Econometría espacial con aplicaciones Profesora: Coro Chasco Yrigoyen Univeridad Autónoma de Madrid Lugar: Universidad Politécnica de Barcelona 12-13, 18-20 de junio, 2007 ©2007, Coro Chasco Yrigoyen All Rights Reserved Overview and Goals 1. 2. 3. Overview: 21 hours Intensive Course lectures + PC training sessions Goals: Sound understanding of basic and more advanced principles of spatial econometrics Offer tools for practical application of the methodology Commonly available software products used in the GIS framework (SpaceStat and GeoDa) will be introduced and practised in the PC training sessions. @ 2007, Coro Chasco Yrigoyen All Rights Reserved 2 Course Index S1: Introduction to spatial econometrics S2: Spatial effects, spatial dependence S3: Spatial autocorrelation tests S4: Exploratory Spatial Data Analysis (ESDA) S5: Specification of spatial dependence models S6: Spatial regression models: OLS estimation and testing PS1: GeoDa: introduction and ESDA S7: Spatial dependence models: estimation and testing S8: Modelling strategies in spatial regression models PS2: SpaceStat: confirmatory spatial data analysis S9: Specification of spatial heterogeneity models S10: Spatial heterogeneity models: estimation and testing PS3: Practical exercise and evaluation @©2003, 2007, Coro Chasco Yrigoyen All Rights Reserved All Rights Reserved 3 Session 1: Introduction to Spatial Econometrics Fundamentals of spatial econometrics Spatial econometrics history Spatial data problems Some applications in spatial econometrics @ 2007, Coro Chasco Yrigoyen All Rights Reserved 4 Session 1 1.1. Fundamentals of spatial econometrics (I) Subfield of classical econometrics Spatial interaction = autocorrelation Spatial instability (heteroskedasticity, regimes) = heterogeneity Spatial effects: autocorrelation & heterogeneity Spatial effects do affect regression models Spatial econometrics ≈ spatial statistics Regional & urban economics Physical phenomena: biology & geology @ 2007, Coro Chasco Yrigoyen All Rights Reserved 5 Session 1 1.1. Fundamentals of spatial econometrics (II) Spatial autocorrelation Spatial heterogeneity Core @ 2007, Coro Chasco Yrigoyen All Rights Reserved 6 Session 1 1.2. Spatial econometrics history Early 70s: Paelink coins the term “spatial econometrics” 1979: Paelinck & Klaasen book - “Spatial econometrics” 1975/1981: Cliff and Ord’s books 1988: Anselin’s book- “Spatial econometrics: Methods and models” 1991/1995/1998/2001: Anselin – SpaceStat versions 1995: Anselin & Florax (eds) – “New directions in spatial econometrics” First 2000s : LeSage’s Spatial Econometrics Toolbox (Matlab), Bivand’s spdep (R) y Rey’s STARS (Python) 2004: Anselin, Florax & Rey (eds) – “Advances in spatial econometrics” 2003: Anselin’s GeoDa and PySAL 2006: Arbia – “Spatial econometrics” @ 2007, Coro Chasco Yrigoyen All Rights Reserved 7 Session 1 1.3. Spatial data Spatial data nature Spatial data problems Spatial data empirical applications @ 2007, Coro Chasco Yrigoyen All Rights Reserved 8 Session 1 1.3.1. Spatial data nature Time: Time continous and unidimensional Past (t-1) Space: Space continuous and bidimensional Northwest West Southwest North i South Northeast Present (t) East Future (t+1) Southeast @ 2007, Coro Chasco Yrigoyen All Rights Reserved 9 HAINING, R. (1995), “Data problems in spatial econometric modeling”. In L. Anselin and R. Florax Session 1 (ed.), “New directions in spatial econometrics”. Springer-Verlag, Berlin; pp. 156-171. 1.3.1. Spatial data nature (II) How discretize space? GIS: Raster & vector data Raster data or “field view”: considers de world as a field, a continuous variation Vector data or “object view”: discrete objects Discretization results in a set of regular or irregular contours, sample points or a spatial partition In the object view the world is depicted as an empty space populated by points, lines and areas @ 2007, Coro Chasco Yrigoyen All Rights Reserved 10 Session 1 1.3.1. Spatial data nature (III) FRECUENCIES Trim. Time series: Month Spatial data: Year SCALES Province Country Year Region @ 2007, Coro Chasco Yrigoyen All Rights Reserved 11 Session 1 1.3.1. Spatial data nature (IV) Type of spatial data: Polygons Points Lines (e.g. roads) Points (e.g. outlets) (e.g. districts) M-30 CONCEPCION CONCEPCION CONCEPCION A GUINDALERA GUINDALERA GUINDALERA LISTA LISTA LISTA VENTAS VENTAS VENTAS GOYA GOYA GOYA FUENTE FUENTE DEL BERRO FUENTEDEL DELBERRO BERRO @ 2007, Coro Chasco Yrigoyen All Rights Reserved Type of time series data: 12 (e.g.months) Session 1 1.3.1. Spatial data nature (V) 1 temporal reference: 2 spatial references: