy X ε β (5) Spatial Econometrics 12 / 47

SAR (Spatial Lag, SLM) Spatial OLS (1) If the omission of spatial lag makes the OLS estimator biased, we should include it. Potentially easy to do: if W is predetermined, one can construct the spatial lag variable Wy upfront and estimate the SAR model y = ρwy + Xβ + ε with OLS (this method is referred to as Spatial OLS): y = [ Wy X ] [ ρ β ] + ε From OLS properties, we know that: ([ ]) E = ˆρˆβ [ ] ρ ( [ ] T [ ] ) 1 ( [ ] ) T + Wy X Wy X E Wy X ε β (5) Spatial Econometrics 12 / 47

SAR (Spatial Lag, SLM) Spatial OLS (1) If the omission of spatial lag makes the OLS estimator biased, we should include it. Potentially easy to do: if W is predetermined, one can construct the spatial lag variable Wy upfront and estimate the SAR model y = ρwy + Xβ + ε with OLS (this method is referred to as Spatial OLS): y = [ Wy X ] [ ρ β ] + ε From OLS properties, we know that: ([ ]) E = ˆρˆβ [ ] ρ ( [ ] T [ ] ) 1 ( [ ] ) T + Wy X Wy X E Wy X ε β (5) Spatial Econometrics 12 / 47

SAR (Spatial Lag, SLM) Spatial OLS (2) In the linear regression model, we( assume that the error terms are [ ] ) T independent of regressors, i.e. E Wy X ε = 0, and this proves the unbiasedness of the OLS estimator in such a model. It holds that E ( X T ε ) = 0, but: [ ] { [W E (Wy) T ε = E (I ρw) 1 Xβ + W (I ρw) 1 ε ] } T ε = { [W = E (I ρw) 1 Xβ ] T [ ε + W (I ρw) 1 ε ] } T ε = = E {ε [ T W (I ρw) 1] } T ε 0 Our model is not the classical regression model, because observations depend on one another (y i depends on the neighbour y j and vice versa). Situation similar to the simultaneous equations models. (5) Spatial Econometrics 13 / 47

SAR (Spatial Lag, SLM) Spatial OLS (2) In the linear regression model, we( assume that the error terms are [ ] ) T independent of regressors, i.e. E Wy X ε = 0, and this proves the unbiasedness of the OLS estimator in such a model. It holds that E ( X T ε ) = 0, but: [ ] { [W E (Wy) T ε = E (I ρw) 1 Xβ + W (I ρw) 1 ε ] } T ε = { [W = E (I ρw) 1 Xβ ] T [ ε + W (I ρw) 1 ε ] } T ε = = E {ε [ T W (I ρw) 1] } T ε 0 Our model is not the classical regression model, because observations depend on one another (y i depends on the neighbour y j and vice versa). Situation similar to the simultaneous equations models. (5) Spatial Econometrics 13 / 47

SAR (Spatial Lag, SLM) Spatial OLS (3) For simplication, consider the SAR model with 1 explanatory variable x: = = ([ ]) E = ˆρˆβ [ ] [ ρ 1 + ( β [ ] det Wy x T [ ] ) Wy x }{{} γ>0 usually>0 [ ] {}}{ ρ γx T x (Wy) T ε + β γx T (Wy) (Wy) T ε }{{} usually<0 x T x x T (Wy) (Wy) T x (Wy) T (Wy) ] (Wy) T ε x T ε }{{} =0 = So, the spatial OLS delivers biased estimates! (ρ usually upward biased, β downward biased). In the multivariate cases, the bias is concentrated on the parameters for variables X whose spatial patterns most resembles the spatial pattern of y. (5) Spatial Econometrics 14 / 47

SAR (Spatial Lag, SLM) Spatial OLS (3) For simplication, consider the SAR model with 1 explanatory variable x: = = ([ ]) E = ˆρˆβ [ ] [ ρ 1 + ( β [ ] det Wy x T [ ] ) Wy x }{{} γ>0 usually>0 [ ] {}}{ ρ γx T x (Wy) T ε + β γx T (Wy) (Wy) T ε }{{} usually<0 x T x x T (Wy) (Wy) T x (Wy) T (Wy) ] (Wy) T ε x T ε }{{} =0 = So, the spatial OLS delivers biased estimates! (ρ usually upward biased, β downward biased). In the multivariate cases, the bias is concentrated on the parameters for variables X whose spatial patterns most resembles the spatial pattern of y. (5) Spatial Econometrics 14 / 47

