The Effects of Immigrant Status and Age at Migration on Changes in - - PDF document

1

The Effects of Immigrant Status and Age at Migration on Changes in Older Europeans’ Health

Introduction During the last half-century, two important processes, international migration and demographic ageing, have greatly influenced the age structure and ethnic composition of national populations in

• Europe. Many immigrants from the early post-World War II period did not go back to their

countries of origin, as expected, but settled down definitively and aged in their countries of destination, mainly in Northwest Europe. This phenomenon has recently elicited interest in the ageing and health status of immigrant populations due to the demographic significance of their middle-aged and older members, which has rapidly increased in the last few decades. From 2008 to 2015, the number of elderly people living in Europe who have aged in countries other than where they were born increased from 7 million to 15 million (AAMEE 2008). In Germany, the most important “old” immigration country, annual registration data show that the older foreign-born population aged 60 or over has doubled roughly every 10 years, from 80,000 in 1970 to 160,000 in 1980, 320,000 in 1992 (after a short- term reduction in the late 1980s), and nearly 670,000 in 2001. In addition, the proportion of older people will increase in the coming decades (White 2006). Another reason for interest in foreign- born people is that a growing body of research has shown that the health status of older adults living in Europe is partly determined by their “immigration status,” since some immigrant groups tend to have poorer health later in life than native-born people (Silveira and Ebrahim 1998; Pudaric et al. 2003; Solé-Aurò and Crimmins 2008; Leão et al. 2009; Vaillant and Wolff 2010). Other studies have documented great variations in the ageing process among immigrants, indicating that certain groups undergo successful ageing, whereas others do not (e.g., Lanari et al. 2015).

2

The “health vulnerability” of immigrants aged 50 and over living in Northern and Western Europe has recently been highlighted (Lanari and Bussini 2012). The authors found that some immigrant groups are more likely to perceive and self-rate worse health and to suffer from depression more than native-born groups, even when demographic and socio-economic variables are taken into account. In particular, the facts of being born and living in a specific country, in addition to duration of residence and citizenship, give rise to an increased health status risk in particular immigrant groups. For example, people born in Eastern Europe living in Germany, France, and Sweden have poorer health than native-born people. A rapid deterioration in the health status of immigrants from Eastern Europe living in Germany has been shown by Ronellenfitsch and Razum (2004), despite their initial health advantages on arrival and improved socio-economic status

• ver time. Other studies conducted in the United States have shown that, according to the

“immigrant health paradox” literature, even when immigrants are relatively healthy when they arrive, this advantage decreases over time and eventually disappears, as some immigrants undergo a relatively fast decline in health and end up being disadvantaged in later life. Specifically, despite a health advantage at age 50, Hispanic young adult immigrants suffer a steeper decline in self-rated health afterward, whereas both non-Hispanic and Hispanic immigrants who migrated in late adulthood undergo much faster health declines in old age (Guberskaya 2014). Socio-economic disadvantages, cultural and linguistic barriers, unequal access to healthcare and social services, discrimination, the psychological stress of living in a new environment, and the lack

• f social and family relationships are all factors that can explain the increased risk of perceiving

worse health among foreign-born groups compared to the majority of the population (Ringbäck et

• al. 1999; Silveira et al. 2002; Ronellenfitsch and Razum 2004). In addition, according to the theory
• f cumulative disadvantage (Dowd and Bengtson 1978), the successive addition of adverse

circumstances over time —such as social and economic disadvantages— does not promote successful ageing but the opposite, potentially resulting in the onset of poor mental and physical health, which deteriorates with length of residence. Also, other research on inequalities in women’s

3

health in the United States support the “weathering” theory, according to which, early health deterioration of African women compared to native-born women is a consequence of the cumulative impact of repeated experience with social, economic, or political exclusion (Geronimus 2001). Other studies focused on the role of negative acculturation, reporting that “greater” degrees of acculturation were associated with problematic health outcomes (Alegría et al. 2008, 2007). The negative acculturation theory places significant importance on changes concerning immigrants’ tastes and preferences (worsening of dietary styles), adoption of risky behaviors (consumption of tobacco and alcohol and lack of physical exercise), and environmental exposure (living conditions) becoming more similar to that of the native population for adapting in the host country (Ceballos and Palloni 2010; Antecol and Bedard 2006; Abraido-Lanza et al. 2005). Additional explanations for the deterioration of immigrants’ health over time may be related to the political and economic context of the host country (racism, xenophobia, poor living conditions) and the disruption of religious and family ties, which could serve as important sources of support and protection against stressors (Viruell-Fuentes 2007; Silveira and Ebrahim 1998). Socio-economic factors in the host country, such as loss of social status and poor working conditions, may contribute to explaining immigrants’ health worsening, since many ageing migrants entered a country with little education and were employed in low-skilled and low-paid manual work (Warnes et al. 2004). Consequently, the higher probability for foreign-born people to be employed in jobs that are more dangerous or to perform riskier tasks than natives could result in more frequent fatalities and work-related injuries

• r other health problems that may explain deterioration in health. For minority populations such as

asylum seekers and refugees, other factors related to the situation in their countries of birth may influence health since they may have experienced trauma, war, discrimination, and poverty (Akhtar 1999). This may be the case for individuals who migrate as adolescents or adults, who are likely to have vivid memories of life prior to migration compared to those who migrated as young children (Portes and Rumbaut 2006). These results emphasize the importance of considering the various

4

factors related to the social and political environments in their areas of origin and the receiving countries in explaining the health disadvantage of immigrants. Another hypothesis is that foreign- born people are less likely to have adequate healthcare coverage or familiarity with and established connections to healthcare systems (Carrasquillo et al. 2000). In general, migrants suffer lower access to specialist and preventive care and higher usage of emergency departments (Morris et al. 2005; Gravelle et al. 2003; Cots et al. 2007). In addition, the lack of language skills can be a great barrier to a proper understanding of the health system and may lead to later diagnosis and less

• ptimal choice of treatment (Davies et al. 2006). Further evidence shows that experience with

discrimination is a decisive factor in access to healthcare services (Agudelo-Suárez et al. 2009). In view of the above demographic trends and the higher probability of some immigrant groups facing health disadvantages, changes in various aspects of their health status during ageing have become a central concern for policy-makers. On one hand, the rising proportion of older people is placing further pressure on the overall healthcare spending and welfare systems of the host countries; on the other, better understanding of immigrants’ health status and behavior is needed so that clearly defined policy measures can be adopted and relevant healthcare services can be

• planned. It is equally important to understand how the effects of immigrant status on health may

vary by country of origin and age at migration. Patterns of migration flows in Europe have changed

• ver time; the size and composition of migrant populations reflects both current and historical

patterns of migration flows. Many immigrants moved either from south to north or from east to west within Europe, mainly for economic or political reasons and limited opportunities in the country of origin. Others came from regions of similarly restricted opportunities, such as North Africa and South-East Asia after the independence of former colonies (Fassmann and Münz 1992). These international migrants have cultural, religious, and socio-economic backgrounds that differ from those of the host populations and that may influence their overall health. Another important key variable related to immigration status is “age at migration,” because it captures the length of exposure in both countries of origin and destination and the ability to

5

maintain good health in old age. For example, since most immigrants from less-developed areas suffered longer exposure to unfavorable living conditions in their countries of origin, this may have caused a health disadvantage for the foreign-born who migrated during adulthood. Younger immigrants who did not encounter these difficulties were more likely to better integrate into their new society and acquire the socio-economic resources or upward mobility necessary to maintain good health during ageing. This paper is an extension of the results of our previous study in that it also examines health trajectories in order to ascertain whether the health disadvantages faced by some immigrant groups increase with age or not among the elderly population when compared to native-born people. Specifically, this study examines the probabilities of transition among health states in self-rated health (SRH), depression (DEP), and activities of daily living (ADL) in middle-aged and older adults living in Europe and how these transitions vary according to immigration-related determinants (country of origin and age at migration). The longitudinal dimension acquired by the Survey of Health, Aging, and Retirement (SHARE) allows us to extend our research on health inequalities to make up for the relatively scarce knowledge of how successful immigrants’ ageing is compared to that of native-born people. Most studies applying SHARE data focus on the health transitions of old people without distinguishing between native-born and foreign-born (see, for example, Verropoulou 2012; Hank et al. 2013; Adena and Myck 2014). Furthermore, since panel surveys suffer from a certain degree of sample attrition between waves, a phenomenon that may bias our transition probabilities, we also conducted a robustness check on this matter. To the best of our knowledge, this is the first time that the longitudinal framework has been used to estimate both declining and recovering health in the two subgroups. In view of the theoretical and practical background described above, we hypothesize that some immigrant groups are more likely to undergo deterioration in health during ageing; that is, the probability that a person with certain characteristics (immigrant born in a specific macro-area) will move from a “healthy state” to a “sick state” is higher when compared to others (non-immigrants). Conversely, we assume that the

6

probability of moving from poor to good health (resolution of sickness) is higher for native-born people. Data and Methods Data Transitions in health status were analyzed with data from SHARE, which included, among others, socio-demographic and economic characteristics, self-rated health indicators, and at-risk behavior by non-institutionalized individuals aged 50 and over1. Three waves of the SHARE survey collected every two years, starting from 2004 to 2010, were used, the first two (collected respectively in 2004/2005 and 2006/2007) and the fourth (2010/2011), since the third wave is SHARELIFE and focuses exclusively on people's life histories. Consequently, the probability of transition from one health state to another can be estimated over a 2- to 6-year interval, a considerable time lag in which substantial changes in health may occur. Our data cover the period 2004–2011 and include individuals who were part of the sample of countries selected in 2004 (our initial state) and who were residing in various regions of Europe from Scandinavian countries (Denmark and Sweden) to Northwestern Europe (Austria, France, Germany, Switzerland, Belgium, The Netherlands) and the

1The SHARE target population consists of all persons aged 50 years and over at the time of

sampling who have their regular domicile in the respective SHARE country. A person is excluded if she or he is incarcerated, hospitalized, or out of the country during the entire survey period, unable to speak the country’s language(s), or has moved to an unknown address. Given the general purpose

• f SHARE and that the prevalence of institutions for elderly differs between European countries, it

was desirable that elderly living in institutions were not explicitly excluded from the research population (Börsch-Supan and Jürges 2005). In some countries like Denmark, Germany, the Netherlands, and Sweden, this became possible; in others it was not (Belgium, France, Greece, Italy, Austria and Spain) (Klevmarken et al. 2005).

