Market Potential, MAUP, NUTS and other spatial mysteries Fernando Bruna Jesus Lopez-Rodriguez Andres Faina 11th International Workshop Spatial Econometrics.

Slides:



Advertisements
Similar presentations
Autocorrelation and Heteroskedasticity
Advertisements

Introduction Describe what panel data is and the reasons for using it in this format Assess the importance of fixed and random effects Examine the Hausman.
Dynamic panels and unit roots
Hedonic Modeling Mats Wilhelmsson Center for Banking and Finance (Cefin)
Econometric Analysis of Panel Data Panel Data Analysis – Random Effects Assumptions GLS Estimator Panel-Robust Variance-Covariance Matrix ML Estimator.
Applied Econometrics Second edition
Econometric Analysis of Panel Data Panel Data Analysis: Extension –Generalized Random Effects Model Seemingly Unrelated Regression –Cross Section Correlation.
PANEL DATA 1. Dummy Variable Regression 2. LSDV Estimator
Panel Data Models Prepared by Vera Tabakova, East Carolina University.
Data organization.
Yvonne Wolfmayr with Martin Falk Services and materials outsourcing to low-wage countries and employment: Empirical evidence from EU countries WORKS Expert.
11 Pre-conference Training MCH Epidemiology – CityMatCH Joint 2012 Annual Meeting Intermediate/Advanced Spatial Analysis Techniques for the Analysis of.
What Explains Germany’s Rebounding Export Market Share Stephan Danninger (IMF Research Department) Fred Joutz (George Washington University) September.
Introduction to Applied Spatial Econometrics Attila Varga DIMETIC Pécs, July 3, 2009.
GIS and Spatial Statistics: Methods and Applications in Public Health
19 th Advanced Summer School in Regional Science GIS and spatial econometrics University of Groningen, 4-12 July 2006 “Income and human capital inequalities.
Is the Financial Safety Net a Barrier to Cross-Border Banking? Ata Can Bertay (Tilburg University and World Bank) Asli Demirgüç-Kunt (World Bank) Harry.
Sustainability of economic growth and inequality in incomes distribution Assistant, PhD, BURZ R ă zvan-Dorin West University of Timisoara, Romania Lecturer,
Comments on: Does Financial Structure Matter for Poverty? Evidence from Developing Countries by Kangni Kpodar and Raju Jan Singh L. Colin Xu World Bank.
Chapter 15 Panel Data Analysis.
Economics 20 - Prof. Anderson1 Summary and Conclusions Carrying Out an Empirical Project.
Tse-Chuan Yang, Ph.D The Geographic Information Analysis Core Population Research Institute Social Science Research Institute Pennsylvania State University.
The Paradigm of Econometrics Based on Greene’s Note 1.
IS415 Geospatial Analytics for Business Intelligence
Empirical Example Walter Sosa Escudero Universidad de San Andres - UNLP.
Spatial Econometric Analysis Using GAUSS 4 Kuan-Pin Lin Portland State University.
[ 1 ] MIGRATION AND PRODUCTIVITY. LESSONS FROM THE UK-SPAIN EXPERIENCES This project is funded by the European Commission, Research Directorate General.
Regression Method.
JDS Special program: Pre-training1 Carrying out an Empirical Project Empirical Analysis & Style Hint.
21/09/2015 Wages and accessibility: the impact of transport infrastructure Anna Matas Josep LLuis Raymond Josep LLuis Roig Universitat Autònoma de Barcelona.