SAR (Spatial Lag, SLM) Spatial 2SLS (1) The simultaneous equation bias in y = ρwy + Xβ + ε can be treated analogously to the case of endogenous regressors: i.e. use the instrumental variables method. This implementation is consistent, unbiased and is referred to as spatial 2-stage least squares (S2SLS). A valid instrumental variable is correlated with the problematic regressor (Wy), but uncorrelated with the error term (ε). Recall that for the SAR model: Wy = WXβ + ρw 2 Xβ + ρ 2 W 3 Xβ +... + Wε + ρw 2 ε + ρ 2 W 3 ε +... }{{} ideal instruments! Step 1: Linear regression of Wy on the matrix including exogenous variables and a certain number of instruments: Π = [ X WX W 2 X... ] (OLS). ( ) 1 Theoretical values: Ŵy = Π Π T Π Π T Wy }{{} P (5) Spatial Econometrics 15 / 47

SAR (Spatial Lag, SLM) Spatial 2SLS (1) The simultaneous equation bias in y = ρwy + Xβ + ε can be treated analogously to the case of endogenous regressors: i.e. use the instrumental variables method. This implementation is consistent, unbiased and is referred to as spatial 2-stage least squares (S2SLS). A valid instrumental variable is correlated with the problematic regressor (Wy), but uncorrelated with the error term (ε). Recall that for the SAR model: Wy = WXβ + ρw 2 Xβ + ρ 2 W 3 Xβ +... + Wε + ρw 2 ε + ρ 2 W 3 ε +... }{{} ideal instruments! Step 1: Linear regression of Wy on the matrix including exogenous variables and a certain number of instruments: Π = [ X WX W 2 X... ] (OLS). ( ) 1 Theoretical values: Ŵy = Π Π T Π Π T Wy }{{} P (5) Spatial Econometrics 15 / 47

SAR (Spatial Lag, SLM) Spatial 2SLS (1) The simultaneous equation bias in y = ρwy + Xβ + ε can be treated analogously to the case of endogenous regressors: i.e. use the instrumental variables method. This implementation is consistent, unbiased and is referred to as spatial 2-stage least squares (S2SLS). A valid instrumental variable is correlated with the problematic regressor (Wy), but uncorrelated with the error term (ε). Recall that for the SAR model: Wy = WXβ + ρw 2 Xβ + ρ 2 W 3 Xβ +... + Wε + ρw 2 ε + ρ 2 W 3 ε +... }{{} ideal instruments! Step 1: Linear regression of Wy on the matrix including exogenous variables and a certain number of instruments: Π = [ X WX W 2 X... ] (OLS). ( ) 1 Theoretical values: Ŵy = Π Π T Π Π T Wy }{{} P (5) Spatial Econometrics 15 / 47

SAR (Spatial Lag, SLM) Spatial 2SLS (2) Step 2: OLS estimation of the SAR model parameters after the replacement of Wy with [ ] ( Ŵy: [ ] T [ ] ) 1 [ ] T = Ŵy X Ŵy X Ŵy X y ˆρˆβ Spatial 2SLS (S2SLS) model <- stsls(y ~ x, listw = W) (5) Spatial Econometrics 16 / 47

SAR (Spatial Lag, SLM) Spatial ML (1) Variant 2: maximum likelihood method M M {}}{{}}{ y = (I ρw) 1 Xβ + (I ρw) 1 u, u N(0, σ 2 ) L (u) = ( ) N ( ) 1 2 σ 2 exp ut u 2π 2σ 2 By the change of variables theorem (multivariate case): L (y) = det L (y) = M 1 {( }} ){ u L [u (y)] y det ( M 1) ( ( ) ) N 1 2 σ 2 exp (y MXβ)T (M 1 ) T (M 1 )(y MXβ) 2π 2σ 2 ˆβ = arg maxl (y) β (5) Spatial Econometrics 17 / 47

