We found no evidence to support our hypothesis of increasing mortality among Turkish nationals resident in Germany. On the contrary, they appear to have strikingly and persistently lower all-cause mortality rates than the German population and an urban population in Turkey. These are unexpected findings. Studies from the UK and the US show two to four-fold increased mortality in minorities and socio-economically disadvantaged groups compared to the general population (Davey Smith et al. 1990; Geronimus et al. 1996). Migrant studies have demonstrated a decreasing mortality differential over time (Williams 1993; Parkin 1993). A decreased mortality similar in size to the one we found has been reported in Vietnamese refugees to the UK; it persisted for 5–10 years after arrival (Swerdlow 1991). Thus, our findings may reflect research artefacts, environmental or biological differences between populations, or selection processes.
Environment and genetics
In many studies, socio-economic status was found to be a strong predictor of morbidity and mortality. Turkish residents in Germany have lower job ratings and lower income which would explain a higher rather than the observed lower mortality. In recent years, numerous Turkish residents have started businesses of their own but this has not yet led to a major change in social composition (Sen 1996).
Regarding cardiovascular risk factor levels, Turkish residents might eat a healthier (Mediterranean) diet than Germans, which, by reducing cardiovascular mortality, would result in a reduction of overall mortality. Available figures, however, suggest that risk factors such as smoking are at least as prevalent among Turkish males as among Germans (Bilgin et al. 1994).
The size of the mortality differential suggests that a selection process may be at work. Our data are based on population registers, not on the follow-up of individuals. While this approach is commonly used in migrant studies (Parkin 1993), it cannot satisfactorily account for individuals entering or leaving the population under study. Some conclusions, however, can be drawn regarding selection by healthy migrant effect, healthy worker effect, change of nationality and re-migration in case of grave illness.
Migrants constitute a self-selected population with better-than-average health status (Weber et al. 1990; Swerdlow 1991; Balzi et al. 1993). However, this effect is likely to wear off over time (Williams 1993; Chaturvedi & McKeigue 1994). Before 1973, Turkish ‘Gastarbeiter’ underwent a health exam before being hired, but this was said to be rather superficial and is unlikely to explain the low mortality decades later. Also, family members who followed were not examined, further diluting a healthy migrant effect over the years.
Another obvious explanation for the differential in mortality would be a ‘healthy worker effect’. Workers usually have lower overall death rates than the general population because chronically ill and disabled people are excluded from the workforce (McMichael 1976). Fifteen years after recruitment, this effect would not account for more than a 10% differential (Fox & Collier 1976); it can even reverse in workers living under adverse conditions (Juel 1994). Substantially larger reductions in mortality have only been observed immediately after recruitment (SMR = 0.37) (Fox & Collier 1976), among airline pilots (SMR = 0.63) who undergo regular health check-ups (Band et al. 1996) and among steel workers in Brazil (SMR < 0.5) where there is a huge socio-economic gap between employed and unemployed (Barreto et al. 1996). Neither would apply to Turkish residents in Germany.
The healthy worker effect is an unlikely explanation for the observed differential for additional reasons. First, the effect would attenuate with ageing of the workers (McMichael 1976; Fox & Collier 1976) which we did not observe. Second, the Turkish population in Germany consists not only of employed persons but also of their family members, of unemployed, chronically ill and jobless individuals. Since 1973, the employment rates of Turkish residents and of Germans have converged. In 1988, the employment rate of male Turkish residents (58.9%) was slightly lower than that of German males (60.5%); that of Turkish females (27.5%) was considerably lower than that of German females (37%). Furthermore, 76% of employed Turkish residents but only 44% of employed Germans had a manual job (Cornelsen 1990) which, in other countries, is associated with a higher mortality (Davey Smith et al. 1990). Immigration, at 1% per year in the age group of interest (Statistisches Bundesamt 1980–1994), cannot explain a persisting healthy worker effect.
Naturalization is a relatively small and recent phenomenon among Turkish residents in Germany; double citizenship is currently not a legal option. In 1994, 4067 Turkish nationals above 25 years of age (0.4% of the denominator population) attained German citizenship (Statistisches Bundesamt 1996). Considering the availability of equal medical treatment options for Turkish residents in Germany, it is unlikely that naturalization would be differential with regard to health status.
In 1994, the re-migration rate to Turkey was 2.9% in the age group 25 years and older (Statistisches Bundesamt 1980–1994). To fully explain the mortality differential, 4% of all returnees would have to re-migrate because of grave illness subsequently leading to death. Ensuing mortality rates of returnees would be 10–20 times those of the Turkish population in Germany, which should have raised attention. Re-migration to die at home has been observed in moribund cancer patients. It is unlikely in conditions such as cardiovascular disease for which medical treatment in Germany is readily available and almost free.
While Turkish residents are socio-economically disadvantaged compared to Germans, they are nonetheless an active and productive group (Sen 1996; Collatz 1996). Relative socio-economic differences within the Turkish population in Germany could be more important and smaller than those in comparison with Germans, leading to lower mortality (Wilkinson 1997). In addition, first-generation Turkish migrants may draw positive experiences from migration (Schiffauer 1991), e.g. achieving economic and social success. Integration into familial and religious networks offers potentially effective ways to deal with stress (Antonovsky 1996; Collatz 1996). Thus, Turkish residents in Germany, perhaps more so than Germans, may perceive a sense of cohesion in their lives that is ‘salutogenic’ (Antonovsky 1996), i.e. a protective resource. Success also influences re-migration plans: elderly Turkish residents planning to stay in Germany are significantly better socially integrated and satisfied with their life than those intending to re-migrate (Fabian & Straka 1993).
We thus propose an additional hypothesis to help explain the mortality differential. Unlike emigration to another continent, ‘Gastarbeiter’ type migration within Europe is temporary by intention (if not necessarily so in practice) and the migrants are comparatively mobile. In such migrant populations a continuous selection process might be taking place. Individuals and families who cope well socially and economically stay in the host country; those who fail (and thus might in future experience a higher mortality) re-migrate even before becoming manifestly ill. As a consequence of this (so far hypothetical) ‘unhealthy re-migration effect’, the population remaining in the host country is constantly selected according to low mortality risk. This effect should be discernible in other European countries with similarly dynamic migrant populations. It would have public health repercussions: effects of preventive activities for migrants would be difficult to evaluate in terms of morbidity and mortality, as those at highest risk return before disease becomes manifest.
To learn more about what distinguishes returnees from persons remaining in the host country, research needs to follow up individuals. Reaching a decision to re-migrate is likely to be a complex process. Thus it will be necessary to carry out qualitative studies in order to obtain information on the range of motives for re-migration.
From our data it should not be concluded that there is no need for activities to improve the health of Turkish residents in Germany. Overall mortality figures may fail to reveal excess mortality from specific causes warranting public health action. For many chronic diseases, morbidity and disability are disproportionately higher than mortality. Common health problems such as psychosomatic, psychiatric and musculo-skeletal disorders are not reflected by mortality statistics. Thus, when populations are compared, lower mortality rates do not necessarily indicate better health.