The Impact of the Salmon Bias on the Migrant Mortality Advantage: New Evidence Using French Pension Data
Michel Guillot, University of Pennsylvania
Myriam Khlat, INED
Irma Elo, University of Pennsylvania
Matthieu Solignac, Univ. Bordeaux; INED
Romeo Gansey, University of Pennsylvania
In the literature on mortality patterns among migrants indeveloped countries, the most systematic finding is the observation thatmigrants tend to exhibit lower mortality than the non-migrant population oftheir host country, a phenomenon termed the Migrant Mortality Advantage (MMA).Among receiving Western countries, this pattern has been reported among migrantgroups in countries as diverse as Australia,1Belgium,2Canada,3France,4,5Germany,6 the Netherlands,7 Switzerland,8the UK,9and US.10 The MMA has been explained using various hypotheses,including in-migration selection effects (healthy migrant effect),out-migration selection effects (salmon bias out-migration of the unhealthy),cultural factors, and data artifacts. However, the relative contribution ofeach of these hypotheses in various contexts remains highly debated in theliterature.
Thispaper focuses on out-migration selection effects as a possible explanation theMMA. Thisexplanation, also referred to as salmon bias, postulates that migrants whoare in poor health may be more likely to return to their country of origin thanhealthier migrants, for reasons ranging from the willingness to seek betterfamily support to the desire to die in ones birthplace.11-18As a result of this unhealthy remigration, the health distribution ofresident migrants will be artificially favored, and in turn mortality ratesamong them will be artificially low. We will examine out-migration selectioneffects using a rich, unique data set from France in which individuals arefollowed worldwide (including when they leave the country) until they die. Thisdata set provides a rare opportunity to directly observe out-migrationselection effects and their impact on the migrant mortality advantage.
Mostof the literature on out-migration selection effects is based on indirectevidence, including analyses of the shape of the mortality curve for migrantsat older ages, or comparisons of mortality risks for migrant groups who are notable to return to their home country vs. migrant groups who are able to return(eg. Cubans vs Mexicans in the US).15 Toour knowledge, only one study (by Turra and Elo)17 wasable to directly observe this mechanism and its impact on mortality rates, focusingon Hispanics in the US, and concluded that some unhealthy return migration isindeed taking place. Our paper contributes to this literature by providing asecond opportunity to directly observe out-migration selection effects. Itbuilds on the Turra & Elo study by: (1) examining the salmon bias in theEuropean context, focusing on France, one of the major receiving countries inEurope; and (2) examining a wider set of countries or origin of migrants,including both more and less developed countries.
Thispaper is based on a longitudinal data set from the Caisse Nationale dAssuranceVieillesse (CNAV), Frances most important pension fund. Thisdata set is a random sample of individuals aged 65 and above who are alive asof December 31, 2008 and receiving a pension from CNAV as of that date. These individuals,who have all been employed in France at some point during their adult life but mayreside in France or abroad at the beginning of the period (12/31/2008), are thenfollowed until December 31, 2014. This follow-up documents changes ofresidence, in France or abroad, during the period of observation. It alsodocuments deaths, in France or abroad, among these individuals, usingregistration information (for deaths in France) or certificate of lifeinformation (for deaths outside France). Pensioners living abroad are indeedrequired to produce a certificate of life once a year, via the local authoritiesand after presentation of proper identification documents, to prove that theyare indeed alive, so that they continue receiving their pension. If anindividual stops producing this certificate, s/he is assumed to be dead and theCNAV stops paying pensions. This sample, which was prepared by the CNAVspecifically for this study, includes 300,000 pensioners born abroad and100,000 pensioners born in France.
Ouranalysis of the CNAV will include both life table analysis and individual-levelsurvival models. We will examine all foreign-born pensioners as well asspecific countries of birth, focusing on the six most important sendingcountries in the French context (Algeria, Morocco, Tunisia, Italy, Spain, Portugal).
Preliminaryresults are presented below in Figure 1. This figure, which pertains to males,compares age-specific mortality rates for foreign-born CNAV pensioners residingoutside France vs. residing in France (using mortality ratios). These resultsdocument a large amount of excess mortality among those foreign-born pensionerswho left France vs. those who stayed. The risk ratio is particularly high(above 1.25) and significant in the age range 65 to 79. These preliminaryresults are indicative of a strong amount of negative out-migration selectionfor foreign-born males in France.
Ratiosof age-specific mortality rates (residing outside France vs. residing inFrance), foreign-born CNAV pensioners, males, 2009-2014
Presented in Session 1185: Mortality and Longevity