Random-Effects Models for Migration Attractivity and Retentivity: A Bayesian Methodology

Abstract

Several studies have proposed methods for deriving summary scores for describing the in-migrant attractivity of areas, as well as out-migrant push (or conversely migrant retentivity). Simple in-migration and out-migrant rates (migrant totals divided by populations) do not correct for spatial separation or the migration context of a particular area, namely the size and proximity of nearby urban areas with populations at risk of migrating to an area, or offering potential destinations for out-migrants from an area. An extended random-effects gravity model is proposed to represent the influences of attractivity and retentivity after controlling for urban structure. Whereas the existing literature is focused on fixed effects modelling (and classical estimation), the focus in this paper is on a Bayesian hierarchical random-effects approach that links estimates of pull-and-push scores across areas (i.e. allows scores to be spatially correlated) and also allows correlation between attractivity and retentivity within areas. As demonstrated by a case-study of English local authorities, a random-effects model may have a lower effective model dimension than a fixed effects model. Copyright (c) 2010 Royal Statistical Society.

Spatial Interaction

“…These feedback effects are known to require a more elaborate estimation effort (Muth, 1971). The heavy attraction of a city for migrants also presumably pushes up the demand for housing and can make the attractive destination have a higher house price, an effect noted by Congdon (2010). Generally, his results confirm the statistical expectations in Table 2.…”

On Distance and the Spatial Dimension in the Definition of Internal Migration

“…Plane, 1984;Stillwell and Congdon, 1991). In brief, distance is treated as a proxy for variables that are difficult to measure, and by including other variables that are expected to be of importance for migration decision-making, estimates can be made regarding their relative impact on attracting or repelling migrants (for some recent examples see Congdon, 2010;Cooke and Boyle, 2011;Biagi et al, 2011;Kalogirou, 2012). While spatial interaction modeling of migration flows incorporate a 5 measure of distance, it tends to be a very coarse measure, usually given little analytical weight.…”

Estimating Population Attribute Values in a Table: “Get Me Started in” Iterative Proportional Fitting

“…When estimating a contingency table of migration data, linear regression models are often the method of choice, in the form of Poisson regression models (Boyle 1993;Bohara and Krieg 1996) or log-linear regression models (Rogers, Little, and Raymer 2010;Raymer, de Beer, and van der Erf 2011). Similarly, spatial interaction models have a well-established place in the estimation of interaction data (Rees, Fotheringham, and Champion 2004;Congdon 2010) with a very useful introduction provided by Dennett (2012). When estimating a multidimensional age by sex by origin by destination table, van Imhoff et al (1997) experimented with both log-linear modeling and IPF.…”