Template-Type: ReDIF-Article 1.0 Author-Name: Estefanía Villar Cheda Author-Name: Mª Esther Calvo Ocampo Author-Name: Mª Esther López Vizcaíno Author-Name: Carlos L. Iglesias Patiño Author-Name: Solmary Silveira Calviño Author-Name: Mª Isolina Santiago Pérez Title: Clasificación de los municipios gallegos según su grado de urbanización Abstract: RESUMEN: En el marco de la estadística oficial y dentro de las clasificaciones que emplean un número reducido de variables geodemográficas, el Instituto Galego de Estatística (IGE) publicó en julio del 2011 una clasificación del grado de urbanización de los municipios gallegos en seis zonas, basada en un estándar europeo. Teniendo en cuenta análisis posteriores (Calvo et al, 2012) se optó por dividir a Galicia en tres zonas, rural, semiurbana y urbana. El objetivo de este trabajo es comprobar cómo esta agregación en 3 clases,  se asemeja a la obtenida con un modelo que emplea características sociodemográficas, sociolaborales y económicas que, a priori, se vinculan con el proceso de urbanización.ABSTRACT:Provide informationat the provincial levelin Galiciadoes not always solvethe problems associated withdesigningand implementing public policies, especiallythose that aim to reduce disparitiesin the level of development achieved. Thisdoes not solve, either, the demands ofprivateinformation.  However, thereis broad agreementthat the habitatcan be determinant insocioeconomic status. Therefore it is interesting to have informationrelatesto the urbanandruralinGaliciaor itsprovinces. Apriori, the difference betweenthe urban areas oftwo provincesseems lower than betweentheurban andruralareasofthe same province.It would be usefulto contrastthis assumptionwithempirical evidence.  Inthis framework and within the officialstatisticsclassificationswhich employ asmall number ofgeo-demographicvariables, the Galician Statistics Institute(IGE) published in July 2011a classificationof the degree ofurbanization of Galician municipalitiesinto six types of areas, based on a European standard. Taking into account subsequent reviews (Calvo et al, 2012) we divideGaliciainto three areas, rural, semi-urbanand urban.Theaim of this studyis to see howthis aggregationinto 3 classes is similar to that obtained witha statistical model thatuses sociodemographic,labour and economic characteristics that, a priori, are linked tothe urbanisation process.  In this case the units of analysis are the 315 municipalities of Galicia and the dependent variable is the classification of municipalities in rural, semi-urban and urban areas. To analyze its relationship with other sociodemographic, labor and economicvariables, we adjusta multinomial logistic regression model which is an extension of the classical logistic regression to the case where the response variable is more than two categories(1-rural classification, 2- semi-urban, 3- urban). In this multinomial logistic regression model, the category of urban is considered as the reference, and each of the remaining categories are compared to it.  For the model variable selection we take into account the results of the three phases of the exploratory analysis: box plots, univariate multinomial logistic models and correlation analysis. We select 12 variables for use as potential explanatory variables in the multinomial model fit, for the first step. After that, we readjust the model with the variables that were significant in the first step and we repeat again what was done in the previous step. We finish the model fit in the third step, in which all the variables included are significant: the average age, the percentage of people registered in Social Security System in services, gross disposable income per capita (GDI) and the social benefits in percentage of GDI.  We calculate also the Odds ratio (OR) that is the relative probability that a municipality isclassifiedin the category k of the response variableYoverit is classifiedinthe reference category(third, inthis case) foran element witha valueofX=1 comparedwith one that hasa valueof X= 0.If the variableXis ordinalor continuous, theORcomparesthe relative probabilityassociated witha unit change inthe variableX. In this case exists interaction between variables, then the estimate of theORassociated witha variabledepend on the valueof the variable thatis interactingwith it.Thisis the caseof the social benefits in percentage of GDI, whichdependsof the ORof the registered in Social Security System in services.  For example, in theruralmodel, theORfor oneunit change inthe social benefits in percentage of GDI dependsonthe value of the ORof the people registered in Social Security System in services through theexpressionexp (1.895) exp(-0.04*PAFIL_SERV), so ,in a municipalitywith a percentageequal to30.2% of registered in Social Security System(minimum value) theORof benefitsis2.011andfora municipality with anpercentage of registered in services equal84.1%(maximum) is 0.233.Theunit increaseinthe social benefits in percentage of GDI increases the probability ofrural versusurban(OR>1)in municipalities withlow values ​​of thepercentageof registered in Social Security System in servicesanddecrease (OR <1) in municipalitieswith highervalues ​​of thisvariable.  Thefit of the modelpermit evaluateif theinitial classificationis similar tothat predictedby the model.It was consideredthe maximum(p1, p2, p3) as a criterionto classifymunicipalities. The estimated modelhas anability to correctly classify85.4% of the municipalities, butthe modelclassifiesbetterruraland semi-urban. In fact, themodel is consistentwith the classification ofIGEin 93.2% of rural municipalities, 73.9% of semi-urbanand 72.2% of urban.It is convenient, however, a more detailedanalysisof the 46municipalities assignedby the model thatdifferfrom the initial categories.  The characterization of the degree of urbanization of Galicia inthree areas(urban, rural and transitional), from theaggregationof a classificationin sixdegrees,thatmainlyuses thepopulation density,isadequately explainedby amultinomiallogistic modelthat usesothersocioeconomic variables. With this,ithas achievedthe goalof the workin its variousfacets. First,the urban spaceinGaliciais broader thanthe ZPDoriginal,enlarged withtheZIPhigh. Second, a classification thatis basedonlyin basic considerationsgeo-demographicis consistentwith anotherthat employs arichersocioeconomic data. Third and last, we have used anew methodology based ona statistical techniquedifferent fromthose previouslyusedtobuilda clearer anda moreparsimonious model.  The methodology usedin this work allowsto obtain conclusionsdirectlyfrom the variablesconsidered in the model, thus avoiding the useof techniquesto reducethe dimension(whichare detrimental tothe interpretabilityof the model)as has been donein other studies ofsimilar characteristics(Iglesiasetal, 2000;Ruaetal, 2003; Penaetal, 2008). The model alsoallows us to conclude, as inother studies, that the presence ofthe tertiary sectorand high levelsofincome, linked to the demographic dynamics, appear to be the determinantsofintra-regionaldisparities in Galicia (Pena etal, 2008). Classification-JEL: R1 Keywords: Galicia, Grado de urbanización, Urbano, Rural, Clasificación, Regresión multinomial, Degree of urbanisation, Urban, Clasification, Multinomial regression Pages: 193-223 Volume: 03 Year: 2013 File-URL: http://www.revistaestudiosregionales.com/documentos/articulos/pdf-articulo-2426.pdf File-Format: Application/pdf Handle: RePEc:rer:articu:v:03:y:2013:p:193-223