Four essays on spatial modelling and welfare analysis

Abstract: There are two different econometric approaches used in thisthesis, random utility econometrics and spatialeconometrics. Random utility econometrics has become an increasinglyimportant tool in transportation, labour economics, housingeconomics, non-market valuation, resource and environmentaleconomics. In a discrete choice context the convenientapproach of using a fictious representative individual forwelfare assessment is inappropriate when the marginal utilityof money is not constant across alternatives. When marginalutility of money is not constant, it has been recognised thatthere have existed no exact methods to calculate hicksianwelfare measures. In the first essay of this thesis, an exactformula is derived which can be used for most real-worldapplications. In the second essay, the role of income in randomutility modelling is reviewed, and it is argued that theproblem has been obscured in the transportation economicsliterature as the choice settings implied have not always beenclearly defined. It is also shown that although Slutskysymmetry is not fulfilled, there exists a ``quasi-demand''function that can be integrated to establish path independenceof the hicksian welfare measures. Local indicators of spatial association have become animportant tool in spatial data analysis, but they are rarelyapplied to interaction flowdata. In the second essay a spatialassociation statistic for flow data is developed bygeneralizing the statistic of Getis-Ord, Gi(and Gi') This local measure of spatial association, Gij, is associated with each origin destination pair.Measures which control for generation or attractionnonstationarity are also defined. The measures are implementedto examine the spatial association of residuals from twodifferent models. In contrast to the normal use of the Gistatistic, the normal approximation is oftenappropriate, but the statistics are still correlated. We alsodefine a test statistic for barrier effects, which is comparedwith the traditional dummy approach for barrier effects. Weargue that the explorative analysis of interaction flow dataadd to our understanding of the model. Spatial econometric methodology has been rarely applied toelectoral data. In the fourth essay we develop a game-theoretic model toinvestigate the political effect of spillovers in a localpublic choice context. It is shown that if there existstrategic interaction among jurisdictions, it may be beneficialto delegate power to a representative to set public spending.As a consequence, public spending levels may deviate in arepresentative and a direct democracy, without anymisrepresentation of preferences. This is in contrast a commonconclusion drawn from empirical investigations on local publicspending data. Spatial econometric methods are used in anexplorative analysis to find evidence of spatial effects inecological electoral data of split voting. The empiricalanalysis has its own value, since spatial data analysis methodshave rarely been applied to electoral data, and never toecological split voting data.