7

South (Spain, Italy, Greece)2. Some of these countries, such as Germany, France, Switzerland, Belgium, and Sweden, became the most important European receiving countries for immigrants. The three Mediterranean countries were also considered, although the proportion of middle-old immigrants was negligible because these three countries have only become subject to immigration since the 1990s (Bonifazi et al. 2009). However, we decided to keep these countries in the sample in

• rder to have the largest number of respondents. A robustness check was also implemented to take

possible bias into account, excluding the three Mediterranean countries. The baseline survey included 27,061 persons, of whom 2,194 (about 8%) were defined as “immigrants” born in a country different from that of their residence (Table 1). A restricted panel sample of 19,935 individuals with all complete information at waves 1 and subsequent waves (W2 and W4), was used for analysis. Information on place of birth and age at migration allowed us to test the different influence of “being immigrant” on health transitions. Three broad immigrant groups were distinguished independently of citizenship: immigrants from Eastern Europe (East- EU), other European countries (Other-EU, (e.g., Western, Northern, and Southern Europe)), and non-European countries (Extra-EU). Classifying countries into geographical regions followed the method used by the Population Division of the United Nations (United Nations 2008). In defining these groups, we tried to obtain a sufficient number of observations in each subgroup and, in particular, to examine the health transition of immigrants from Eastern Europe, the most vulnerable

• group. We also had information on 1,490 individuals who died between waves 1 and 4.

Health measures

2We did not consider Israel in the original dataset, as it is not a European country. It is important to

point out that Greece did not participate in wave 4, so for this country, the analysis is related to the period 2004–2007.

8

Our health measures were SRH, DEP, and ADL. Following the approach of Thielke and Diehr (2012), each health variable was dichotomized into “Healthy” and “Sick,” with cut-offs validated in the international scientific literature. The “US global version” was used to examine self-rated health, since the very similarly constructed European version was unavailable for subsequent waves

• f the survey. Individuals who reported excellent, very good, or good health were classified as

“Healthy” and those who reported fair or poor health as “Sick.” SRH is one of the most widely used indicators of health in survey research and a useful tool for identifying individuals and groups at risk for poor health and for monitoring health changes in populations (de Bruin et al. 1996). The advantages of using a subjective general health measure include its simplicity, ease of availability in most health surveys, capacity to provide a holistic approach to the concept of health, and reduced burden and costs. SRH is also a strong predictor of mortality patterns (Idler and Benyamini 1997), use of healthcare services, and future health of older people (Bath 1999). However, SRH is a subjective assessment and, even with agreement on the structure and wording of the question, it is likely that answers will be less precise and culturally influenced across differing ethnic groups, compared to more objective health measures (Robine et. al.2003). Depression (DEP) was measured by self-reports according to the Euro-D scale (from 0 to 12), counting whether the individual reported having problems in a list of negative feelings such as having sad or depressed moods, pessimism, guilt, suffering from lack of concentration and interest, sleeping disorders, fatigue, irritability, loss of appetite, lack of enjoyment, tearfulness, or thoughts of suicide3. Individuals’ answers were then set in two categories: three or more symptoms, or less than three. This cut-off point had been validated in an earlier cross-European study of depression prevalence (EURODEP)

3The importance of measuring mental health is evidenced by Robine et al. (2003, p. 8): “Population

indicators of ‘mental health’ are being developed since mental disorders are felt to be underreported, under-diagnosed and under-treated and are now recognized as one of the principal causes of disability, consuming a significant proportion of the health budget in Western countries.”

9

against other clinically significant indicators. People reporting three or more depressive symptoms were likely to be diagnosed as suffering from depression and thus defined as “Sick,” for whom medical intervention would be desirable (Prince et al. 1999a,b). ADL are activities related to personal care which include bathing or showering, dressing, getting in or out of bed or a chair, using the toilet, eating, and getting around inside the home. For limitations in ADL, people who suffered from one or more impairments were considered to be “Sick” in that domain. The motivation to examine ADL (1+) is the smaller samples size of respondents with more than one ADL limitation when analyzing data per country of origin or age at migration. However, we conducted a robustness check for ADL, setting a higher threshold of three or more limitations (e.g., Adena and Myck 2014; Leveille et al., 2000). For all three indicators of health, modality 1 was assigned to the “Sick” dimension and modality 0 for the “Healthy” one. Independent Variables The independent variable of interest was the migration background indicator, which combines the status of immigrants with their countries of origin and age on arrival in the destination country, independent of citizenship. In particular, interaction variables between immigrants’ countries of

• rigin and age at migration were created, since the great heterogeneity of immigrants affects

economic development, cultural, religious, and at-risk behavior and may consequently influence their overall health status. Three immigrant groups were distinguished from the native-born people serving as the reference group: immigrants from Eastern Europe (East-EU), Other Europe (Other- EU), and Extra-Europe (Extra-EU). We also considered the proportion of life that immigrants had spent in the host country in three “age at migration categories,” with the native-born people serving as the reference (Angel et al. 2001). These categories were: people who had migrated before the age

• f 15 (childhood), between ages 15–34 (young adulthood), and those who had arrived in their new

country at age 35 and over (mature adulthood). This classification aimed at reflecting the different

• pportunities for social, economic, and cultural integration for people who came to their new

10

country earlier in life, accelerated by the possibility of schooling in the host country, compared to those who arrived during adulthood. The following socio-demographic variables were included to serve as controls. We included gender (reference category male), age (as a continuous variable), and marital status (married as reference). In order to examine how socio-economic status (SES) influences health status, we controlled for educational level and the perceived economic condition of household. Educational level was based

• n self-reporting of the highest level of education and reclassified according to the UNESCO

International Standard Classification of Education (ISCED) to homogenize education systems across countries (UNESCO 1997). The original ISCEDs were recoded into three broader levels: “low” (pre-primary to lower secondary education), “medium” (upper secondary and post-secondary, non-tertiary education), and “high” (first and second stages of tertiary education). The last was used as the modality reference. The perceived economic conditions of household were built starting from responses to the question: “Is your household able to make ends meet?” which included four decreasing modalities of the difficulty to meet needs with respect to monthly household income (easily, fairly easily, with some difficulty, with great difficulty). This categorical variable was then reclassified into two modalities of perceived economic resources: high/intermediate (reference category) and low (with some or great difficulty). Social support received from outside the household in the 12 months prior the interview was also taken into account. Social support refers here to the help received from a member of the family, friend, or neighbor for personal care, practical household help, or help with paperwork. We also ascertained risky health behavior, smoking, and alcohol consumption. The variable representing smoking habits dealt with whether a respondent was a regular smoker or was a non-smoker (i.e., never smoked daily for at least one year). Drinking behavior was based on a binary indicator showing whether the respondent had been drinking more than two glasses of alcoholic beverages almost every day or at least 5 or 6 days a week for the 6 months prior to the interview. The number of years between baseline and follow-up measurements of health was also considered, since the probability of experiencing any health

11

change should increase with the time elapsed between SHARE interviews (see Hank et al. 2013). Lastly, we included country-fixed effects in the analyses of health status in order to take into account the specificity of the respondents’ country of residence. A detailed description of variables is reported in Appendix, Table A.1. Statistical Analysis Transition probability models were used to assess the prevalence and incidence of healthy and sick states for each of the three health measures (SRH, DEP, and ADL) for various groups distinguished according to immigrant status (native-born/foreign-born), country of origin (East-EU, Other-EU, and Extra-EU), and age at migration (childhood/young adulthood/mature adulthood). In order to describe and compare prevalence, we first calculated the prevalence of the healthy state as the percentage of people who declared that they were healthy. We then estimated transition probabilities as changes to healthy/sick states (or vice versa), conditional on being in a good or bad state at the starting time-point. Note that these probabilities are conditioned on survival since only surviving individuals add observations to the probability model. In order to make predictions comparable to the underlying nonparametric counting estimates, we augmented the two health states with an additional third state: “Dead.” This is an absorbing state. The natural ordering of the health states motivated us to estimate the transition health status models with an ordered logit model. Additionally, we were interested in analyzing the determinants of the health transition probability

• f immigrant groups compared to native-born groups (i.e., declining and recovering health) and

evaluating their statistical significance. For this purpose, we then used logistic regressions to estimate the odds of moving from healthy to sick status and vice versa for each health indicator used, stratified by the immigrants’ countries of origin and conditioning the vector of characteristics such as socio-demographic variables (age, gender, marital status, education, and perceived economic conditions), at-risk behavior (drinking and smoking habits), social support (help received