Spain: coordinates B.C. Greenwich: longitude 0 (-300,4200) (0,0) Year 0 Equator: Axis X: latitude 0 Longitude A.C. Axis Y: Latitude @ 2007, Coro Chasco Yrigoyen All Rights Reserved 13 Session 1 1.3.2. Spatial data problems Aggregation and spatial arrangement Changes of supports Identification - Micro-macro aggregation problem – Econometrics Modifiable Areal Unit Problem (MAUP) - Geography Ecological fallacy - Sociology Change of Support Problem (COSP) - Geostatistics @ 2007, Coro Chasco Yrigoyen All Rights Reserved 14 ANSELIN, L (1988), “Spatial econometrics: methods and models”, Kluwer Academic Publishers Session 1 1.3.2. Spatial data problems (II) Modifiable Areal Unit Problem (MAUP) The statistical measures for cross-sectional data are sensitive to the way in which the spatial units are organised. The level of aggregation and the spatial arrangement in zones (i.e., combinations of contiguous units) affects the magnitude of various measures of association, such as spatial autocorrelation coefficients and parameters in a regression model. Spatial data aggregation only lead to sensical conclusions when the underlying phenomenon is homogeneous. If not, the inherent spatial heterogeneiy and structural instability should be accounted for in the various aggregation schemes. a1 a2 A @ 2007, Coro Chasco Yrigoyen All Rights Reserved b1 B 15 b2 GOTWAY, CA and LJ YOUNG (2002), “Combining incompatible spatial data”. Journal of the American Statistical Association 97, 632-648 Session 1 1.3.2. Spatial data problems (III) The different types of spatial data (point, line, area, surface), occurring naturally or as a result of the measurement process, potentially allow many ways of integrating these data. Changing the support of a spatial variable (volume, geometrical size, shape & spatial orientation) -typically by aggregation/averagingcreates a new variable. @ 2007, Coro Chasco Yrigoyen All Rights Reserved 16 Session 1 1.3.2. Spatial data problems (IV) From an econometric standpoint, MAUP is also an identification problem, problem since there is insufficient information in the data to allow for the full specification of the simultaneous interaction over space. In this sense, a formulation of linear spatial association can be considered as a special case of a system of simultaneous linear equations, with one observation for each equation. SYSTEM OF SIMULTANEOUS LINEAR EQUATIONS LINEAR SPATIAL AUTOREGRESSIVE MODEL N N yi = ρ ∑ wij yi yi = ∑ ρi wij yi i =1 i =1 @ 2007, Coro Chasco Yrigoyen All Rights Reserved 17 Session 1 1.3.3. Spatial data empirical applications Empirical applications with spatial data imply the use of: Statistical data GIS & digital maps Spatial econometrics software @ 2007, Coro Chasco Yrigoyen All Rights Reserved 18 1.3.3. Spatial data empirical applications (II) Statistical data: data Primary data: from private or public surveys... Secondary data. E.g. in Spain: Regional data: INE, IVIE-BBVA, “la Caixa”... Municipal data: INE, “la Caixa”-LR Klein Institute... Districts, census tracts: INE, Regional Statistics Offices Others: Expérian, Mosaic, Data Segmento, Sitesa... @ 2007, Coro Chasco Yrigoyen All Rights Reserved 19 1.3.3. Spatial data empirical applications (III) ¾ GIS: Managment, analysis and visualization of spatial data. It is estructures in interactive maps, spatial data, geoprocessing models, data models and metadata” @ 2007, Coro Chasco Yrigoyen All Rights Reserved 20 1.3.3. Spatial data empirical applications (IV) Digital maps: maps Instituto Nacional de Estadística: regions, provinces, municipalities, districts, census tracts, streets... ESRI: Muniview, Censalview, Arcópolis MapInfo Others: Maptel... @ 2007, Coro Chasco Yrigoyen All Rights Reserved 21 1.3.3. Spatial data empirical applications (V) Spatial econometrics software: Focused on ESDA: GEODA: GEODA https://www.geoda.uiuc.edu/downloadin.php STARS: STARS http://regal.sdsu.edu/index.php Focused on confirmatory analysis: SpaceStat: SpaceStat http://www.terraseer.com Spatial Econometrics Toolbox (Matlab): http://www.spatial-econometrics.com spdep (R): http://cran.r-project.org PySAL (Python): http://sal.uiuc.edu @ 2007, Coro Chasco Yrigoyen All Rights Reserved 22 GEODA: GEODA https://www.geoda.uiuc.edu/downloadin.php Created by Luc Anselin, Anselin University of Illinois in Urbana-Champaign @ 2007, Coro Chasco Yrigoyen All Rights Reserved 23 GEODA: GEODA https://www.geoda.uiuc.edu/downloadin.php GIS + ESDA + Spatial regression Exploratory Spatial Data Analysis (AEDE): - Data managment - Mapping and descriptive statistics - Spatial association: global and local spatial