Measuring Sovereign Contagion in Europe Presented by Jingjing XIA Caporin, Pelizzon, Ravazzolo, and Rigobon (2013)
ERES 2015 | Main Sessions A Hedonical Spatial Office Rent Index An Application for Madrid Market Ramiro J. Rodríguez A presentation for ERES.
Spatial Econometric Analysis Using GAUSS 1 Kuan-Pin Lin Portland State University.
Spatial Statistics in Ecology: Area Data Lecture Four.
Determinants of Credit Default Swap Spread: Evidence from the Japanese Credit Derivative Market.
Spatial and non spatial approaches to agricultural convergence in Europe Luciano Gutierrez*, Maria Sassi** *University of Sassari **University of Pavia.
THE IMPACT OF INTERNATIONAL OUTSOURCING ON EMPLOYMENT: EMPIRICAL EVIDENCE FROM EU COUNTRIES Martin Falk and Yvonne Wolfmayr Austrian Institute of Economic.
1 Prof. Dr. Rainer Stachuletz Summary and Conclusions Carrying Out an Empirical Project.
Remittances and competitiveness: Evidence for Latin America Migration and Development Thematic Group Seminar Humberto Lopez November 26, 2006 Presentation.
Panel Data Analysis Using GAUSS
CMSSE Summer School Dots to boxes: Do the size and shape of spatial units jeopardize economic geography estimations? A.Briant, P.-P. Combes, M. Lafourcade.
Statistical methods for real estate data prof. RNDr. Beáta Stehlíková, CSc
ERSA Summer school Groningen 2006 Convergence among regions of the enlarged Europe: impact of spatial effects Nicolas DebarsyCem Ertur.
Lecture 1 Introduction to econometrics
Hugo Storm and Thomas Heckelei Institute for Food and Resource Economics (ILR), University of Bonn 150th EAAE Seminar “The spatial dimension in analysing.
11.1 Heteroskedasticity: Nature and Detection Aims and Learning Objectives By the end of this session students should be able to: Explain the nature.
Experimental Evaluations Methods of Economic Investigation Lecture 4.
Reasons for Instability in Spatial Dependence Models Jesús Mur(*), Fernando López (**) and Ana Angulo(*) (*) Department of Economic Analysis University.
Adam Storeygard, Tufts University
Vera Tabakova, East Carolina University
Luciano Gutierrez*, Maria Sassi**
Vera Tabakova, East Carolina University
Spatial Econometric Analysis Using GAUSS
Spatial Modeling Lee Rivers Mobley, Ph.D..
Spatial spillovers and innovation activity in European regions
Large and Small Sample Properties of the MESS Specification
Kakhramon Yusupov June 15th, :30pm – 3:00pm Session 3
PANEL DATA 1. Dummy Variable Regression 2. LSDV Estimator
Econometric methods of analysis and forecasting of financial markets
For the World Economy Availability of business services and outward investment: Evidence from French firms Holger Görg Kiel Institute for the World Economy,
Fundamentals of regression analysis 2
STOCHASTIC REGRESSORS AND THE METHOD OF INSTRUMENTAL VARIABLES
Chapter 15 Panel Data Analysis.
The value of public research: a time series perspective
Sven Blank (University of Tübingen)
Spatial Econometric Analysis
Spatial Econometric Analysis
SPATIAL ANALYSIS IN MACROECOLOGY
5/5/2019 Financial dependence and industry growth in Europe: Better banks and higher productivity Robert Inklaar and Michael Koetter University of Groningen.
Innovation and Employment: Evidence from Italian Microdata
Presentation transcript:

