Introduction and Research QuestionThe Model
Results
Determinats of Regionals Convergence(Divergence)
Insights from Intradistribution Dynamics
Fabrizi E.1 Guastella G.2 Timpano F.1
1Dep. of Economics and Social SciencesFaculty of Economics - Catholic University, Piacenza
2Doctoral School in Economic PolicyCatholic University, Piacenza
AISRe Annual Conference, Aosta, 2010
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Outline
1 Introduction and Research QuestionMotivationBackgroundResearch Question
2 The ModelThe Multinomial Response ModelBinary response model
3 ResultsTransition probabilitiesRegression OutputConclusion
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
Outline
1 Introduction and Research QuestionMotivationBackgroundResearch Question
2 The ModelThe Multinomial Response ModelBinary response model
3 ResultsTransition probabilitiesRegression OutputConclusion
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
Motivation
Transition dynamics approach has been introduced as analternative test for convergenceConvergence (in the long run) is considered to be theresult of movements within the distributionThe determinants of regional development are however notconsideredThis work is a first attempt to use information fromintradistribution dynamics to discuss determinants ofregional growth
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
Motivation
Transition dynamics approach has been introduced as analternative test for convergenceConvergence (in the long run) is considered to be theresult of movements within the distributionThe determinants of regional development are however notconsideredThis work is a first attempt to use information fromintradistribution dynamics to discuss determinants ofregional growth
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
Motivation
Transition dynamics approach has been introduced as analternative test for convergenceConvergence (in the long run) is considered to be theresult of movements within the distributionThe determinants of regional development are however notconsideredThis work is a first attempt to use information fromintradistribution dynamics to discuss determinants ofregional growth
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
Motivation
Transition dynamics approach has been introduced as analternative test for convergenceConvergence (in the long run) is considered to be theresult of movements within the distributionThe determinants of regional development are however notconsideredThis work is a first attempt to use information fromintradistribution dynamics to discuss determinants ofregional growth
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
Outline
1 Introduction and Research QuestionMotivationBackgroundResearch Question
2 The ModelThe Multinomial Response ModelBinary response model
3 ResultsTransition probabilitiesRegression OutputConclusion
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
The standard approach
Regional development is analyzed by mean of growthregression
Conditional convergence (Institutions and structuralcharacteristics)Externalities and spilloversDeterminants of development (HC, R&D, Agglomerationeconomies,...)
β-convergence is however generally not sufficientσ-convergence only focuses on the SD of the incomedistribution
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
The standard approach
Regional development is analyzed by mean of growthregression
Conditional convergence (Institutions and structuralcharacteristics)Externalities and spilloversDeterminants of development (HC, R&D, Agglomerationeconomies,...)
β-convergence is however generally not sufficientσ-convergence only focuses on the SD of the incomedistribution
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
The standard approach
Regional development is analyzed by mean of growthregression
Conditional convergence (Institutions and structuralcharacteristics)Externalities and spilloversDeterminants of development (HC, R&D, Agglomerationeconomies,...)
β-convergence is however generally not sufficientσ-convergence only focuses on the SD of the incomedistribution
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
The standard approach
Regional development is analyzed by mean of growthregression
Conditional convergence (Institutions and structuralcharacteristics)Externalities and spilloversDeterminants of development (HC, R&D, Agglomerationeconomies,...)
β-convergence is however generally not sufficientσ-convergence only focuses on the SD of the incomedistribution
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
The standard approach
Regional development is analyzed by mean of growthregression
Conditional convergence (Institutions and structuralcharacteristics)Externalities and spilloversDeterminants of development (HC, R&D, Agglomerationeconomies,...)
β-convergence is however generally not sufficientσ-convergence only focuses on the SD of the incomedistribution
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
The standard approach
Regional development is analyzed by mean of growthregression
Conditional convergence (Institutions and structuralcharacteristics)Externalities and spilloversDeterminants of development (HC, R&D, Agglomerationeconomies,...)