12

from outside the household), time between interviews, and country-fixed effects. This strategy has also been used to evaluate changes among older adults, including self-rated health and depression (Diehr and Patrick 2001; Thielke et al. 2010). Formally, the equations of the estimated model are given as:

{ 𝐛. 𝛒 (yWt = 1 |yW1 = 0) = 𝐆 (𝛽XW1 + 𝛾ImW1) 𝐜. 𝛒 (yWt = 0 |yW1 = 1) = 𝐆 (𝛽XW1 + 𝛾ImW1) with Wt = W2, W4 (1)

where, in the first probability model, yW1 = 0 stands for being in the healthy state in wave 1, and yWt = 1 indicates being in the sick state in waves 2 or 4 (i.e., declining health). XW1 is a vector of controls measured at the time of wave 1, and ImW1 is a stratified variable of the immigrants’ countries of origin defined at the time of wave 1. Function F(∙), which in our estimations is a logistic function, takes values between 0 and 1. Conversely, equation 1.b models the transition from sick to healthy states of immigrants compared to natives (i.e., recovering health). We also extended the framework in equation (1), modeling the interactions between country of

• rigin and age at migration, to explore more in depth the effect of immigrant-status variables on
• health. Lastly, we examined various other specifications to test the robustness of our results.

Results Descriptive findings Table 1 lists descriptive statistics and the prevalence of a healthy state by birth status for the three indicators used (SRH, DEP, and ADL). All the prevalence values of each variable were calculated as the percentage of persons who stated they were healthy, as defined above. The native-born people had a significantly higher prevalence of healthy states. For example, for SRH, about 68% of the native-born people were “Healthy,” since they stated they were in “good, very good, or

13

excellent” health. For immigrants, the prevalence of a healthy state was lower at 62.2%, but more importantly, the health conditions of immigrants born in Eastern Europe were the worst when compared to both native-born and other groups of immigrants: only about 54% stated that they were "Healthy.” The same pattern was found for DEP, the native-born people having a significantly higher prevalence of a healthy state (not depressed) compared to immigrants (76.2% versus 71.9%), showing differences from 3 to 7 percentage points, when the different countries of origin of the immigrants were taken into account. There were no significant differences between native-born people and immigrants for ADL. The last line of Table 1 shows the number of observations (transition pairs). (Table 1 about here) Table 2 shows the estimated probabilities of transition in health status for each measure analyzed for native-born people and immigrants, using ordered logit specifications that include the control variables listed in Appendix (Table A.1). We also reported confidence intervals for inference issues. The first two lines show the transition probabilities for the native-born who were initially healthy or sick; the second part of the table refers to the whole sample of the foreign-born, divided into three immigrant subgroups in order to examine the role of country of birth on health transitions. For example, in regard to the self-rated health indicator, the native-born who were healthy (who reported excellent, very good, or good health) had a 79.44% probability of remaining healthy, 18.11% of becoming sick (poor or fair health), and 2.44% of dying. For immigrants, the probability

• f staying healthy was much lower (74.6%), and that of moving from a healthy to a sick state was

23.75%, more than 5 percentage points higher than that of native-born people. In particular, the data showed that East-EU people had a higher risk of health deterioration when compared to the native- born and other immigrant groups, since only about 70.7% of them remained in excellent/very good/good health but 28.24% moved toward the “Sick” category, the highest value recorded.

14

Confidence intervals at 95% indicate that the differences between native-born people and immigrants or sub-groups are statistically significant. Conversely, the transition probabilities for the native-born people who were initially sick showed a higher probability of recovery. Specifically, the probability of moving from a sick to a healthy status for the native-born was 28.91% versus 20.33% for immigrants, a statistically significant result. In regard to DEP, in the two transitions of remaining healthy and becoming sick, the estimated transition probabilities showed a deterioration in health, in particular for immigrants from East-EU, characterized by the lowest percentages of people staying healthy (81.2%) and highest proportions

• f becoming sick (17.9%). In this case, the statistical inference suggests that the estimated

probabilities for native-born people are not different from those of immigrants. However, the transition probabilities for persons who were initially sick indicated that the probability of recovering from sickness was lower for immigrants from Other-EU and significantly different when compared to Extra-EU immigrants and native-born people. We interpreted the probability of remaining sick as a negative transition because the person did not recover, although from another point of view, this may be considered as a positive transition because the person did not die. Smaller and less-consistent differences between the native-born people and immigrants were found for ADL. (Table 2 about here) Transitions in health: estimates from logit specifications Table 3 shows the estimated relationship between health transition and immigrant status in three SHARE waves (2004, 2006, and 2010). Specifically, logistic regressions were run to estimate the transition probabilities from one state (healthy or sick) to another (sick or healthy) for each health measure (SRH, DEP, and ADL). We were thus able to predict the probabilities of transition from

15

good to bad health and vice versa in two separate regressions, firstly distinguishing between natives and all immigrants and then by immigrants’ countries of origin (East-EU, Other-EU, or Extra-EU). In all specifications, in addition to the immigration-related variables, we checked for a broad range

• f controls potentially related to health transitions (see previous section). The first part of Table 3

shows odds ratios interpreted as the effect of being an immigrant on the probability of transitioning from healthy to sick states versus the native-born people (reference group) and revealed a deterioration in SRH for all immigrant groups except for Extra-European immigrants. We did not find significant differences between natives and immigrants for DEP, whereas Extra-EU immigrants were more likely to experience negative changes in ADL. (Table 3 about here) East-EU immigrants were about 57% more likely to experience a negative change in SRH (from healthy to sick) than the native-born people. A health deterioration in terms of ADL also appeared in Extra-EU immigrants, with a statistically significant odds ratio of 1.725. The probability of transition in recovering from the sick state for SRH was the mirror image of those from healthy to sick states, since all immigrant groups were less likely to recover from sickness and move toward a healthy state. For the DEP measure, we found a lower probability of recovery for immigrants from Other-EU; for ADL, the estimated effects were not statistically significant. This was probably due to the small number of transitions available for people with ADL limitations. Each of the control variables of the model, such as gender, age, marital status, education, and the perceived economic conditions revealed the expected signs of ORs and was statistically significant4. Older people tended to undergo a negative transition toward poor self-rated health. In addition, for all health indicators, respondents with little education and low perceived economic resources experienced a

4The estimates of the model with explicit controls are presented in the Appendix, Table A.2.

16

deterioration in health. Being female was also very strongly associated with the probability of transition from healthy to sick states for DEP; with respect to marital status, being married had a significant protective effect against depression. For all health indicators, a strong association between receiving practical social support from someone outside the family and the probability of negative health changes was evidenced. It is plausible that elderly people receiving meaningful support are those with health problems and may need extra help. Table 4 shows the results of our transition models where we included interaction variables between country of origin and age at migration. The probability of moving to poor self-rated health was higher among East-EU immigrants who had migrated during their young and mature adulthood (ORs=2.134 and 1.561, respectively). The same pattern, although to a lesser extent, appeared for young adult immigrants who arrived from Other-EU and Extra-EU when they were between the ages of 15 and 34. In regard to DEP, East-EU immigrants in the age class of 15 to 34 had a significantly higher probability of undergoing a deterioration in health than the reference group (ORs=2.485). A similar deterioration appeared in immigrants aged 35+ from Other-EU. Lastly, our results indicate higher odds of experiencing the onset of ADL limitations for Extra-EU immigrants who arrived in Europe during childhood and mature adulthood. Conversely, looking at transitions from sick to healthy states, East-EU immigrants arriving in the host countries when they were over 15 were less likely to recover from bad health in the subsequent waves according SRH, as well as immigrants from Other-EU arriving either during childhood or in young adulthood who displayed a lower probability of recovering. In the domain of DEP, immigrants from Other-EU who arrived during childhood and when they were 35 and older showed reduced and significant rates of

• recovery. Also, Eastern European immigrants who arrived during childhood were less likely to

move from healthy to sick states. In regard to ADL, no significant statistical differences between native-born and foreign-born people in evaluating their health condition were found. (Table 4 about here)

17

Heterogeneity effects We conducted a sensitivity analysis to test the robustness of our estimates by repeating them only for transitions from healthy to sick states for two groups according to gender (men/women) and age class (younger respondents aged 50–64/65 and older)5. Age was divided into two categories to split younger respondents from those considered “elderly” or older persons in most developed countries, who are usually retired. It is worth noting that the transitions for these subgroups are consistently

• large. In fact, the number of transitions for the three health outcomes range from the minimum of

27,331 for DEP to the maximum of 28,167 for SHR, and the share of transitions for the binary modalities of these variables is between 46% and 54 %. The results shown in the Appendix (Figure A.1 and Figure A.2) do not differ from our conclusions, although some important findings emerge. The horizontal lines in Figures A.1 and A.2 represent the baseline of the odds ratio, equal to 1, which implies there is no difference between immigrants and natives in the probabilities of worsening health. Deterioration in self-rated health was characterized in particular by women from Eastern Europe who arrived when they were 15–34, and their odds ratio was almost three times higher than that of the native-born (ORs=2.641); in women, the probability of recording a deterioration in SRH was significant also for those from Other-EU aged 15–34 on arrival. In regard to DEP, our estimates show a steeper deterioration in health for Other European men who arrived during mature adulthood (ORs=2.701) and for East-EU women who arrived aged 15–34 (ORs=2.642). Similarly, in ADL, Extra-EU women migrating during youth and mature adulthood were found to be more probably affected by a decline in health (Figure A.1). In regard to age, the status of immigrant seems to be more important for older respondents (>=65 years), since a decrease in SRH is evident for such immigrants from East-EU and Other-EU who

5Results for transitions from sick to healthy states are available upon request from the

corresponding author.