autocorrelation statistics Spatial regression: - Basic OLS model - Spatial lag model - Spatial error model @ 2007, Coro Chasco Yrigoyen All Rights Reserved 24 GEODA: GEODA https://www.geoda.uiuc.edu/downloadin.php @ 2007, Coro Chasco Yrigoyen All Rights Reserved 25 STARS: STARS http://regal.sdsu.edu/index.php Created by Serge Rey, Rey San Diego State University. @ 2007, Coro Chasco Yrigoyen All Rights Reserved 26 STARS (II): (II) http://regal.sdsu.edu/index.php @ 2007, Coro Chasco Yrigoyen All Rights Reserved 27 STARS (III): (III) http://regal.sdsu.edu/index.php @ 2007, Coro Chasco Yrigoyen All Rights Reserved 28 spdep for R: http://cran.r-project.org Created by Roger Bivand, Bivand Norwegian School of Economics and Business Administration,. @ 2007, Coro Chasco Yrigoyen All Rights Reserved 29 PYSAL: PYSAL http://sal.uiuc.edu Future Luc Anselin,’s Anselin project: combination of GEODA- STARS-SPDEP opensource 30 software. @ 2007, Coro Chasco Yrigoyen All Rights Reserved PYSAL (II) @ 2007, Coro Chasco Yrigoyen All Rights Reserved 31 Spatial Econometrics Toolbox for Matlab: Matlab http://www.spatial-econometrics.com Created by James LeSage, LeSage University of Toledo. Main contents: 2 Spatial autoregressive models 2.1 The rst-order spatial AR model 2.2 The mixed autoregressiveregressive model 2.3 The spatial errors model 2.4 The general spatial model 3 Bayesian Spatial autoregressive models 4 Locally linear spatial models 4.1 Spatial expansion 4.2 DARP models 4.3 GWR 4.4 A Bayesian Approach to GWR 5 Limited dependent variable models 5.2 The Gibbs sampler 5.3 Heteroscedastic models 6 VAR and Error Correction Models 6.1 VAR models 6.2 Error correction models 6.3 Bayesian variants 6.4 Forecasting the models @ 2007, Coro Chasco Yrigoyen All Rights Reserved 32 Main contents: SpaceStat: SpaceStat 1 Data management and algebra 2 Spatial weight matrices http://www.terraseer.com 2.1 Creation & characteristics 2.2 Spatial correlogram 2.3 Spatial lags Created by Luc Anselin 3 ESDA 3.1 Basic descriptive statistics 3.2 Joint count, Moran’s I & Geary’s c 3.3 G statistics & QAP measures 4 SPATIAL REGRESSION 4.1 Basic OLS regression 4.2 Spatial lag & spatial error models 4.3 Instrumental Variables estimation 4.4 Heteroskedastic model 4.5 Trend surface model 4.6 Spatial regimes & S-ANOVA 4.7 Spatial expansion @ 2007, Coro Chasco Yrigoyen All Rights Reserved 33 Session 1 1.4. Some applications in spatial econometrics (I) Cancer mortality rate model Homicide rate Spatial β-convergence regional model Household per capita income @ 2007, Coro Chasco Yrigoyen All Rights Reserved 34 Session 1 1.3. Some applications... (II) Haining: Cancer mortality rate model (spatial-error model) y: cancer mortality rate i: one of the 87 Glasgow community medicine areas x: deprivation index W: spatial weight matrix β0, β1, λ: parameters to be estimated u, ε: stochastic error terms . HAINING, R. (1995), “Data problems in spatial econometric modeling”. In L. Anselin and R. Florax (ed.), “New directions in spatial econometrics”. Springer-Verlag, Berlin; pp. 156-171. @ 2007, Coro Chasco Yrigoyen All Rights Reserved 35 Session 1 1.3. Some applications... (III) Baller, Anselin, Messner, Deane and Hawkings: Homicide rate (spatial-lag model) H: i: D: P: V: U: E: homicide rate U.S. county resource deprivation index population structure percent divorced unemployment rate median age W: spatial weight matrix ρ: spatial autoregressive parameter β0, β1, β2, β3, β5, β6: exogenous vars. Parameters ε: i.i.d. stochastic error term . BALLER, R., L. ANSELIN, S. MESSNER and D. HAWKINS (2001), “Structural covariates of U.S. county homicide rates: incorporating spatial effects”. Criminology (próxima publicación). @ 2007, Coro Chasco Yrigoyen All Rights Reserved 36 Session 1 1.3. Some applications... (IV) Rey & Montouri: Spatial β-convergence regional model (spatial cross-regressive model) yi,t: real per capita income in state i year t k: year period W: spatial weight matrix α, β, γ : parameters to be estimated u: i.i.d. stochastic error term . REY, S. and B. MONTOURI (1999), “US regional income convergence: a spatial econometric perspective”. Regional Studies, vol. 33.2; pp. 143-156. @ 2007, Coro Chasco Yrigoyen All Rights Reserved 37 Session 1 1.3. Some applications... (V) Chasco: Household per capita income (spatial-lag model with two spatial regimes) Spatial lag . CHASCO, C. (2003), “Econometría espacial aplicada a la predicción-extrapolación de datos microterritoriales”. Consejería de Economía e Innovación Tecnológica de la Comunidad de Madrid. @ 2007, Coro Chasco Yrigoyen All Rights Reserved 38