β-convergence is however generally not sufficientσ-convergence only focuses on the SD of the incomedistribution
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
The alternative approach
Markov chains and long-run distributionmovements within different parts of the distributiontransition probabilitiesergodic distribution and equilibrium analysis
Markov or not Markov?classes boundaries and sensitivity of resultstime homogeneity (to make inference about equilibriumdistribution)
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
The alternative approach
Markov chains and long-run distributionmovements within different parts of the distributiontransition probabilitiesergodic distribution and equilibrium analysis
Markov or not Markov?classes boundaries and sensitivity of resultstime homogeneity (to make inference about equilibriumdistribution)
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
The alternative approach
Markov chains and long-run distributionmovements within different parts of the distributiontransition probabilitiesergodic distribution and equilibrium analysis
Markov or not Markov?classes boundaries and sensitivity of resultstime homogeneity (to make inference about equilibriumdistribution)
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
The alternative approach
Markov chains and long-run distributionmovements within different parts of the distributiontransition probabilitiesergodic distribution and equilibrium analysis
Markov or not Markov?classes boundaries and sensitivity of resultstime homogeneity (to make inference about equilibriumdistribution)
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
The alternative approach
Markov chains and long-run distributionmovements within different parts of the distributiontransition probabilitiesergodic distribution and equilibrium analysis
Markov or not Markov?classes boundaries and sensitivity of resultstime homogeneity (to make inference about equilibriumdistribution)
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
The alternative approach
Markov chains and long-run distributionmovements within different parts of the distributiontransition probabilitiesergodic distribution and equilibrium analysis
Markov or not Markov?classes boundaries and sensitivity of resultstime homogeneity (to make inference about equilibriumdistribution)
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
The alternative approach
Markov chains and long-run distributionmovements within different parts of the distributiontransition probabilitiesergodic distribution and equilibrium analysis
Markov or not Markov?classes boundaries and sensitivity of resultstime homogeneity (to make inference about equilibriumdistribution)
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
Markov chain and the determinants of development
Probabilities give a clearer idea of the developmentprocessEven sustained growth may in fact be not sufficient totransitateHowever we know which regions transitate but not why
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
Markov chain and the determinants of development
Probabilities give a clearer idea of the developmentprocessEven sustained growth may in fact be not sufficient totransitateHowever we know which regions transitate but not why
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
Markov chain and the determinants of development
Probabilities give a clearer idea of the developmentprocessEven sustained growth may in fact be not sufficient totransitateHowever we know which regions transitate but not why
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
Outline
1 Introduction and Research QuestionMotivationBackgroundResearch Question
2 The ModelThe Multinomial Response ModelBinary response model
3 ResultsTransition probabilitiesRegression OutputConclusion
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
A first attempt to explain transition
Transition is the result of very sustained growthWe aim to find a link between
the probability of transition andthe determinants of development
very sustained growth
It is necessary to ensure that transition is not the result of asimple statistical effect!
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
A first attempt to explain transition
Transition is the result of very sustained growthWe aim to find a link between
the probability of transition andthe determinants of development
very sustained growth
It is necessary to ensure that transition is not the result of asimple statistical effect!
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
A first attempt to explain transition
Transition is the result of very sustained growthWe aim to find a link between
the probability of transition andthe determinants of development
very sustained growth
It is necessary to ensure that transition is not the result of asimple statistical effect!
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
A first attempt to explain transition
Transition is the result of very sustained growthWe aim to find a link between
the probability of transition andthe determinants of development
very sustained growth
It is necessary to ensure that transition is not the result of asimple statistical effect!
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
MotivationBackgroundResearch Question
A first attempt to explain transition
Transition is the result of very sustained growthWe aim to find a link between
the probability of transition andthe determinants of development
very sustained growth
It is necessary to ensure that transition is not the result of asimple statistical effect!