Market Potential, MAUP, NUTS and other spatial mysteries Fernando Bruna Jesus Lopez-Rodriguez Andres Faina 11th International Workshop Spatial Econometrics and Statistics November 2012 Avignon – France

Motivation

ECONOMIC REASONS The relative power of the various economic agglomerating and spreading forces are not scale-neutral but heterogeneous. Different economic forces (theories) are active at different spatial scales => Analyses at different scales provide different insights: the MAUP is only a problem when it is not recognized (ESPON, 2006). STATISTICAL REASONS – The two sides of MAUP: Scale effect (ecological fallacy): for a given space, results can depend on the number of units representing it. Zoning (or aggregation ) effect: for a given scale, results can depend on how the study area is divided up. Motivation: reasons for the MAUP

Is a general form of the wage equation robust to different aggregation levels of European data and different non spatial econometric specifications?: –Long-term relationships: cross-section (variables in levels) –Short-term relationships: Panel data with fixed effects (growth rates) Is the MAUP affecting the estimation of these relationships with spatial econometric models? –SEM and SAR How does the sample selection affect the results? –Broad sample: 25 countries (260 NUTS 2 regions) –Restricted sample: 15 countries (206 NUTS 2 regions) Software – R packages: "spdep" (Bivand 2012); "plm" (Croissant and Millo, 2008) and "splm" (Millo and Piras, 2012). Amelia II (Honaker et al., 2011). Motivation: empirical questions

New Economic Geography: The wage equation

Specifications: variables and notation

Spatial distribution of the variables

Market Potential (lagged one year) is meaningful but its presence does not alter dramatically the results. Residuals are spatially autocorrelated for NUTS 1 and 2: a positive spatial autocorrelation tends to increase with the disaggregation level Pooled estimations with time dummies: broad sample OLS

And the winner is… the SEM model! => OLS estimates are not efficient Particular cases: Contradiction Morans I-LM tests for NUTS 0 in the restricted sample Both robuts tests are highly significant in some cases: thought the decision rule choses the SEM, caution with misspecification. Lagrange Multiplier tests for spatial dependence In the pooled OLS estimations with time dummies and lagged Market Potential Broad sampleRestricted sample

SEM: one year cross-section (1) and pooling with time effects (2) Broad sample Broad sample Restricted sample ML estimation

SAR: one year cross-section (1) and pooling with time effects (2) Broad sample Restricted sample Broad sample ML estimation

Pooled (1) and fixed effects (2) estimations with time effects Broad sample Restricted sample Broad sample OLS

SEM: Pooled (1) and fixed effects (2) estimations with time effect Broad sample Restricted sample Broad sample ML

Broad sample Restricted sample Broad sample SAR: Pooled (1) and fixed effects (2) estimations with time effect ML

With the exception of the fixed effects estimation in the restricted sample,, residuals are autocorrelated and their autocorrelation and estimated spatial parameters increase with disaggregation. The general wage equation is very robust to the short-and-long-run specifications, to this three NUTS levels and to the broad and the restricted sample. Many test of the wage equation in the literature do not distinguish the short-and-long-run specifications. But the estimation with individual effects give a whole different view (Acemoglu et al., 2008). Preliminary conclusions

Results from NUTS 1 and 2: the estimated elasticities are very robust for the non spatial and the SEM and SAR models (FE non checked) => No problem with MAUP (but we have not studied NUTS 3!). Results from NUTS 0 are more sensitive to sample selection. Maybe higher heterogeneity than when pooling regions from different countries at NUTS 1-3. Some of the detected patterns in the change of estimates by NUT level are economically meaningful: at least from NUTS 1 to NUTS 2 the elasticity to Market Potential always increases => More severe problems if this variable is omitted at higher levels of disaggregation. Preliminary conclusions

Sensitivity analysis (at least in the pooled model): –Kelejian and Pruchas (1998) instrumentation of the spatially lagged dependent variable in the SAR model –spatial heteroskedasticity and autocorrelation consistent (HAC) estimators –A graphical W instead of a matrix of the 5 nearest neighbours - but LeSage and Pace (2012)!- – Now annual data: Short-run models for several years panels GWR ( conditional parametric approach) – local variation of estimates: At each NUTS level, what countries are de drivers of the fixed estimates? The zoning effect internal to each MAUP – The areas by country at each NUTS level: Does size matters? –Weighted regression –Recalculate Market Potential: with distances among centroids, bigger regions are further apart from their markets Current research and possible extensions

Results change more using NUTS 0: thoughts welcomed. Similar elasticities in the not spatial and in the SEM and SAR models in spite of being a simple equation. Thoughts: Is this because the SAR was not recommended by the LM tests?. So much effort with spatial models for this?.... Endogeneity – Proper instruments for Market Potential. Endogeneity – In the SAR model both market potential and the endogenous spatial lag of the dependent variable are endogenous: How to deal with this? Which would be the best W matrix to compare models using data with different aggregation? Results of the pooled estimation different when using spdep or splm R packages: why? Questions

COMMENTS WELCOMED THANK YOU Fernando Bruna University of A Coruña, Spain