Spatial dependence in unemployment in Great Britain

In this study I find strong evidence that there is spatial dependence in unemployment in Great Britain. Spatial dependence is a situation where outcomes in a given area appear to depend on outcomes or other factors elsewhere and my study suggests that spatial dependencies or spatial spillovers are present in short-term, long-term and total unemployment in Great Britain. One of the key results which has helped to establish this conclusion is that high levels of human capital in a region are associated with lower unemployment rates, not only in that region but in neighbouring regions also. This is a new result as although past research has examined unemployment in Great Britain for spatial dependence, my study is the first to attempt to examine the nature of these dependencies rather than just establish their presence. The other insights from my research include that assuming that dependencies exist between regions up to 50km in distance from each other appears to give a better approximation of reality than when 25km, 100km, 200km or 300km are used as thresholds up to which dependencies are present. My study also suggests that the dependencies apparent in short-term, long-term and total unemployment are very similar. These results are of clear societal relevance as they point towards two major policy suggestions. The first is that public authorities governing proximate regions, particularly those within 50km of each other, should collaborate on unemployment reduction policies so that beneficial spatial spillovers are accounted for in decision making. The second is that human capital improvements appear to be a beneficial policy option when dealing with unemployment, especially if combined with the collaboration also suggested. My study was able to yield such relevant results by establishing a strong theoretical framework and designing a robust methodology. My theoretical framework made the case that as job-seekers carry out multi-regional job searches resulting in job-matches which, other things being equal, lower unemployment, there will be dependencies in unemployment between regions. I also argued that the job-search behaviour of a job-seeker in a particular region is affected by their distance to this region, the attractiveness of working in this region and their perceived probabilities of getting a job in the region. This argument allowed me to establish a set of factors to include in my model by focussing on what could affect attractiveness and perceived success probabilities while remembering the role of distance. 4 The final element in my theoretical framework was the case I made for spatial dependencies to potentially differ for short-term and long-term unemployment. My argument for this adjustment was that job-search behaviour differs between the short- and long-term unemployed which may lead to differing dependencies in short- and long-term unemployment given that job-search helps explain these dependencies. This theoretical framework helped me to assemble an appropriate dataset and design a suitable methodology. This methodology consisted of exploratory spatial analysis using the global and local versions of Moran's I which allowed me to demonstrate that unemployment in Great Britain is positively spatially correlated meaning proximate areas have similar unemployment rates. This correlation was apparent when considering the whole of Great Britain but appeared to be particularly driven by four large clusters three of which were high unemployment clusters in the North while the fourth was a Southern, low unemployment cluster. Following this exploratory analysis I designed and carried out some econometric analysis. Baseline results using non-spatial panel models provided initial evidence regarding the links between human capital and unemployment but this was expanded on by the estimation of spatial models. A Spatial Durbin Model specification was used as this closely matched my theoretical model of unemployment and was statistically supported by the results of a number of tests. The Spatial Durbin Models I estimated were able to provide evidence for the spatial dependencies I set out to examine which in turn allowed me to offer the conclusions and policy recommendations mentioned above. Full details can be found in the rest of my study which should be of relevance to anyone interested in unemployment in Great Britain or in spatial analysis in general.