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Outline
1 Introduction and Research QuestionMotivationBackgroundResearch Question
2 The ModelThe Multinomial Response ModelBinary response model
3 ResultsTransition probabilitiesRegression OutputConclusion
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Multinomial Logistic Regression
With Multinomial model it is possibleto model the transition from different origins
different factors are important in different stages ofdevelopment
to get coefficient estimates which are destination specificsome factors determine larger transitions
to normalize coefficientcoefficients represent the change in probabilities to move toanother classwrt the probability to stay in the origin class
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Multinomial Logistic Regression
With Multinomial model it is possibleto model the transition from different origins
different factors are important in different stages ofdevelopment
to get coefficient estimates which are destination specificsome factors determine larger transitions
to normalize coefficientcoefficients represent the change in probabilities to move toanother classwrt the probability to stay in the origin class
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Multinomial Logistic Regression
With Multinomial model it is possibleto model the transition from different origins
different factors are important in different stages ofdevelopment
to get coefficient estimates which are destination specificsome factors determine larger transitions
to normalize coefficientcoefficients represent the change in probabilities to move toanother classwrt the probability to stay in the origin class
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Multinomial Logistic Regression
With Multinomial model it is possibleto model the transition from different origins
different factors are important in different stages ofdevelopment
to get coefficient estimates which are destination specificsome factors determine larger transitions
to normalize coefficientcoefficients represent the change in probabilities to move toanother classwrt the probability to stay in the origin class
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Multinomial Logistic Regression
With Multinomial model it is possibleto model the transition from different origins
different factors are important in different stages ofdevelopment
to get coefficient estimates which are destination specificsome factors determine larger transitions
to normalize coefficientcoefficients represent the change in probabilities to move toanother classwrt the probability to stay in the origin class
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Multinomial Logistic Regression
With Multinomial model it is possibleto model the transition from different origins
different factors are important in different stages ofdevelopment
to get coefficient estimates which are destination specificsome factors determine larger transitions
to normalize coefficientcoefficients represent the change in probabilities to move toanother classwrt the probability to stay in the origin class
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Multinomial Logistic Regression
With Multinomial model it is possibleto model the transition from different origins
different factors are important in different stages ofdevelopment
to get coefficient estimates which are destination specificsome factors determine larger transitions
to normalize coefficientcoefficients represent the change in probabilities to move toanother classwrt the probability to stay in the origin class
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
The MLG: Problems
Trade off betweennumber of classes (detail of the analysis)degree of freedom (for each regression)
Low number of transition for more than 1 classthe transition is the result of a statistical effect due to classboundariesthe choice of boundaries should guarantee a sufficientnumber of transitionwith 1 class transition the model reduces to a simple logisticregression
ConclusionClasses boundaries are chose according to results: sensitivityof results
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
The MLG: Problems
Trade off betweennumber of classes (detail of the analysis)degree of freedom (for each regression)
Low number of transition for more than 1 classthe transition is the result of a statistical effect due to classboundariesthe choice of boundaries should guarantee a sufficientnumber of transitionwith 1 class transition the model reduces to a simple logisticregression
ConclusionClasses boundaries are chose according to results: sensitivityof results
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
The MLG: Problems
Trade off betweennumber of classes (detail of the analysis)degree of freedom (for each regression)
Low number of transition for more than 1 classthe transition is the result of a statistical effect due to classboundariesthe choice of boundaries should guarantee a sufficientnumber of transitionwith 1 class transition the model reduces to a simple logisticregression
ConclusionClasses boundaries are chose according to results: sensitivityof results
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
The MLG: Problems
Trade off betweennumber of classes (detail of the analysis)degree of freedom (for each regression)