18

arrived when they were 15–34 years old. Older East-EU immigrants who arrived when they were adults also had a higher probability of undergoing a decline in DEP, as shown by the higher odds ratios compared to native-born people. Although to a lesser extent, the same holds for Other-EU immigrants who arrived during mature adulthood. Less consistent results were found for ADL (Figure A.2). Robustness tests A robustness check was conducted in order to examine how the baseline results changed if we excluded from the analysis the three Mediterranean countries (Spain, Italy, and Greece), in which the number of elderly immigrants was low compared to the “old” immigration countries in Europe such as Germany, France, Austria, etc. The estimated probabilities of transition among health states for all three indicators (SRH, DEP, and ADL) without the inclusion of the three countries of Southern Europe are presented in the Appendix, Table A.3, and are very similar to the ones listed in Table 4. Our second sensitivity analysis relates to the possibility that the health threshold at the value of

• ne or more limitations in ADL could be bad, since our choice was rather arbitrary, whereas for

depressive symptoms, the cut-off used was validated in the literature as discussed above. We examined whether the results would change if we set a higher threshold of three or more limitations in ADL, which was also used in the literature (Adena and Myck 2014). Our results, reported in the Appendix, Table A.4, do not differ substantially from the baseline specification (ADL 1+) with the exception of a lack of statistical significance in the analysis of health deterioration using the three ADL limitations threshold. This is due to the small sample size when considering three or more limitations in ADL, since only a relatively small fraction of our sample experienced an onset of ADL limitations between the baseline and follow-up. We also present in the Appendix (Table A.5) estimations of transition probabilities between wave 1 and wave 2 to examine how baseline results change when considering a short period of

19

• bservation (i.e., from 2004 to 2006). This sensitivity analysis confirms the goodness of the main

findings, showing that the odds ratios estimates of the restricted sample are close to the baseline models, irrespective of whether health transitions started from healthy to sick or vice versa. Another potential problem with the empirical approach involves attrition in the original sample, which may bias transition probabilities. When we hypothesize that people with poor health have a higher probability of dropping out during follow-up, the average health outcomes of survivors will be higher than that of the population as a whole. Applied to our analysis, this means that if immigrant groups report poorer health than the native-born, the dropouts in this subgroup would lead to an under-estimation of the decline in health status. Table 5 lists the number of individuals for each year and the corresponding number of dropouts between years expressed as survival and attrition rates. As expected, attrition rates by nativity status were higher for immigrants than for the native-born. In particular, immigrants from East-EU have an attrition rate close to 40%, while the native-born have rates of the 25%. (Table 5 about here) Despite the relatively severe attrition, the main questions are whether attrition significantly influences the transition probabilities of health status indicators and whether the different strengths

• f immigrants dropping out of later surveys compared to the native-born definitely generates bias in
• estimations. Although various approaches have been proposed to test attrition6, we follow that of

Verbeek and Nijman (1992) to compare results based on balanced and unbalanced panel data, modified to account for individual health status data which are missing because of non-response or

6 For example, Contoyannis et al. (2004) proposed adding extra variables, such as the number of

times an individual was observed in the sample and whether that individual responded each year and in the subsequent year, to the baseline specification.

20

• death. Sample attrition is in fact driven by these two factors. The last column of Table 5 shows the

share of attrition due to death. The statistics suggest no large difference between native-born people and immigrants in leaving the SHARE survey because of death. Our strategy accounts for attrition driven by death in wave 4 by correcting health outcomes for attrition in wave 2 as “Sick,” under the assumption that death is a reflection of bad health status. Like Adena and Myck (2014), we then tested for the sensitivity of the results to sample attrition by treating those respondents absent from waves 2 and 4 as in poor health in the first estimation set (i.e., Attrition 1) and as healthy in the other set (Attrition 2). Table 6 lists the odds ratios for the estimated transition probabilities of immigrant subgroups on health status, comparing the baseline estimates with those obtained following Adena and Myck’s (2014) approach. We estimated two separate regressions. In the first regression, we included a dummy variable for all immigrants and natives (reference category), while in the second one we distinguished among East-EU, Other-EU, and Extra-EU with natives being the reference. Even if we add observations with health outcomes measured with errors, which may be upward- or downward-biased, the estimated odds ratios for all immigrants from different areas of Europe or Extra-Europe are robust and close to the baseline estimation. Based on these results, we may conclude that panel attrition is unlikely to affect our results. (Table 6 about here) Conclusions Although a considerable amount of research on health differentials among elderly people in Europe exists, less attention has been paid to the health status of the increasing number of older immigrants who are ageing in their European host countries. Torres (2006) asserted that ageing in a country as

21

an immigrant and migrating in old age are two different things. Although the problems and health disadvantages of older people are becoming more compelling, they are not widely recognized by national governments. As evidenced by Ruspini (2009), migrants are mostly perceived as a “hidden population” that may represent a social burden on national welfare and consequently a “problem” or a social risk because of their “otherness.” After having established that some groups of immigrants rate poorer health status than native- born people, this work aims to highlight the differences in the health trajectories between the native- born and foreigners living in Europe in terms of self-rated health, depression, and ADLs from 2004 to 2011. In particular, the roles played by immigration-related variables such as nativity, age at migration, and country of origin were examined. Clearly, the results of this study show that transition probabilities vary according to immigrant status. In all domains of health, older immigrants’ health status deteriorated more frequently than those of the native-born over the 6-year span of the survey. Native-born people and immigrants undergo different types of health changes

• ver time, the former showing better health and less sickness. The probability of moving from

healthy to sick states in SRH was significantly higher for all groups of immigrants. In particular,

• ur results emphasize the heterogeneity among immigrants, as people born in Eastern Europe are

more likely to suffer worsening health, having the highest probabilities of transiting from a healthy to a sick status in SRH and DEP compared to the native-born and other foreign-born groups. Moreover, Eastern European immigrants are also less likely to recover from sickness. Specifically,

• ur estimations showed that East-EU immigrants are about 57% more likely to undergo negative

changes in SRH (from healthy to sick) than native-born people, whereas for Other European immigrants, the risk of transition to the sick state was lower, about 36%. Compared with the native- born, Extra-EU immigrants are 72% more likely to suffer a deterioration in health, as measured by

• ADL. This pattern of immigrants’ health disadvantage was also found when we analyzed health

improvements, evidencing the higher probability of the native-born of staying healthy and recovering from bad health in almost all domains.

22

We also examined whether the increased risk of health deterioration among immigrants is related to their age at the time of migration. Our results showed the existence of a relationship between the immigrants’ age on arrival and deterioration in health status, as foreign-born people who arrived in childhood were “protected” from negative transitions toward bad health, with the exception of Extra-EU immigrants for ADL. Conversely, some immigrants who arrived during adulthood experienced a relatively fast decline in health, being disadvantaged in later life. This was the case of East-EU immigrants who arrived during young adulthood, who showed deterioration in both SRH and DEP, with a probability double that of the native-born. Immigrants from Other-EU and Extra- EU arriving when they were aged 15–34 deteriorated in SRH, like DEP for other European immigrants aged 35+ on their arrival in the host countries. A negative transition in health (from healthy to sick) was also recorded in ADL for Extra-EU immigrants who arrived during childhood

• r mature adulthood. Difficulties in recovering a good health status, with very few exceptions, were

mainly characteristic of immigrants who arrived during adulthood from East and Other Europe. We believe that the health advantage of immigrants arriving in their host countries during childhood lies in the opportunity for better cultural, linguistic, and social integration that they acquire, for example, thanks to the possibility of schooling during childhood in the host country. The formal, compulsory education offered by the host country has definitely contributed to ensuring that barriers to integration, first of all linguistic, are removed. This explanation is in line with the views of Sandford and Seeborg (2003), who found that immigrants who arrived in the destination country as children gained greater returns on human capital investments than immigrants who arrived as young adults. For example, language proficiency and schooling received in the host country are important elements affecting immigrants’ standards of living and may consequently positively influence health trajectories over time. We suggest that many of the health status differences between young and mature immigrants and the native-born can be explained by the fact that most immigrants reached Northern and Western Europe as labor immigrants, the most deprived and socially excluded group (Warnes et al. 2004).