SAR (Spatial Lag, SLM) Spatial ML (2) Standard errors evaluated on the basis of Hessian matrix at the maximum point of the likelihood function (typical for ML). If M = I, the likelihood function identical as in the linear model. ML for the SAR model in R model <- lagsarlm(y ~ x, listw = W) The same model is estimated when the formula argument in the function spautolm (pure SAR) is supplied with additional regressors. (5) Spatial Econometrics 18 / 47

SAR (Spatial Lag, SLM) Tests: linear model vs SAR (1) This illustration demonstrates the univariate case (θ scalar). (5) Spatial Econometrics 19 / 47

SAR (Spatial Lag, SLM) Tests: linear model vs SAR (2) LM ρ = N tr[(w T +W)W]+ 1 (WXˆβ) T ˆε T [I X(X T X)X T ](WXˆβ) ˆε χ 2 (1) ( ) ˆε T 2 Wy ˆε T ˆε H 0 : linear model (ρ = 0) H 1 : SAR (5) Spatial Econometrics 20 / 47

Plan prezentacji 1 Linear model vs SAR/SLM (Spatial Lag) 2 Model SEM (Spatial Error) 3 SLX model 4 Combining point GIS data with regional statistics (5) Spatial Econometrics 21 / 47

SEM model with global error dependence Flow of impacts in SEM model (5) Spatial Econometrics 22 / 47

SEM model with global error dependence SEM model relation to other models (5) Spatial Econometrics 23 / 47

SEM model with global error dependence SEM model relation to other models (5) Spatial Econometrics 24 / 47

SEM model with global error dependence SEM model specication It is not the dependent variable, but the error term, that is subject to spatial autocorrelation the dierence is analogous to the dierence between AR and MA models. y = Xβ + ε ε = λwε + u In the absence of regressors X, the model would be equivalent to (pure) SAR. Spatial clustering in unobservables (shocks). (5) Spatial Econometrics 25 / 47

SEM model with global error dependence SEM model specication It is not the dependent variable, but the error term, that is subject to spatial autocorrelation the dierence is analogous to the dierence between AR and MA models. y = Xβ + ε ε = λwε + u In the absence of regressors X, the model would be equivalent to (pure) SAR. Spatial clustering in unobservables (shocks). (5) Spatial Econometrics 25 / 47

SEM model with global error dependence SEM model specication It is not the dependent variable, but the error term, that is subject to spatial autocorrelation the dierence is analogous to the dierence between AR and MA models. y = Xβ + ε ε = λwε + u In the absence of regressors X, the model would be equivalent to (pure) SAR. Spatial clustering in unobservables (shocks). (5) Spatial Econometrics 25 / 47

SEM model with global error dependence SEM model estimation (1) OLS estimator is inecient (and the standard errors biased), because: y = Xβ + ε ε = λwε + u, czyli ε = (I λw) 1 u Var (ε) = E ( εε T ) = (I λw) 1 E ( uu T ) [ (I λw) 1] T = σ 2 (I λw) 1 [ (I λw) 1] T σ 2 I Variant 1: as usually with non-spherical errors, the solution is Generalised Least Squares estimation: ˆβ = ( X T Ω 1 X ) 1 X T Ω 1 y with given Ω = (I λw) 1 [ (I λw) 1] T W known, λ estimated based on errors derived from the consistent OLS estimation (details of the procedure: Kelejian and Prucha, 1998; Arbia, 2014). ) Var (ˆβ = ˆσ ( 2 X T Ω 1 X ) 1 (5) Spatial Econometrics 26 / 47

SEM model with global error dependence SEM model estimation (2) Spatial GLS in R model4 <- GMerrorsar(y ~ x, listw = W) (5) Spatial Econometrics 27 / 47