Low number of transition for more than 1 classthe transition is the result of a statistical effect due to classboundariesthe choice of boundaries should guarantee a sufficientnumber of transitionwith 1 class transition the model reduces to a simple logisticregression
ConclusionClasses boundaries are chose according to results: sensitivityof results
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
The MLG: Problems
Trade off betweennumber of classes (detail of the analysis)degree of freedom (for each regression)
Low number of transition for more than 1 classthe transition is the result of a statistical effect due to classboundariesthe choice of boundaries should guarantee a sufficientnumber of transitionwith 1 class transition the model reduces to a simple logisticregression
ConclusionClasses boundaries are chose according to results: sensitivityof results
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
The MLG: Problems
Trade off betweennumber of classes (detail of the analysis)degree of freedom (for each regression)
Low number of transition for more than 1 classthe transition is the result of a statistical effect due to classboundariesthe choice of boundaries should guarantee a sufficientnumber of transitionwith 1 class transition the model reduces to a simple logisticregression
ConclusionClasses boundaries are chose according to results: sensitivityof results
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
The MLG: Problems
Trade off betweennumber of classes (detail of the analysis)degree of freedom (for each regression)
Low number of transition for more than 1 classthe transition is the result of a statistical effect due to classboundariesthe choice of boundaries should guarantee a sufficientnumber of transitionwith 1 class transition the model reduces to a simple logisticregression
ConclusionClasses boundaries are chose according to results: sensitivityof results
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
The MLG: Problems
Trade off betweennumber of classes (detail of the analysis)degree of freedom (for each regression)
Low number of transition for more than 1 classthe transition is the result of a statistical effect due to classboundariesthe choice of boundaries should guarantee a sufficientnumber of transitionwith 1 class transition the model reduces to a simple logisticregression
ConclusionClasses boundaries are chose according to results: sensitivityof results
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Outline
1 Introduction and Research QuestionMotivationBackgroundResearch Question
2 The ModelThe Multinomial Response ModelBinary response model
3 ResultsTransition probabilitiesRegression OutputConclusion
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Logistic regression
Transition is modelled according toMove forward (1) vs stay (0)Move backward (1) vs stay (0)
no differentiation according to origin classdifferentiation based on NMSDifferentiation based on income level
high number of classeslow sensitivity to boundariesstill enought to ensure ergodic properties of TPM
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Logistic regression
Transition is modelled according toMove forward (1) vs stay (0)Move backward (1) vs stay (0)
no differentiation according to origin classdifferentiation based on NMSDifferentiation based on income level
high number of classeslow sensitivity to boundariesstill enought to ensure ergodic properties of TPM
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Logistic regression
Transition is modelled according toMove forward (1) vs stay (0)Move backward (1) vs stay (0)
no differentiation according to origin classdifferentiation based on NMSDifferentiation based on income level
high number of classeslow sensitivity to boundariesstill enought to ensure ergodic properties of TPM
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Logistic regression
Transition is modelled according toMove forward (1) vs stay (0)Move backward (1) vs stay (0)
no differentiation according to origin classdifferentiation based on NMSDifferentiation based on income level
high number of classeslow sensitivity to boundariesstill enought to ensure ergodic properties of TPM
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Logistic regression
Transition is modelled according toMove forward (1) vs stay (0)Move backward (1) vs stay (0)
no differentiation according to origin classdifferentiation based on NMSDifferentiation based on income level
high number of classeslow sensitivity to boundariesstill enought to ensure ergodic properties of TPM
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Logistic regression
Transition is modelled according toMove forward (1) vs stay (0)Move backward (1) vs stay (0)
no differentiation according to origin classdifferentiation based on NMSDifferentiation based on income level
high number of classeslow sensitivity to boundariesstill enought to ensure ergodic properties of TPM
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Logistic regression
Transition is modelled according toMove forward (1) vs stay (0)Move backward (1) vs stay (0)
no differentiation according to origin classdifferentiation based on NMSDifferentiation based on income level
high number of classeslow sensitivity to boundariesstill enought to ensure ergodic properties of TPM