23

Since these immigrants mainly came from less-developed areas, their health disadvantages may depend both on the conditions in their countries of origin (economic and social deprivation) and in the host countries (worsening of dietary styles, adoption of at-risk behavior, less access to healthcare, discrimination, and hard or dangerous working conditions). Eastern European people and other immigrants who arrived as children probably benefited from an improved environmental state of health, since they did not have to experience the poor living conditions, deprivation, ethno- political conflicts, or civil wars that so greatly characterized the life histories of other immigrants who arrived later in life. The greater risk of faster health decline of those who migrated when young

• r mature may also be due to the fact that this group was mainly made up of migrant workers. On
• ne hand, work could facilitate acceptance in the new society by engaging immigrants in the local

culture and structures through the provision of health insurance or improvements in language or socio-economic position. However, labor migrants may also be exposed to many negative factors, such as exploitation, discrimination, dangerous or hard working conditions, lower salaries, limited legal rights, and limited access to healthcare services. The existing literature shows that most labor migrants arriving in their host countries after the Second World War were poorly educated and entered low-skilled and low-paid manual work. As suggested by Warnes et al. (2004, p. 312): "In short, in comparison to the host populations, they have had a lifetime of disadvantage and deprivation, including poor health care and housing conditions, few opportunities to learn the local language, and very often the insults of cultural and racial discrimination.” This analysis showed that elderly migrants are comprised of very heterogeneous groups with different kinds of life-courses and immigration trajectories and that they have lived through various states according to both their countries of origin and to that of their destinations. The social and health policies of their receiving countries must recognize these differences if they are to address the specific needs of elderly migrants. We have shown that some groups, such as Eastern European immigrants, suffer from disadvantages due to their higher probability of health deterioration. Policy measures should take into account their vulnerability and create different policies targeting specific

24

immigrant groups to guarantee better integration and that their needs are met during their ageing process in their host countries. Acknowledgement "This paper uses data from SHARE Waves 1, 2 and 4 (DOIs: 10.6103/SHARE.w1.500, 10.6103/SHARE.w2.500, 10.6103/SHARE.w4.500), see Börsch-Supan et al. (2013) for methodological details. The SHARE data collection has been primarily funded by the European Commission through FP5 (QLK6-CT-2001-00360), FP6 (SHARE-I3: RII-CT-2006-062193, COMPARE: CIT5-CT-2005-028857, SHARELIFE: CIT4-CT-2006-028812) and FP7 (SHARE- PREP: N°211909, SHARE-LEAP: N°227822, SHARE M4: N°261982). Additional funding from the German Ministry of Education and Research, the U.S. National Institute on Aging (U01_AG09740-13S2, P01_AG005842, P01_AG08291, P30_AG12815, R21_AG025169, Y1-AG- 4553-01, IAG_BSR06-11, OGHA_04-064) and from various national funding sources is gratefully acknowledged (see www.share-project.org)." References AAMEE (Active Ageing of Migrant Elders across Europe). 2008. Bonn Memorandum: Active Ageing of Migrant Elders across Europe, adopted on 02 October 2008 by the first European Conference “Active Ageing of Migrant Elders – from Challenges to Opportunities” at the World Conference Center in Bonn (WCCB). Bonn: Ministry for Intergenerational Affairs, Family, Women and Integration of the State of North Rhine-Westphalia. Abraido-Lanza, A.F., M.T. Chao, and K.R. Florez. 2005. Do Healthy Behaviors Decline with Greater Acculturation? Implications for the Latino Mortality Paradox, Social Science and Medicine 61: 1243-1255.

25

Adena, M. and M. Myck. 2014. Poverty and transitions in health in later life, Social Science and Medicine 116: 202–210. Agudelo-Suárez, A., D. Gil-González, E. Ronda-Pérez, V. Porthé, G. Paramio-Pérez, A.M. García, and A. Garì. 2009. Discrimination, work and health in immigrant populations in Spain, Social Science and Medicine 68(10): 1866–1874. Akhtar, S. 1999. The immigrant, the exile, and the experience of nostalgia, Journal of Applied Psychoanalytic Studies 1: 123–130. Alegría, M., G. Canino, P.E. Shrout, M. Woo, N. Duan, D. Vila, and X-L. Meng. 2008. Prevalence

• f mental illness in immigrant and non-immigrant U.S. Latino groups, American Journal of

Psychiatry 165: 359–369. Alegría, M., P.E. Shrout, M. Woo, P. Guarnaccia, W. Sribney, D. Vila, and G. Canino. 2007. Understanding differences in past year psychiatric disorders for Latinos living in the U.S, Social Science and Medicine 65: 214–230. Angel, J.L., C.J. Buckley, and A. Sakamoto. 2001. Duration or disadvantage? Exploring nativity, ethnicity, and health in midlife, Journal of Gerontology: Social Sciences 56B (5): 275-284. Antecol, H. and K. Bedard. 2006. Unhealthy Assimilation: Why Do Immigrants Converge to American Health Status Levels? Demography 43(2): 337-360. Bath, P.A. 1999. Self-rated health as a risk factor for prescribed drug use and future health and social service use in older people, J. Gerontol. Ser. A-Biol. Sci. Med. 54: 565–570. Bonifazi, C., F. Heins, S. Strozza, and M. Vitiello. 2009. The Italian transition from emigration to immigration country, IRPPS Working Paper No. 24. Börsch-Supan, A. and H. Jürges. 2005. The Survey of Health, Aging, and Retirement in Europe – Methodology, Mannheim Research Institute for the Economics of Aging. Börsch-Supan, A., M. Brandt, C. Hunkler, T. Kneip, J. Korbmacher, F. Malter, B. Schaan, S. Stuck, and S. Zuber. 2013. Data Resource Profile: The Survey of Health, Ageing and Retirement in Europe (SHARE), International Journal of Epidemiology, doi: 10.1093/ije/dyt088.

26

Carrasquillo, O., A.I. Carrasquillo, and S. Shea. 2000. Health insurance coverage of immigrants living in the United States: differences by citizenship status and country of origin, American Journal of Public Health 90(6): 917–23. Ceballos, M. and A. Palloni. 2010. Maternal and infant health of Mexican immigrants in the United States: the effects of acculturation, duration, and selective return migration, Ethn Health. 15(4): 377–396. Contoyannis, P., A.M. Jones, and N. Rice. 2004. The dynamics of health in the British Household Panel Survey, Journal of Applied Econometrics 19(4): 473–503. Cots, F., X. Castells, O. García, M. Riu, A. Felipe, and O. Vall. 2007. Impact of immigration on the cost of emergency visits in Barcelona (Spain), BMC Health Services Research 7:9 http://www.biomedcentral.com/1472-6963/7/9. Davies, A.A., A. Basten, and C. Frattini. 2006. Migration: A social determinant of migrants’ health, Eurohealth 16 (1): 10-12. de Bruin, A., H.S.J. Picavet, and A. Nossikov. 1996. Health interview surveys. Towards international harmonization of methods and instruments. WHO, Geneva, Regional Publications European Series no 58. Diehr, P. and D.L. Patrick. 2001. Probabilities of transition among health states for older adults, Quality of Life Research 10: 431–442. Dowd, J. and V. Bengtson. 1978. Aging in minority populations: an examination of the double jeopardy hypothesis, Journal of Gerontology33: 427–436. Fassmann, H. and R. Münz. 1992. Patterns and trends of international migration in Western Europe, Population and Development Review 18(3): 457–480. Geronimus, A.T. 2001. Understanding and eliminating racial inequalities in women's health in the United States: the role of the weathering conceptual framework, Journal of the American Medical Women’s Association 56(4):133–136, 149-50.

27

Gravelle, H., M. Sutton, S. Morris, F. Windmeijer, A. Leyland, C. Dibben, and M. Muirhead. 2003. Modelling supply and demand influences on the use of health care: implications for deriving a needs-based capitation formula, Health Economics 12 (12): 985–1004. Gubernskaya, Z. 2014. Age at migration and self-rated health trajectories after age 50: understanding the older immigrant health paradox, Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, doi: 10.1093/geronb/gbu049. Hank, K., C. Deindl, and M. Brandt. 2013. Changes in older Europeans’ health across two waves of SHARE: life-course and societal determinants, Population Ageing 6: 85–97. Idler, E.L. and Y. Benyamini. 1997. Self-rated health and mortality: A review of twenty-seven community studies, J. Health Soc. Behav 38: 21–37. Klevmarken, N.A., B. Swensson, and P. Hesselius. 2005. The share sampling procedures and calibrated design weights. In A. Börsch-Supan and H. Jürges (Eds.), The survey of health, aging, and retirement in Europe. Mannheim: Mannheim Research Institute for the Economics of Aging.

• pp. 28–69.

Lanari, D. and O. Bussini. 2012. International migration and health inequalities in later life, Ageing and Society 32(6): 935–962. Lanari, D., O. Bussini, and L. Minelli. 2015. Self-rated health among Eastern European immigrants

• ver 50 living in Western Europe, International Journal of Public Health 60(1): 21–31.

Leão, T.S., J. Sundquist, S–E. Johansson, and K. Sundquist. 2009. The influence of age at migration and length of residence on self-rated health among Swedish immigrants: a cross-sectional study, Ethnicity and Health 14(1): 93–105. Leveille, S.G., H.E. Resnick, and J. Balfour. 2000. Gender differences in disability: evidence and underlying reasons, Aging Clin. Exp. Res. 12(2): 106–112. Morris, S., M. Sutton, and H. Gravelle. 2005. Inequity and inequality in the use of health care in England: an empirical investigation, Social Science & Medicine 60 (6): 1251-1266.

