Mortality of immigrants and their descendants in Britain

Matthew Wallace, University of Stockholm, Hill Kulu, University of St Andrews, Paul Williamson, University of Liverpool and Gemma Catney, Queen's University Belfast

(Project no. 0301579, previously 30157)

The aim of this project is to investigate causes of mortality differences between immigrants, their descendants and the 'native' British population using survival analysis to calculate mortality rates with and without controlling for socio-economic characteristics. Frailty models will be used to detect and control for unobserved heterogeneity. The risk population will be post-war immigrant groups in Britain for those aged 18+ from 1971-2011

Statistical purpose: Matching data sets, individual-level longitudinal data, survival analysis, frailty models

The aim of this project is to examine mortality patterns among immigrants and their descendants in Britain and to investigate causes of mortality differences between immigrants, their descendants and hosts. We will focus on the mortality of the main post-war immigrant groups in the UK: Irish, Indian, Pakistani, Bangladeshi and Afro-Caribbean. We will build upon previous research by using large-scale register data to examine mortality patterns of immigrants and their descendants. We will apply frailty models to gain a better understanding of the expected 'healthy-migrant' effect (i.e. the combination of individual positive health selection and process of acculturation over time). We will also examine the selectivity of return migrants by health status - the 'salmon-bias effect' (i.e. remigration through illness/poor health). Finally, we will analyse the descendants of post-war immigrants; there is little research on mortality of the second-generation largely because of the limited information available. The ONS LS combined with information on place of birth and ethnic identity offers the opportunity to do this. (See research proposal attached in full). Access to individual-level data is vital for the following methodological reasons: first, a set of survival models will be estimated with and without controlling for socioeconomic characteristics of the population; second, frailty models will be fitted and sensitivity of the results to different specification of the baseline hazard (the 'force of mortality') and that of frailty variance will be investigated; third, simultaneous-equations hazard models will be used to detect and control for the effect of selective return migration; and finally, the time at risk for return migrants will be imputed and it will be investigated how sensitive mortality estimates are for undocumented migration.