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Logistic regression
Transition is modelled according toMove forward (1) vs stay (0)Move backward (1) vs stay (0)
no differentiation according to origin classdifferentiation based on NMSDifferentiation based on income level
high number of classeslow sensitivity to boundariesstill enought to ensure ergodic properties of TPM
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
Logistic regression
Transition is modelled according toMove forward (1) vs stay (0)Move backward (1) vs stay (0)
no differentiation according to origin classdifferentiation based on NMSDifferentiation based on income level
high number of classeslow sensitivity to boundariesstill enought to ensure ergodic properties of TPM
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
ESPON dataset 1999-2000
Dependent: per capita gdp in PPS (1999-2007)Regressors
share of employment in services, industry and agricolturelong-term unemploymentpopulation densityred
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
ESPON dataset 1999-2000
Dependent: per capita gdp in PPS (1999-2007)Regressors
share of employment in services, industry and agricolturelong-term unemploymentpopulation densityred
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
ESPON dataset 1999-2000
Dependent: per capita gdp in PPS (1999-2007)Regressors
share of employment in services, industry and agricolturelong-term unemploymentpopulation densityred
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
ESPON dataset 1999-2000
Dependent: per capita gdp in PPS (1999-2007)Regressors
share of employment in services, industry and agricolturelong-term unemploymentpopulation densityred
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
ESPON dataset 1999-2000
Dependent: per capita gdp in PPS (1999-2007)Regressors
share of employment in services, industry and agricolturelong-term unemploymentpopulation densityred
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
ESPON dataset 1999-2000
Dependent: per capita gdp in PPS (1999-2007)Regressors
share of employment in services, industry and agricolturelong-term unemploymentpopulation densityred
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
ESPON dataset 1999-2000
Regressorsroadkm and intaccfunds received up to 1999
More data?country dummy: fixed effects capturing also some dep varother structural characteristics: need for data reduction
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
ESPON dataset 1999-2000
Regressorsroadkm and intaccfunds received up to 1999
More data?country dummy: fixed effects capturing also some dep varother structural characteristics: need for data reduction
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
ESPON dataset 1999-2000
Regressorsroadkm and intaccfunds received up to 1999
More data?country dummy: fixed effects capturing also some dep varother structural characteristics: need for data reduction
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
ESPON dataset 1999-2000
Regressorsroadkm and intaccfunds received up to 1999
More data?country dummy: fixed effects capturing also some dep varother structural characteristics: need for data reduction
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
ESPON dataset 1999-2000
Regressorsroadkm and intaccfunds received up to 1999
More data?country dummy: fixed effects capturing also some dep varother structural characteristics: need for data reduction
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
The Multinomial Response ModelBinary response model
ESPON dataset 1999-2000
Regressorsroadkm and intaccfunds received up to 1999
More data?country dummy: fixed effects capturing also some dep varother structural characteristics: need for data reduction
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Outline
1 Introduction and Research QuestionMotivationBackgroundResearch Question
2 The ModelThe Multinomial Response ModelBinary response model
3 ResultsTransition probabilitiesRegression OutputConclusion
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Table of probabilities - ML estimates
finalinit 0.6 0.741 0.834 0.922 1 1.07 1.13 1.22 1.38 Inf0.6 0.828 0.1720.741 0.050 0.750 0.150 0.0500.834 0.200 0.440 0.200 0.120 0.0400.922 0.381 0.381 0.095 0.095 0.0481 0.538 0.308 0.077 0.038 0.0381.0 0.435 0.087 0.0431.13 0.091 0.318 0.273 0.3181.22 0.036 0.036 0.250 0.250 0.321 0.1071.38 0.042 0.250 0.625 0.083Inf 0.040 0.280 0.680
ergodic 0.6 0.741 0.834 0.922 1 1.07 1.13 1.22 1.38 Inf0.05 0.173 0.173 0.186 0.13 0.106 0.04 0.05 0.061 0.03
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Outline
1 Introduction and Research QuestionMotivationBackgroundResearch Question
2 The ModelThe Multinomial Response ModelBinary response model
3 ResultsTransition probabilitiesRegression OutputConclusion
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Results with 4 classes
ML estimatesForward Backward
Estimate z value Estimate z value(Intercept) -29.7546 (-3.345)*** -3.0315 (-0.419)seragri 0.4398 (0.843) 0.9713 (1.937).ltu 1.2463 (1.089) 3.5127 (3.577)***popd -0.7294 (-1.634) -1.0499 (-2.408)*educ 1.6455 (1.893). -1.6331 (-2.545)*roadkm -0.2027 (-1.319) 0.2372 (1.434)red 0.2684 (0.564) -0.1564 (-0.450)intacc 0.4276 (0.326) -0.2256 (-0.240)funds 0.9949 (3.939)*** -0.2653 (-1.292)nms 17.0169 (3.853)*** -7.0361 (-1.801).Note: .,*,**,*** indicate significance at 90%, 95%, 99%, 99.9%.