28

Portes, A. and R.G. Rumbaut. 2006. Immigrant America: a portrait. Berkeley: University of California Press 3rd ed. Prince, M.J., F. Reischies, A.T.F. Beekman, R. Fuhrer, C. Jonker, S.–L. Kivelä, B. Lawlor, A. Lobo,H. Magnússon, I. Fichter, H. van Oyen, M. Roelands, I. Skoog, C. Turrina, and. J.R.M. Copelan.(1999a). Development of the EURO–D scale – a European union initiative to compare symptoms of depression in 14 European Centres, British Journal of Psychiatry 174: 330–338. ———. (1999b). Depression symptoms in late life assessed using the EURO-D scale, British Journal of Psychiatry174: 339–345. Pudaric, S., J. Sundquist, and S–E. Johansson 2003. Country of birth, instrumental activities of daily living, self-rated health and mortality: a Swedish population-based survey of people aged 55-7, Social Science and Medicine 56: 2493–2503. Ringbäck, G., A. Gullberg, A. Hjern, and M. Rosén. 1999. Mortality statistics in immigrant research: method for adjusting underestimation of mortality, International Journal of Epidemiology 28: 756–763. Robine, J-M., C. Jagger and Euro-REVES group. 2003. Creating a coherent set of indicators to monitor health across Europe. The Euro-REVES 2 project, European Journal of Public Health 13(3 supplement): 6-14. Ronellenfitsch, U. and O. Razum. 2004. Deteriorating health satisfaction among immigrants from Eastern Europe to Germany, International Journal for Equity in Health 3: 4. Ruspini, P. 2009. Elderly migrants in Europe: an overview of trends, policies and practices. Article drawn from a preliminary report prepared in 2009 for the European Committee on Migration of the Council of Europe (CDMG). Sandford, J. and M.C.Seeborg. 2003. The assimilation of immigrants who arrived in the United States as children, Journal of Business and Economics Research 1(3): 27–36.

29

Silveira, E.R. and S. Ebrahim. 1998. Social determinants of psychiatric morbidity and well-being in immigrant elders and whites in East London, International Journal of Geriatric Psychiatry 13: 801–812. Silveira, E.R., I. Skoog, V. Sundh, P. Allebeck, and B. Steen. 2002. Health and well-being among 70-year-old migrants living in Sweden – Results from the H 70 gerontological and geriatric population studies in Göteborg, Social Psychiatry and Psychiatric Epidemiology 37: 13–22. Solé-Auró, A. and E.M. Crimmins. 2008. Health of Immigrants in European Countries, International Migration Review 42(4):861–876. Tielke, S. and P. Diehr. 2012. Transitions among health states using 12 measures of successful aging in men and women: results from the Cardiovascular Health Study, Journal of Aging Research doi:10.1155/2012/243263. Torres, S. 2006. Elderly Immigrants in Sweden: ‘Otherness’ Under Construction, Journal of Ethnic and Migration Studies 32(8): 1341–1358. UNESCO (Institute for Statistics, International Standard Classification

• f

Education). 1997.<http://www.uis.unesco.org/TEMPLATE/pdf/isced/ISCED_A.pdf>. United Nations. 2008. Population Division: The Standard Country or Area Codes for Statistical Use, Revision 4. Vaillant, N. and F.–C Wolff. 2010. Origin differences in self-reported health among older migrants living in France, Public Health 124(2): 90–8. Verroupoulou, G. 2012. Determinants of change in self-rated health among older adults in Europe: a longitudinal perspective based on SHARE data, European Journal of Aging 9: 305-318. Verbeek, M. and T. Nijman. 1992. Testing for Selectivity Bias in Panel Data Models, International Economic Review, 33(3): 681-703. Viruell-Fuentes, E. 2007. Beyond Acculturation: Immigration, Discrimination, and Health Research among Mexicans in the United States, Social Science and Medicine 65: 1524-1535.

30

Warnes, A.M., K. Friedrich, L. Kellaher, S. Torres. 2004. The diversity and welfare of older migrants in Europe, Ageing and Society 24(3): 307–26. Warnes, A.M., and A. Williams. 2006. Older Migrants in Europe: A New Focus for Migration Studies, Journal of Ethnic and Migration Studies 32(8): 1257–1281. White, P. 2006. Migrant Populations Approaching Old Age: Prospects in Europe, Journal of Ethnic and Migration Studies 32(8): 1283–1300.

31

Tables

Table 1 Descriptive statistics: sample size and prevalence of "Healthy" state by nativity status

Notes: SRH= self-rated health (Healthy: excellent, very good, good health status); DEP= depression (Healthy: less than three symptoms); ADL= activities of daily living (Healthy: no limitations). A restricted panel sample of 19,935 individuals with all relevant information at waves 1 and subsequent waves (W2 and W4), was used for analysis (column 2). Last three columns show prevalence of healthy state among native-born and immigrants, distinguished by country of

• rigin.

Source: Authors’ calculations according to SHARE data, waves 1-2-4, individuals aged 50 and over. Nativity status Total sample Sample used Died by W1 for analysis W4|W1 SRH DEP ADL Native-born 24,867 18,539 1,404 68.2 76.2 89.8 All immigrants 2,194 1,396 86 62.2 71.9 88.8 East-EU 381 230 14 54.2 69.2 99.3 Other-EU 1,197 788 45 62.3 73.0 87.4 Extra-EU 616 377 27 66.7 71.3 91.5 Sample size 27,061 19,935 1,490 Healthy Prevalence

32

Table 2 Estimated transition probabilities by nativity status

Notes: SRH= self-rated health (Sick: fair or poor health status); DEP= depression (Sick: at least three symptoms); ADL= activities of daily living (Sick: at least one limitation). Estimated transition probabilities were performed by

• rdered logit for native-born and immigrants and separately by country of origin. Healthy and sick states were

augmented with death as the additional (absorbing) state. Transition probabilities were obtained including the following control variables: gender, age, age squared, marital status, education, perceived economic resources, social support received from outside family, lifestyle (drinking, smoking), time between interviews, and country of residence considered as a fixed effect. Confidence intervals are shown in bracket. Source: see Table 1.

State at t0 Healthy Sick Dead Healthy Sick Dead Healthy Sick Dead Native-born Healthy 79.44 18.11 2.44 83.90 13.33 2.76 90.87 6.47 2.65

[79.2 ; 79.6] [17.9 ; 18.3] [2.4 ; 2.5] [83.7 ; 84.1] [13.2 ; 13.5] [2.7 ; 2.8] [90.6 ; 91.2] [6.3 ; 6.7] [2.6 ; 2.7]

Sick 28.91 65.38 5.69 44.76 50.08 5.11 42.82 46.94 10.24

[28.5 ; 29.3] [65.1 ; 65.6] [5.5 ; 5.8] [44.2 ; 45.3] [49.6 ; 50.5] [4.9 ; 5.3] [41.2 ; 44.3] [46.6 ; 47.4] [9.6;10.9]

All immigrants Healthy 74.60 23.75 2.05 83.41 14.34 2.25 89.93 7.82 2.25

[73.4 ; 75.7] [22.8 ; 24.6] [1.8 ; 2.2] [82.6 ; 84.1] [13.7 ; 14.9] [2.1 ; 2.3] [89.6 ; 90.3] [7.5 ; 8.1] [2.1 ; 2.3]

Sick 20.33 75.25 4.40 39.20 56.60 4.19 43.10 49.43 7.47

[19.2 ; 21.4] [74.4 ; 76.1] [4.1 ; 4.7] [36.9 ; 41.5] [54.6 ; 58.5] [3.4 ; 4.9] [42.9 ; 43.4] [48.8; 50.1] [7.1 ; 7.9]

East-EU Healthy 70.69 28.24 1.05 81.2 17.9 0.79 90.69 7.59 1.72

[67.4 ; 73.9] [25.4 ; 31.1] [0.4 ; 1.6] [77.7 ; 84.8] [14.7 ; 21.1] [0.1 ; 1.4] [89.7 ; 91.6] [7.2 ; 7.9] [1.3 ; 2.1]

Sick 20.99 75.5 3.45 38.4 57.8 3.7 50.00 46.15 3.85

[17.2 ; 24.7] [72.2 ; 78.8] [2.4 ; 4.4] [32.6 ; 44.1] [52.4 ; 63.1] [0.3 ; 7.2] [48.1 ; 51.9] [45.4 ; 47.0] [3.1 ; 4.5]

Other-EU Healthy 74.64 23.45 1.90 83.12 14.63 2.23 89.96 8.03 2.01

[73.2 ; 76.1] [22.1 ; 24.7] [1.7 ; 2.0] [81.9 ; 84.2] [13.7 ; 15.5] [2.0 ; 2.4] [89.7 ; 90.3] [7.8; 8.2] [1.6 ; 2.4]

Sick 19.47 75.65 4.88 36.63 59.44 3.91 42.28 50.41 7.32

[17.3 ; 21.6] [73.8 ; 77.4] [4.3 ; 5.4] [33.1 ; 40.2] [56.4 ; 62.4] [3.2 ; 4.5] [39.8 ; 44.7] [48.4 ; 52.4] [6.8 ;7.8]

Extra-EU Healthy 76.26 20.87 2.86 85.73 11.08 3.18 89.39 7.55 3.06

[73.6 ; 78.9] [18.7 ; 22.9] [2.2 ; 3.5] [83.6 ; 87.8] [9.52 ; 12.6] [2.5 ; 3.8] [89.1 ; 89.7] [7.4 ; 7.7] [2.8 ; 3.4]

Sick 21.7 73.57 4.70 45.95 48.41 5.64 40.00 48.00 12.00

[18.5 ; 24.8] [70.3 ; 76.8] [3.5 ; 5.9] [40.6 ; 51.3] [41.8 ; 54.9] [2.10 ; 9.14] [38.9 ; 41.2] [46.5 ; 49.5] [10.8 ; 13.2]

State at time t ADL (1+) DEP (3+) SRH

33

Table 3 Probability models of health transitions according to nativity status

Notes: Abbreviations as above. Following variables controlled for: gender, age, age squared, marital status, education, perceived economic resources, social support received from outside family, lifestyle (drinking, smoking), time between interviews, and country

• f residence considered as a fixed effect.

