Modelling Migration Histories: Getting started

Modelling Migration Histories
Juliet Harman, Brian Francis and Richard Davies
Centre for Applied Statistics, Lancaster University

The main substantive questions

1 Are some people more likely to move than others?
What factors determine an individual's propensity to migrate? Are there people who are likely never to move?
2 Does an individual's migration behaviour vary with time?
Do people tend to move at certain ages, at particular life events (marriage, children, schooling), for employment opportunities, or as a response to external factors such as the economic climate or the housing market?
3 How can we separate different temporal effects?
Differing patterns of migration behaviour with age are likely for different birth cohorts, as individual life histories take place in different and changing economic conditions. Cumulative inertia effects (the increasing tendency to stay as length of residence in the same place increases) may complicate the variation of migration propensity with age. How can we disentangle the three temporal effects: age, calendar year and duration of stay?

What data set is analysed?

ITEM To address these substantive questions, we need a data set on each of a large number on individuals, with information for each individual on their migration history, their marital history, their employment history and their family history.

ITEM Such historical information is needed from the start of each individual's adult life until the date of data collection.

ITEM We can use the UK Data Archive online catalogue to find a suitable data set. An example has been constructed on how to search for such a data set on migration.

ITEM The data set chosen is a large retrospective survey of life and work histories carried out in 1986 under the Social Change and Economic Life Initiative (SCELI), funded by the ESRC.

Will I understand this module?

ITEM We assume that you have a certain amount of statistical knowledge already. The most important requirement is to be able to understand the output of a multiple regression. A basic knowledge of logistic regression and Poisson regression (regression models for count data) would also be useful, but this is not essential. We provide an explanation of new technical terms, and explain results through the use of graphs.

Give me a quick overview of this module

ITEM We first analyse a summary data set containing the total number of moves for each individual, and demonstrate the limitations of such cross-sectional analysis for drawing inference about the dynamics of migration.

ITEM We then explore the longitudinal data set containing the life and work histories, and model the annual binary migration data using a conventional logistic model. We discuss the limitations of using conventional models for longitudinal data and demonstrate the importance of controlling for individual specific explanatory variables omitted from the analysis.

What software do I need?

ITEM You will need to use SABRE, which is a statistical software package for the analysis of discrete longitudinal data. SABRE runs on all Windows machines and also on UNIX and Linux platforms. SABRE and the teaching data sets can be downloaded from here free of charge.

ITEM SABRE is a specialist package, with a restricted range of commands; it has no facility for instance to plot graphs. However, the parameter estimates from model fitting can be copied into other packages. We use the statistical package GLIM to supplement SABRE.

How do I use this module?

The best way is to follow the module page by page on the Web, loading the data set into SABRE in a new window, and following the instructions onscreen. Alternatively, it is possible to download the entire module as an ADOBE portable document file.

Acknowledgement

This example is based on research work carried out by R. B. Davies and R. Flowerdew (1992) and by Haghighi A. Borhani and R. B. Davies (1999a, 1999b), using data collected under the Social Change and Economic Life Initiative funded by the ESRC. The work by Haghighi Borhani and Davies was partially supported by ESRC research grant L315253007.

NEXT:Table of contents

Home page

1	Are some people more likely to move than others? What factors determine an individual's propensity to migrate? Are there people who are likely never to move?
2	Does an individual's migration behaviour vary with time? Do people tend to move at certain ages, at particular life events (marriage, children, schooling), for employment opportunities, or as a response to external factors such as the economic climate or the housing market?
3	How can we separate different temporal effects? Differing patterns of migration behaviour with age are likely for different birth cohorts, as individual life histories take place in different and changing economic conditions. Cumulative inertia effects (the increasing tendency to stay as length of residence in the same place increases) may complicate the variation of migration propensity with age. How can we disentangle the three temporal effects: age, calendar year and duration of stay?