SEM model with global error dependence SEM model estimation (3) Variant 2: maximum likelihood method M {}}{ y = Xβ + (I λw) 1 u, u N(0, σ 2 ) L (u) = ( ) N ( ) 1 2 σ 2 exp ut u 2π 2σ 2 By the change of variables theorem (multivariate case): L (y) = det L (y) = M 1 {( }} ){ u L [u (y)] y det ( M 1) ( ( ) ) N 1 2 σ 2 exp (y Xβ)T (M 1 ) T (M 1 )(y Xβ) 2π 2σ 2 ˆβ = arg maxl (y) β (5) Spatial Econometrics 28 / 47

SEM model with global error dependence SEM model estimation (4) Standard errors evaluated on the basis of Hessian matrix at the maximum point of the likelihood function (typical for ML). If M = I, the likelihood function identical as in the linear model. ML for SEM model in R model <- errorsarlm(y ~ x, listw = W) The same model will also be estimated, if the formula in the function spautolm (pure SAR) supplied with regressors. (5) Spatial Econometrics 29 / 47

SEM model with global error dependence Both SAR and SEM collapse to pure SAR without regressors X SAR y = ρwy + Xβ + ε y = ρwy + ε y ρwy = ε (I ρw) y = ε y = (I ρw) 1 ε SEM y = Xβ + (I λw) 1 u β = 0 y = (I λw) 1 u (5) Spatial Econometrics 30 / 47

SEM model with global error dependence LM tests: linear model vs SEM LM λ = N 2 tr[(w T +W)W] H 0 : linear model (λ = 0) H 1 : SEM ( ) 2 ût Wû û T û χ2 (1) (5) Spatial Econometrics 31 / 47

SEM model with global error dependence Robust LM tests (1) In LM thests for SAR and SEM specications (respectively): 1 H 0 : linear model (ρ = 0), H 1 : SAR 2 H 0 : linear model (λ = 0), H 1 : SEM Problem: each pair of hypotheses leaves out of sight the alternative hypothesis from the other pair of the other test. Consequence: test 1 rejects H 0 even under false H 1 (but true H 1 from test 2). And vice versa. RLMlag and RLMerr Anselin et al. (1996) propose robust test statistics LMρ and LMλ, which by construction exclude the possibility that an incorrect process is captured by the alternative hypothesis (see Arbia, 2014). LMρ = LM ρλ LM λ LMλ = LM ρλ LM ρ (5) Spatial Econometrics 32 / 47

SEM model with local error dependence Global vs local SEM model (1) The previously presented SEM model stipulated a global dependence between unobservables: The local SEM version: y = Xβ + ε ε = λwε + u y = Xβ + ε ε = λwu + u What is the dierence? Consider spatial multiplier matrices of y with respect to u in both cases: local SEM: y = Xβ + ε, ε = (I + λw)u M = y u = (I + λw) global SEM: y = Xβ + ε, ε = (I λw) 1 u M = y u = (I λw) 1 Algebraically, note that: multiplier SEM glob {}}{ (I λw) 1 = multiplier SEM loc {}}{ I + λw + λ 2 W 2 + λ 3 W 3 +... (5) Spatial Econometrics 33 / 47

SEM model with local error dependence Global vs local SEM model (2) Example: Canada, USA, Mexico; W = λ = 0.4; shock u = 1 occurs in Mexico. Spatial multiplierrs for local SEM: 0 0.5 0.5 0 I + 0.4 1 0 0 0 = 1 0 0 1 1 0.2 0.2 0.4 1 0 0.4 0 1 0 0 1 = 0.2 0 1 US CA 0 0.5 MX 0.5 1 0 0 1 0 0 y MX = 1, y US = 0.2, no eect for Canada. Shock in u aected y in the directly linked units. ; (5) Spatial Econometrics 34 / 47

SEM model with local error dependence Global vs local SEM model (3) Spatial multipliers for global SEM: 1 0 0.5 0.5 0 I 0.4 1 0 0 0 1 0 0 1 1 0.2 0.2 0.4 1 0 0.4 0 1 1.19 0.24 0.24 0.48 1.10 0.10 0.48 0.10 1.10 1 0 0 1 0.24 0.10 1.10 = y MX > 1, y US > 0.2, there is (weak but positive) eect for Canada The impulse spills over to the related units, and then to their own related units, etc. (including the feedback into the impulse region). (5) Spatial Econometrics 35 / 47