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Results with 10 classes
ML estimatesForward Backward
Estimate z value Estimate z value(Intercept) -13.35170 (-2.405)* 16.16831 (3.249)**seragri 0.13173 (0.366) 0.35112 (1.068)ltu -0.22029 (-0.353) 0.38235 (0.869)popd -0.20579 (-0.668) -0.25530 (-0.953)educ 1.07204 (1.886). -1.37568 (-2.959)**roadkm -0.24873 (-2.309)* 0.14600 (1.503)red 0.41517 (1.320) 0.24917 (0.940)intacc 0.06951 (0.080) -1.56432 (-2.124)*funds1 0.55746 (3.366)*** -0.42325 (-2.776)**nms 11.04681 (3.804)*** -25.69953 (-0.027)Note: .,*,**,*** indicate significance at 90%, 95%, 99%, 99.9%.
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Results with NMS regimes - Forward
ML estimatesNMS NON-NMS
Estimate z value Estimate z value(Intercept) -9.54762 (-0.777)*** -25.53175 (-3.463)***seragri 3.93361 (1.841). -0.91786 (-1.859).ltu -0.72519 (-0.288) 0.02364 (0.029)popd -1.23960 (-0.756) 0.30936 (0.852)educ 0.97002 (0.637)** 2.23724 (2.653)**roadkm 0.27350 (0.417)* -0.27109 (-2.302)*red 2.07724 (1.651) -0.12609 (-0.328)intacc 1.32019 (0.582). 1.74943 (1.655).funds1 0.52072 (0.976)** 0.62902 (3.143)**Note: .,*,**,*** indicate significance at 90%, 95%, 99%, 99.9%.
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Results with NMS regimes - Backward
ML estimatesNMS NON-NMS
Estimate z value Estimate z value(Intercept) -1.857e+01 (-0.001)** 1.617e+01 (3.249)**seragri -4.170e-12 (-2.02e-15) 3.511e-01 (1.068)ltu -1.815e-11 (-3.31e-15) 3.823e-01 (0.869)popd 3.625e-12 (1.47e-15) -2.553e-01 (-0.953)educ -8.648e-13 (-2.77e-16)** -1.376e+00 (-2.959)**roadkm 3.190e-13 (2.89e-16) 1.460e-01 (1.503)red -7.394e-13 (-4.05e-16) 2.492e-01 (0.940)intacc -4.382e-12 (-1.53e-15)* -1.564e+00 (-2.124)*funds1 -1.010e-12 (-1.47e-15)** -4.233e-01 (-2.776)**
Note: .,*,**,*** indicate significance at 90%, 95%, 99%, 99.9%.
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Convergence analysis
ML estimatesForward Backward
Estimate z value Estimate (z value)(Intercept) -5.8301 (-0.838) 25.09524 (2.715)**seragri 0.4779 (0.986) -0.38279 (-0.728)ltu -0.3206 (-0.326) 0.20090 (0.357)popd -0.3812 (-0.874) 0.35340 (0.841)educ 1.4155 (1.808). -1.13022 (-1.561)roadkm -0.2068 (-1.388) 0.19029 (1.170)red 0.7479 (1.822). 0.15476 (0.399)intacc -1.5327 (-1.279) -5.47048 (-3.101)**funds1 0.4583 (2.154)* -0.07864 (-0.353)nms 9.4957 (2.478)* -16.98825 (-0.017)Note: .,*,**,*** indicate significance at 90%, 95%, 99%, 99.9%.
