The statistical analysis and modelling of internal migration flows within England and Wales

Abstract

Migration flows in recent decades suggest that Britain is a nation of people on the move. The combination of information technology and structured data collection allow today a very close and detailed examination of trends in internal migration in England and Wales. However, there has been a relative dearth in the analysis of migration data using quantitative techniques. In this thesis I attempt to analyse migration patterns and to model migration moves in order to explain the main factors affecting individuals' migration decisions. I try to link this work to existing research in this research field by revising and applying recently developed quantitative methods. My main aim in this thesis is to provide empirical evidence that the effects of many socio-economic factors on individuals' migration decisions are non-stationary across space. More specifically, there are four sections of data analysis in this thesis: the exploration of migration flows using data visualization and local statistics; the analysis of the effects of socio-economic factors on out-migration rates; the analysis of the attraction of migrants from areas with varying socio-economic profiles via the construction of global and local models; and the examination of model residuals. The main objectives of this work are twofold: the first is to provide a thorough investigation of internal migration using a rich dataset on annual migration during the 1980s and the 1990s; the second is to remove the inaccuracy traditional global models introduce by assuming that the processes being examined are stationary over space. I do this through the use of newly developed local statistical methods. Former attempts made to provide local forms of statistical analysis have limitations. Geographically Weighted Regression is used in this thesis to allow for local modelling. This method provides not only a technique for best model fit but also for the evaluation of the results using modem goodness of fit statistics such as the Akaike Information Criterion. The temporal dimension of my data allows the examination of the stability in migration flows over time. It also provides a means of checking the consistency of the significance of the spatial variation of local parameter estimates derived from each annual set of migration data. The migration data themselves are disaggregated in 14 sex/age groups.The age disaggregation reflects the stage of life individuals are at (e.g . people forming a family are young adults aged 25 - 29 years old). In order to facilitate the examination of migratory moves over time, I introduce a new way of visualising in-, out- and net migration rates, the heat map. The results for the migration models show that the parameter estimates of some of the migration deterininants exhibit significant spatial variation. This suggests that the effect of some determinants on migration decisions in both origin and destination of a migratory move vary across space. The spatial patterns of the local parameter estimates usually show a North-South or a Northwest - Southeast divide. When out-migration models are concerned, there is strong evidence for a spatially variable effect of employment rate for all migrant groups and percentage non-white population along with percentage long-distance commuters for mature male adults. When destination choice models are concerned, there is strong evidence for a spatially variable effect of destination accessibility, house prices, listed buildings, vacant and derelict dwellings, distance and total population. These new findings on local migration modelling are of high interest and potential benefit to policy makers. The spatial and temporal migration trends confirm the continuation of the counterurbanisation phenomenon in England and Wales. The local out-migration models suggest the effect of some ecological conditions on out-migration is associated with the location of the origin. The local destination choice models suggest that there are differences on what determines short migration moves and longer moves. They also suggest that the behaviour of those leaving an area is not stationary for all England and Wales. Finally, similarly to out-migration, there are instances where the effect of some ecological conditions on destination choice is associated with the location of the destination. This thesis also presents an attempt for constructing more robust migration models, signalling the need for additional migration determinants.