Significance levels: *** values below p < 0.01; **values below p < 0.05; *values below p < 0.1. Standard errors, clustered at individual level, in brackets. Source: see Table 1.

Nativity status OR s.e. OR s.e. OR s.e. Native-born (ref.) All immigrants 1.368 (0.132)*** 1.047 (0.118) 1.254 (0.162)* Native-born (ref.) East-EU 1.572 (0.360)** 1.308 (0.326) 1.217 (0.335) Other-EU 1.363 (0.174)*** 1.080 (0.154) 1.031 (0.183) Extra-EU 1.226 (0.221) 0.771 (0.189) 1.725 (0.395)** Sample size OR s.e. OR s.e. OR s.e. Native-born (ref.) All immigrants 0.654 (0.087)*** 0.772 (0.101)** 1.032 (0.235) Native-born (ref.) East-EU 0.642 (0.174)* 0.686 (0.185) 1.222 (0.489) Other-EU 0.598 (0.108)*** 0.693 (0.118)** 0.912 (0.256) Extra-EU 0.793 (0.201) 1.018 (0.269) 1.553 (0.981) Sample size 1,577 4,251 4,885 SRH DEP (3+) ADL (1+) From healthy to sick state From sick to healthy state 12,682 13,569 16,651

34

Table 4 Probability models of health transitions according to nativity status and age at migration

Notes: Abbreviations as above. Following variables controlled for: gender, age, age squared, marital status, education, perceived economic resources, social support received from outside family, lifestyle (drinking, smoking), time between interviews, and country

• f residence considered as a fixed effect.

Significance levels: *** values below p < 0.01; **values below p < 0.05; *values below p < 0.1. Standard errors, clustered at individual level, in brackets. Source: see Table 1.

Age at migration * nativity status OR s.e. OR s.e. OR s.e. Native-born (ref.) Less than 15 years * East-EU 1.253 (0.476) 0.654 (0.330) 1.488 (0.593) 15-34 years * East-EU 2.134 (0.769)** 2.485 (0.908)** 1.181 (0.546) >= 35 years * East-EU 1.561 (0.669)* 1.116 (0.505) 0.797 (0.485) Sample size Less than 15 years * Other-EU 1.130 (0.251) 0.921 (0.226) 0.765 (0.238) 15-34 years *Other-EU 1.491 (0.258)** 0.891 (0.187) 1.101 (0.274) > = 35 years * Other-EU 1.590 (0.505) 2.214 (0.632)** 1.407 (0.532) Sample size Less than 15 years * Extra-EU 1.186 (0.429) 0.704 (0.317) 2.399 (0.916)** 15-34 years * Extra-EU 1.249 (0.296)* 0.874 (0.247) 1.149 (0.405) >= 35 years * Extra-EU 1.288 (0.530) 0.571 (0.406) 2.618 (0.197)** Sample size OR s.e. OR s.e. OR s.e. Native-born (ref.) Less than 15 years * East-EU 1.046 (0.374) 0.492 (0.206)* 1.955 (1.095) 15-34 years * East-EU 0.399 (0.218)** 0.559 (0.249) 0.981 (0.580) >= 35 years * East-EU 0.406 (0.226)* 1.787 (0.998) 0.541 (0.625) Sample size Less than 15 years * Other-EU 0.589 (0.188)* 0.694 (0.227)* 0.726 (0.431) 15-34 years *Other-EU 0.597 (0.146)** 0.763 (0.165) 1.048 (0.364) > = 35 years * Other-EU 0.605 (0.277) 0.473 (0.211)* 0.481 (0.375) Sample size Less than 15 years * Extra-EU 0.502 (0.237) 0.734 (0.337) 0.325 (0.426) 15-34 years * Extra-EU 1.113 (0.366) 1.309 (0.533) 3.590 (4.236) >= 35 years * Extra-EU 0.506 (0.411) 0.902 (0.435)

• Sample size

SRH DEP (3+) ADL (1+) From healthy to sick state From sick to healthy state

12,564 12,801 12,935 13,316 13,179 12,936 12,635 12,873 13,011 3,864 4,009 4,100 3,838 3,985 4,071 3,939 4,093 4,179

35

Table 5 Sample size, drop-outs and attrition rates by nativity status

Notes: The number of longitudinal observations used in estimation starting from wave 1 (W1), which are not missing in wave 2 (W2) and /or wave 4 (W4), is 19,935. Source: see Table 1.

• N. individuals (W1)

Survival rate Drop-outs Attrition rate Attrition due to death % % %

Native-born

24,867 74.5 6,328 25.4 5.6

All immigrants

2,194 63.6 798 36.4 3.9

East-EU

381 60.4 151 39.6 3.6

Other-EU

1,197 65.8 409 34.2 3.7

Extra-EU

616 61.2 239 38.8 4.4

Total Sample

27,061 73.6 7,126 26.3 5.5

36

Table 6 Sensitivity to attrition: odds ratios of immigrant status variables in probability models of transitions in health

Notes: Abbreviations as above. Attrition 1 lists baseline estimates when respondents absent from waves 2 and 4 treated as sick; Attrition 2= healthy. In the specifications used for evaluating attrition, we include control variables as in baseline specification. Significance levels: *** values below p < 0.01; **values below p < 0.05; *values below p < 0.1. Standard errors, clustered at individual level, in brackets. Source: see Table 1. Nativity status Baseline Attrition 1 Attrition 2 Baseline Attrition 1 Attrition 2 Baseline Attrition 1 Attrition 2 Native-born (ref.) All immigrants 1.368*** 1.374*** 1.362*** 1.047 1.037 1.035 1.254* 1.274* 1.218* (0.132) (0.127) (0.130) (0.118) (0.117) (0.111) (0.162) (0.162) (0.150) Native-born (ref.) East-EU 1.572** 1.595** 1.528** 1.308 1.296 1.211 1.217 1.239 1.185 (0.360) (0.351) (0.348) (0.326) (0.316) (0.297) (0.335) (0.340) (0.314) Other-EU 1.363*** 1.389*** 1.337** 1.080 1.075 1.081 1.031 1.062 0.996 (0.174) (0.170) (0.170) (0.154) (0.150) (0.148) (0.183) (0.185) (0.171) Extra-EU 1.226 1.274 1.198 0.771 0.871 0.818 1.725** 1.795** 1.632** (0.221) (0.223) (0.203) (0.189) (0.196) (0.180) (0.395) (0.399) (0.360) Sample size

12,682 13,008 13,782 13,569 13,685 14,507 16,651 16,861 17,424

From sick to healthy state Native-born (ref.) All immigrants 0.654*** 0.663*** 0.665*** 0.772** 0.771** 0.783** 0.957 0.949 0.972 (0.087) (0.079) (0.088) (0.101) (0.091) (0.099) (0.217) (0.184) (0.232) Native-born (ref.) East-EU 0.642* 0.597** 0.698* 0.686 0.674 0.781 1.222 1.000 1.267 (0.174) (0.156) (0.172) (0.185) (0.188) (0.194) (0.489) (0.385) (0.498) Other-EU 0.598*** 0.575** 0.608*** 0.693** 0.660*** 0.712*** 0.912 0.918 0.941 (0.108) (0.094) (0.096) (0.118) (0.111) (0.110) (0.256) (0.224) (0.227) Extra-EU 0.793 0.718 0.916 1.018 0.925 1.140 1.553 1.415 1.685 (0.201) (0.211) (0.220) (0.269) (0.208) (0.287) (0.981) (0.619) (1.002) Sample size

4,885 5,992 5,219 4,251 5,168 4,346 1,577 2,350 1,775

SRH DEP (3+) ADL (1+) From healthy to sick state

37

Appendix Table A.1 Description of variables

Variables Modalities Descriptions Dependent variables Self-rated health Less than good (fair, poor) Depression 3+ depressive symptoms 3+ symptoms forming the EURO-D scale: sadness-depression, pessimism, suicidal tendency, guilt, trouble sleeping, lack of interest, irritability, loss of appetite, fatigue, lack of concentration, lack of enjoyment, tearfulness Activities of daily living 1+ ADLs 1+ difficulties performing any of the following ADLs: walking across a room, dressing, bathing or showering, eating, getting in

• r out of bed, and using the toilet

Independent variables Gender Male Female Age (Continuous) Country of origin East-Europe Born in Eastern European countries Other-Europe Born in Southern, Western and Northern European countries Extra-Europe Born in non-European countries Age at migration during childhood <15 years during young adulthood 15-35 years during mature adulthood >35 years Marital status married never married widowed, divorced, separated Smoking Regular smokers ever smoked cigarettes, cigars, cigarillos or a pipe daily for a period of at least one year non-smokers Alcohol consumption >2 glasses almost every day Alcoholic beverages (beer, cider, wine, spirits, cocktails) consumed last six months <2 glasses Education Low from 0 to 2 ISCED levels Medium from 3 to 4 ISCED levels High from 5 to 6 ISCED levels Perceived economic conditions Low household able to make ends meet with great/some difficulty High/intermediate household able to make ends meet easily/fairly easily Social support Received received help (personal care, practical household help, help with paperwork) from outside the household (family member, neighbor, friend) in the last twelve months Not received

38

Table A.2 Probability models of health transitions including controls (from healthy to sick state)

Notes: SRH=self-rated health (Sick: fair or poor health status); DEP= depression (Sick: at least three symptoms); ADL=

activities of daily living (Sick: at least one limitation).