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Outline
1 Introduction and Research QuestionMotivationBackgroundResearch Question
2 The ModelThe Multinomial Response ModelBinary response model
3 ResultsTransition probabilitiesRegression OutputConclusion
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Summary of Results
Regional development in Europe is characterized bydivergence - low mobility of regionsRelative importance of infrastructure, agglomerationeconomies and labor market
Short run vs Long run effectNot only benefits
Strong relevance of Human capital and technologicalinfrastructuresStructural changes in NMSseveral evidences supporting the role of funds
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Summary of Results
Regional development in Europe is characterized bydivergence - low mobility of regionsRelative importance of infrastructure, agglomerationeconomies and labor market
Short run vs Long run effectNot only benefits
Strong relevance of Human capital and technologicalinfrastructuresStructural changes in NMSseveral evidences supporting the role of funds
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Summary of Results
Regional development in Europe is characterized bydivergence - low mobility of regionsRelative importance of infrastructure, agglomerationeconomies and labor market
Short run vs Long run effectNot only benefits
Strong relevance of Human capital and technologicalinfrastructuresStructural changes in NMSseveral evidences supporting the role of funds
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Summary of Results
Regional development in Europe is characterized bydivergence - low mobility of regionsRelative importance of infrastructure, agglomerationeconomies and labor market
Short run vs Long run effectNot only benefits
Strong relevance of Human capital and technologicalinfrastructuresStructural changes in NMSseveral evidences supporting the role of funds
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Summary of Results
Regional development in Europe is characterized bydivergence - low mobility of regionsRelative importance of infrastructure, agglomerationeconomies and labor market
Short run vs Long run effectNot only benefits
Strong relevance of Human capital and technologicalinfrastructuresStructural changes in NMSseveral evidences supporting the role of funds
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Summary of Results
Regional development in Europe is characterized bydivergence - low mobility of regionsRelative importance of infrastructure, agglomerationeconomies and labor market
Short run vs Long run effectNot only benefits
Strong relevance of Human capital and technologicalinfrastructuresStructural changes in NMSseveral evidences supporting the role of funds
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Summary of Results
Regional development in Europe is characterized bydivergence - low mobility of regionsRelative importance of infrastructure, agglomerationeconomies and labor market
Short run vs Long run effectNot only benefits
Strong relevance of Human capital and technologicalinfrastructuresStructural changes in NMSseveral evidences supporting the role of funds
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Policy relevance
Human Capital is the real driver of developmentBe careful with interpretation of infrastrucure
Finland and Sweden have high growth but not so manyinfrastructuresIn NMS infrastructures are important!
Need for a more detailed analysis of funds and cohesionpolicy
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Policy relevance
Human Capital is the real driver of developmentBe careful with interpretation of infrastrucure
Finland and Sweden have high growth but not so manyinfrastructuresIn NMS infrastructures are important!
Need for a more detailed analysis of funds and cohesionpolicy
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Policy relevance
Human Capital is the real driver of developmentBe careful with interpretation of infrastrucure
Finland and Sweden have high growth but not so manyinfrastructuresIn NMS infrastructures are important!
Need for a more detailed analysis of funds and cohesionpolicy
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Policy relevance
Human Capital is the real driver of developmentBe careful with interpretation of infrastrucure
Finland and Sweden have high growth but not so manyinfrastructuresIn NMS infrastructures are important!
Need for a more detailed analysis of funds and cohesionpolicy
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)
Introduction and Research QuestionThe Model
Results
Transition probabilitiesRegression OutputConclusion
Policy relevance
Human Capital is the real driver of developmentBe careful with interpretation of infrastrucure
Finland and Sweden have high growth but not so manyinfrastructuresIn NMS infrastructures are important!
Need for a more detailed analysis of funds and cohesionpolicy
Fabrizi-Guastella-Timpano Determinats of Regionals Convergence (Divergence)