CeLSIUS

Geographic and social mobility: trajectories of selective sorting by health status and ethnic group over time

Fran Darlington and Paul Norman, University of Leeds and Dimitris Ballas, University of Sheffield

(Project no. 0301634, previously 30163)

This project will track the changing health status and social and residential circumstances of individuals over time with the intent of exploring whether prior health status sorts people into different social groups and areas and, whether this differs by ethnicity. Planned statistical methods include logistic regression and multilevel modelling. 

It is widely documented that health follows social and spatial gradients within the UK at the macro-spatial scale. These gradients are demonstrated by a wealth of empirical evidence utilising a wide range of datasets, including the ONS LS. However, despite significant academic and public interest in the determinants of inequalities in health, there is convincing evidence to suggest that these gradients are widening. An incomplete understanding of what drives changing health gradients can only serve to undermine attempts to close this health gap, consequently, further research seeking to more fully understand what contributes to changing health gradients and how this operates in contemporary society is required. 

Herein lies the intent of this project: to contribute to the growing evidence-base which seeks to understand why and how health gradients change over time, and crucially, how the processes driving these changes vary across the population. While analysis of the LS will constitute the main analytical section for this PhD thesis, as a census-based piece of empirical work it will also make use of the Samples of Anonymised Records (SARs). However, the SARs do not allow one to track changes to an individual’s life over time and therefore while necessary for the research, are by no means sufficient. Analysis of both the SARs and the LS will be further supplemented by use of the Health Survey for England and a local birth cohort study, the Born in Bradford Cohort.  

An under-researched aspect in population and health research is that, over time, changing health gradients may be attributable to a process of selective sorting between social groups and areas. This questions whether prior health status functions as a sorting mechanism, selectively distributing healthy or unhealthy individuals between different social groups and area-types. For example, are healthy individuals more likely to move up the social hierarchy or move into more advantaged areas than unhealthy individuals? Consequently, this research is specifically concerned with social class, deprivation, residential mobility and health as measured by morbidity and mortality. However, another dimension will also be considered: ethnicity. 

The increasing diversity of contemporary society means ethnicity is an ever-more salient issue in population research. Furthermore, as health also varies by ethnicity, and these variations cannot be explained away by genetic or cultural differences, considering ethnicity in research into inequalities in health is important. As different ethnic groups are distinctive in their experiences of the social system, the types of areas in which they live and their residential mobility patterns, it is likely that their experiences of selective sorting will also be distinctive. 

As such, the broad aims of the overall research project are to:

(a) explore the association between changes in health, social status and place;

(b) investigate the extent to which health functions as a selection mechanism, sorting people of different health status between area-types and social classes; 

(c) analyse whether there are increased health (dis)benefits resulting from simultaneous (un)favourable social and residential mobility;

(d) assess whether this contributes to changing health gradients; and

(e) determine whether the  process of selective sorting or the resultant likelihood of increased health (dis)benefits varies by ethnic group.

In order to fully explore how changing health and social and residential circumstances interact with each other, and whether these interactions differ by ethnicity, it is essential to be able to track individual’s movements through space, society and time. The success of this study is therefore contingent on the ability to identify an individual’s social status, experience of area-based deprivation, geographic location and ethnicity at different points in time, i.e. the 1991, 2001 and 2011 (when available) censuses. As such, the ONS LS represents an ideal data source, providing both the coverage and detail required to gain meaningful results and fully account for the social and residential mobility of individuals while simultaneously analysing their health status and ethnicity.

It should also be noted that age is an important feature of this research insofar as it is recognised that health, propensity for mobility and experiences of the social system vary significantly with age. For example, the importance of prior health status as a selection mechanism may vary by age as well as ethnicity. This will be considered within the analysis.

Firstly, transition categories will be defined for individuals distinguishing change or stability in social and residential circumstances over time. For example, have mobile individuals moved to differently deprived areas or, have immobile individuals experienced changing deprivation. Similarly, have individuals moved up or down the social hierarchy as measured by social class. Standardised mortality and illness ratios will be calculated for each transition category by ethnic group. Incidentally, while previous studies using the LS which investigate selective migration tend to exclude those in poorer health at the start of the study period, no such exclusions will be made within this analysis. The experiences of these groups may provide important insights into the process of selective sorting and its subsequent contribution to changing health gradients. 

Secondly, a combination of binary logistic regression and multinomial or ordinal logistic regression (depending on the data type of the outcome variable of interest) will be used with the intent of quantifying the probability of (un)favourable social or residential mobility by ethnic group and health status, and, to try and quantify the extent of the effect of selective sorting on changing health gradients. It should be noted that while the broad aims for this analysis are clear, final decisions as to the most appropriate statistical methods have not yet been reached. It is likely, however, that logistic regression will be the principle method of analysis. 

Page last modified on 28 jul 16 15:45 by Joanne Tomlinson