Significance levels: *** values below p < 0.01; **values below p < 0.05; *values below p < 0.1. Standard errors, clustered at individual level, in brackets. Source: see Table 1. SRH DEP (3+) ADL (1+) Variables OR s.e. OR s.e. OR s.e. Native-born (ref.) All immigrants 1.368 0.132*** 1.047 0.118 1.254 0.162* Age (continuous) 1.078 0.041** 0.886 0.035*** 0.891 0.045** Age (squared) 0.999 0.001 1.001 0.001*** 1.001 0.001*** Gender Male (ref.) Female 1.100 0.061* 1.843 0.117*** 1.088 0.092 Marital Status Married (ref.) Never married 1.072 0.097 1.199 0.095** 1.198 0.155 Widowed, divorced, separated 0.994 0.066 1.071 0.087 1.132 0.104 Education High (ref.) Intermediate 1.474 0.110*** 1.225 0.091*** 1.143 0.137 Low 1.772 0.131*** 1.034 0.104 1.711 0.195*** Perceived economic resources High/intermediate (ref.) Low 1.725 0.101*** 1.636 0.101*** 1.714 0.135*** Social support Received 1.674 0.103*** 1.684 0.111*** 2.391 0.188*** not received (ref.) Smoking Regular smokers 1.101 0.059* 1.055 0.063 1.061 0.086 Non-smokers (ref.) Alcohol consumption > 2 glasses almost every day 0.968 0.073 0.999 0.083 1.083 0.117 < 2 glasses (ref.) Country of residence Austria (ref.) Germany 1.143 0.144 1.205 0.164 0.928 0.154 Sweden 1.074 0.123 0.874 0.118 0.571 0.094*** The Netherland 0.891 0.109 0.864 0.119 0.445 0.080*** Spain 1.266 0.167* 1.706 0.238*** 0.740 0.124* Italy 1.060 0.132 1.443 0.195*** 0.502 0.085*** France 1.067 0.127 1.398 0.185** 0.588 0.097*** Denmark 0.602 0.085*** 0.789 0.121 0.514 0.101*** Greece 0.372 0.052*** 0.293 0.052*** 0.135 0.033*** Switzerland 0.463 0.074*** 0.864 0.146 0.544 0.117*** Belgium 0.611 0.071*** 1.071 0.135 0.746 0.111** time between interviews 1.018 0.014 1.101 0.017*** 1.081 0.022*** constant 0.001 0.002*** 0.626 0.868 0.142 0.255 Sample size From healthy to sick state

12,682 13,569 16,651

39

Table A.3 Probability models of health transitions for the restricted sample obtained by excluding the Southern European countries (Greece, Italy and Spain)

Notes: Abbreviations as above. Following variables controlled for: gender, age, age squared, marital status, education, perceived economic resources, social support received from outside family, lifestyle (drinking, smoking), time between interviews, and country

• f residence considered as a fixed effect.

Significance levels: *** values below p < 0.01; **values below p < 0.05; *values below p < 0.1. Standard errors, clustered at individual level, in brackets. Source: see Table 1.

Age at migration * nativity status OR s.e. OR s.e. OR s.e. Native-born (ref.) Less than 15 years * East-EU 0.984 (0.402) 0.537 (0.298) 1.218 (0.496) 15-34 years * East-EU 2.694 (1.069)** 2.772 (1.026)*** 1.192 (0.483) >= 35 years * East-EU 1.381 (0.656) 1.148 (0.519) 0.722 (0.444) Sample size Less than 15 years * Other-EU 1.069 (0.244) 0.934 (0.228) 0.843 (0.243) 15-34 years *Other-EU 1.523 (0.265)*** 0.929 (0.199) 1.157 (0.254) > = 35 years * Other-EU 1.642 (0.517) 2.152 (0.625)*** 1.295 (0.502) Sample size Less than 15 years * Extra-EU 1.157 (0.486) 0.576 (0.302) 2.428 (0.875)** 15-34 years * Extra-EU 1.284 (0.299)* 0.908 (0.294) 1.367 (0.432) >= 35 years * Extra-EU 1.252 (0.784) 0.659 (0.597) 2.137 (1.241)* Sample size OR s.e. OR s.e. OR s.e. Native-born (ref.) Less than 15 years * East-EU 0.941 (0.374) 0.576 (0.259) 2.503 (1.710) 15-34 years * East-EU 0.385 (0.213)* 0.654 (0.329) 1.154 (0.805) >= 35 years * East-EU 0.392 (0.221)* 1.715 (0.988) 0.498 (0.580) Sample size Less than 15 years * Other-EU 0.555 (0.204) 0.693 (0.249) 0.936 (0.567) 15-34 years *Other-EU 0.526 (0.149)** 0.645 (0.139)** 1.151 (0.400) > = 35 years * Other-EU 0.726 (0.331) 0.466 (0.226) 0.441 (0.313) Sample size Less than 15 years * Extra-EU 0.837 (0.401) 1.106 (0.556) 0.750 (1.026) 15-34 years * Extra-EU 1.190 (0.439) 1.498 (0.729) 4.263 (4.631) >= 35 years * Extra-EU 0.321 (0.357) 0.288 (0.239)

• Sample size

SRH DEP (3+) ADL (1+) From healthy to sick state From sick to healthy state

9,545 9,868 9,499 9,713 10,043 9,669 9,762 10,140 9,811 2,379 2,479 2,392 2,475 2,529 2,572 2,465 2,515 2,622

40

Table A.4 Probability models of health transitions for ADL (3+)

Notes: ADL= activities of daily living (Sick: at least three limitations). Following variables controlled for: gender, age, age squared, marital status, education, perceived economic resources, social support received from outside family, lifestyle (drinking, smoking), time between interviews, and country of residence considered as a fixed effect. Significance levels: *** values below p < 0.01; **values below p < 0.05; *values below p < 0.1. Standard errors, clustered at individual level, in brackets. Source: see Table 1.

Nativity status OR s.e. OR s.e. Native-born (ref.) All immigrants 1.444 (0.445) 0.676 (0.180) Native-born (ref.) East-EU 1.911 (1.018) 0.524 (0.289) Other-EU 1.179 (0.508) 0.751 (0.249) Extra-EU 1.645 (0.851)* 0.592 (0.353) Sample size ADL (3+) From sick to healthy state From healthy to sick state

17,924 300

41

Table A5 Sensitivity for wave selection - probability models of health transitions: baseline estimates (waves 2004-2006-2010) versus restricted sample (waves 2004-2006)

Notes: Abbreviations as above. Following variables controlled for: gender, age, age squared, marital status, education, perceived economic resources, social support received from outside family, lifestyle (drinking, smoking), time between interviews, and country of residence considered as a fixed effect. Significance levels: *** values below p < 0.01; **values below p < 0.05; *values below p < 0.1. Standard errors, clustered at individual level, in brackets. Source: see Table 1.

Nativity status Baseline Restricted Baseline Restricted Baseline Restricted Native-born (ref.) All immigrants 1.368*** 1.450*** 1.047 1.043 1.254* 1.299* (0.132) (0.166) (0.118) (0.143) (0.162) (0.205) Native-born (ref.) East-EU 1.572** 1.727** 1.308 1.362 1.217 1.576 (0.360) (0.438) (0.326) (0.403) (0.335) (0.502) Other-EU 1.363*** 1.428*** 1.080 1.089 1.031 0.969 (0.174) (0.217) (0.154) (0.189) (0.183) (0.217) Extra-EU 1.226 1.283 0.771 0.708 1.725** 1.694** (0.221) (0.276) (0.189) (0.218) (0.395) (0.471) Sample size

12,682 8,911 13,569 9,385 16,651 11,639

Native-born (ref.) All immigrants 0.654*** 0.658** 0.772** 0.724** 0.957 1.045 (0.087) (0.113) (0.101) (0.101) (0.217) (0.278) Native-born (ref.) East-EU 0.642* 0.653* 0.686 0.626 1.222 0.871 (0.174) (0.201) (0.185) (0.190) (0.489) (0.458) Other-EU 0.598*** 0.643** 0.693** 0.631** 0.912 1.132 (0.108) (0.160) (0.118) (0.134) (0.256) (0.378) Extra-EU 0.793 0.647 1.018 1.077 1.553 1.462 (0.201) (0.228) (0.269) (0.299) (0.981) (0.906) Sample size

4,885 3,239 4,251 3,109 1,577 1,172

SRH DEP (3+) ADL (1+) From healthy to sick state From sick to healthy state

42

Figure A.1 Immigration and transitions in health: transitions from healthy to sick states by gender -

• dds ratios by immigrants’ country of origin and age at migration.

Note: Bars represent odds ratios by immigrants’ country of origin and age at migration dummy variables; spikes are 95% confidence intervals. Estimates include control variables as in baseline specifications.

43

Figure A.2 Immigration and transitions in health: transitions from healthy to sick states by age -

• dds ratios by immigrants’ country of origin and age at migration.

44

Note: Bars represent odds ratios by immigrants’ country of origin and age at migration dummy variables; spikes are 95% confidence intervals. Estimates include control variables as in baseline specifications.