Plan prezentacji 1 Linear model vs SAR/SLM (Spatial Lag) 2 Model SEM (Spatial Error) 3 SLX model 4 Combining point GIS data with regional statistics (5) Spatial Econometrics 36 / 47

SLX model Flow of impacts in the SLX model (5) Spatial Econometrics 37 / 47

SLX model SLX model relation to other models (5) Spatial Econometrics 38 / 47

SLX model SLX model relation to other models (5) Spatial Econometrics 39 / 47

SLX model SLX model specication Direct impact of causes in the neighbourhood on the consequence in the observed region spatial spillovers:: y = Xβ + WXθ + ε Consistent, ecient and unbiased estimation with OLS. (5) Spatial Econometrics 40 / 47

SLX model SLX model specication Direct impact of causes in the neighbourhood on the consequence in the observed region spatial spillovers:: y = Xβ + WXθ + ε Consistent, ecient and unbiased estimation with OLS. (5) Spatial Econometrics 40 / 47

Plan prezentacji 1 Linear model vs SAR/SLM (Spatial Lag) 2 Model SEM (Spatial Error) 3 SLX model 4 Combining point GIS data with regional statistics (5) Spatial Econometrics 41 / 47

Example: location of Biedronka markets GIS data about the markets Biedronka Source: poiplaza.com POI: points of interest (usually published for the users of car GPS navigation sets) Point data about location of individual markets in Poland. Longitude and latitude. Question: what regional criteria do the managers / owners of Biedronka use when locating their markets? In other words, is there a relationship between the number of Biedronka markets (per capita) and local socio-economic characteristics from Local Data Bank, e.g. on the level of poviats? (5) Spatial Econometrics 42 / 47

Example: location of Biedronka markets GIS data about the markets Biedronka Source: poiplaza.com POI: points of interest (usually published for the users of car GPS navigation sets) Point data about location of individual markets in Poland. Longitude and latitude. Question: what regional criteria do the managers / owners of Biedronka use when locating their markets? In other words, is there a relationship between the number of Biedronka markets (per capita) and local socio-economic characteristics from Local Data Bank, e.g. on the level of poviats? (5) Spatial Econometrics 42 / 47

Example: location of Biedronka markets GIS data about the markets Biedronka Source: poiplaza.com POI: points of interest (usually published for the users of car GPS navigation sets) Point data about location of individual markets in Poland. Longitude and latitude. Question: what regional criteria do the managers / owners of Biedronka use when locating their markets? In other words, is there a relationship between the number of Biedronka markets (per capita) and local socio-economic characteristics from Local Data Bank, e.g. on the level of poviats? (5) Spatial Econometrics 42 / 47

Example: location of Biedronka markets Aggregation of points on the predened map (5) Spatial Econometrics 43 / 47

Example: location of Biedronka markets Aggregation of points on the predened map Andrzej Function Torój ClassIntervals watch outinstitute the style of Econometrics parameter. Department So far, weofhave Applied been Econometrics (5) Spatial using Econometrics quantile division into classes of equal count for the purpose of presentation. 44 / 47

Example: location of Biedronka markets Estimates of 3 models Fit the following regression models of the number of Biedronka markets per 10 thousand residents on labour market statistics (unemployment, wages): linear SLM SEM SLX Compare the models as regards the signicance of variables, AIC criterion, log-likelihood value at maximum and the presence of unremoved spatial autocorrelation. Which model is the best? (5) Spatial Econometrics 45 / 47

Homework Exercise Derive the likelihood function for SEM model with local error dependence. Develop an R code for the estimation of such a model. (5) Spatial Econometrics 46 / 47

Homework Homework 5 Build a set of potential regressors (X) and merge it with the spatial dataset used in homework 1 (y). Using the spatial weight matrix W constructed in homework 2 estimate the pure SAR model for the variable from homework 1. Is this result in line with the testing results from homework 3? Estimate models SAR, SLX, SEM for the considered X, y and W. Visualise the residuals from these models on a map and test them for the presence of spatial autocorrelation in residuals. (5) Spatial Econometrics 47 / 47