Social health inequalities: a French analysis based on the migrant population

Transcription

1 Socal health nequaltes: a French analyss based on the mgrant populaton Carolne Berchet * Florence Jusot Paul Dourgnon Abstract: Usng a representatve survey of the French populaton, the Health, Health Care and Insurance Survey (ESPS: Enquête sur la santé et la protecton socale ), ths artcle ams to study the lnks between mgraton, regon of orgn and health status n France. Frstly, we have compared the health status between mgrants and the natve populaton n dscernng an dentfable dfference between frst-generaton and second-generaton mgrants. Followng ths, n order to explan the heterogenety of health status amongst the mgrant populaton, we have refned our analyss by ntegratng ther country of orgn nto our estmaton and then explorng the health dfferences between ndvduals who have emgrated from South-East Medterranean (SEM) countres and ndvduals who have emgrated from all other countres. Our fndngs show that there exst health nequaltes that are related to mmgraton, when compared wth the health status of the natve populaton n France. Frst and second generaton mgrants have a hgher rsk than the natve French born populaton to report a poor health status. By ntroducng country of orgn nto our analyss we are able to confrm the health heterogenety wthn both groups of mgrants. Indvduals comng from SEM countres are more lkely to report poor health status than the natve French born populaton (for both generaton mgrants) and ths rsk seems sgnfcantly hgher for ndvduals who have emgrated from Turkey. These nequaltes are partly explaned by the poor soco-economc condtons of the mgrant populaton and a general lack of socal ntegraton n France. * LEDa-LEGOS (Unversté Pars-Dauphne), Place du Maréchal-de-Lattre-de-Tassgny, Pars Cedex 16. LEDa-LEGOS (Unversté Pars-Dauphne) et IRDES, Place du Maréchal-de-Lattre-de-Tassgny, Pars Cedex 16. IRDES, 10 rue Vauvenargues, Pars. 1

2 1. Introducton: Ths study explores the statstcal relatonshp between an ndvdual s mgratory status, ther country of brth and ther health status. Socal health nequaltes are well documented n general populaton n France, however few studes have focused on mgrant populaton due manly to the lack of nformaton on natonalty and country of brth provded by most health surveys (Jusot & al, 2009; Fassn, 2000). In 2004, mgrants represent represented 8.1% of the French populaton and there s evdence to suggest that the mgrant populaton s fundamentally dfferent from natve populaton n relaton to ther health condton. Poor soco-economc status, language dffcultes, stress due to new lvng condtons or dscrmnaton and a lack of specfc knowledge or nformaton about the structure and organsaton of the health care system are all factors that contrbute to mgrants lower health status and go some of the way of explanng socal health nequaltes (Sender, 2008; Attas- Donfut & Tesser, 2005; Chaouch, Casu & Caussder, 2006). Based on these factors, the mgrant populaton s consdered at frst glance as a hgh rsk group n socety wth regard to health. Paradoxcally a number of studes have shown that the mgrant populaton s on average n better health than the natve populaton n relaton to a number of key health ndcators. Ths Healthy Mgrant Effect suggests that people born overseas have generally better health than the natve born populaton. Ths hypothess, whch can be consdered as a selecton effect, assumes that only people wth good health status or who are ntally wealthy are more able and lkely to mgrate. The Healthy Mgrant Effect s well documented n both French and nternatonal lterature, however the fndngs are not smlar and there s no general consensus. For example, n the USA, Canada, Australa and the Unted Kngdom the mmgrant populaton s on average healther than the natve populaton (McDonald & Kennedy, 2004; Kennedy & al, 2006; Rubalcava & al, 2008). However n France, the results of recent studes prove that the mgrant or foregn populaton s more unhealthy than natve French-born populaton (Jusot & al, 2009; Attas-Donfut & Tesser, 2005; Lert, Melchor & Vlle, 2007). Ths selecton effect can be offset over tme by the deleterous effects of mgraton such as lonelness, a loose of socal support or poorer lvng condtons and unfavourable socoeconomc status. In fact, health captal models suggest that soco-economc condtons represent one of the most mportant socal determnants of an ndvdual s health (Grossman, 2000). A large body of lterature shows that an ndvdual s socal status wthn socety and ther materal lvng condtons are strongly correlated to the ndvdual s health status (Goldberg and al, 2002; Marmot and Wlknson, 2006). Studes on mgrant health have emphassed that such populatons are more lkely to be affected by unemployment, to have lower ncomes and a lower level of educaton (Newbold & Danforth, 2003; Attas-Donfut & Tesser, 2005; Jusot & al, 2009). In France for example, the unemployment rate among mmgrant s n 2007 s twce 2

3 as hgh as than of the natve born populaton, and the level of unemployment s even more pronounced for people who have emgrated from Turkey or Tunsa (Perrn-Haynes, 2008). The mmgrant populaton of France s over-represented amongst people employed n unsklled occupatons. Ths s largely due to the fact that many mgrants to France have no formal qualfcatons, whch s demonstrated by the fact that more than sxty percent of mmgrants from Turkey have no qualfcatons (Perrn-Haynes, 2008). Furthermore, a Canadan study has shown not only that mgrant populatons have more unfavourable socoeconomc condtons but also that ths s an mportant determnant whch helps to explan the dfference between mgrant health and that of the natve born populaton (Dunn and Dyck, 2000). Apart from the nfluence of materal lvng condtons on health status, some studes have stressed the mportance of factors relatng to socal ntegraton and more generally to psychosocal resources when explanng dfferences n levels of health. Psycho-socal resources refer to socal captal, socal relatonshps or emotonal and fnancal support. Accordng to Putnam (1995), socal captal refers to features of socal organsaton, such as trust, norms and networks that can mprove the effcency of socety by facltatng coordnated actons. Socal captal encompasses the nsttutons, relatonshps and norms that shape the qualty and quantty of a socety s socal nteracton and t may refer to an ndvdual s socal characterstcs that enable prvate returns va nteracton wth others (Glasear & al, 2002; Durlauf, 2002). Socal captal s actually consdered as a potental explanatory factor of an ndvdual s health status snce socal nteracton, trust and recprocty facltate people to access resources and to have expectatons towards others. Numerous studes have therefore suggested that a hgh level of socal captal enhances populaton health outcomes and reduces health dfferences (Golberg & al., 2002; Jusot, Grgnon & Dourgnon, 2008; Folland, 2007; Islam, 2007; Srven 2006). Due to adaptaton dffcultes n the host country, a lack and loss of socal or emotonal support, an mmgrant populaton may present hgh levels of psychosocal stress, whch n turn lead to a poorer health status. In fact, socal captal and psychosocal determnants appear to be a partcularly relevant health determnant for vulnerable populatons, of whch the mgrant populaton s, snce t consttutes nformal nsurance aganst health rsks, enablng a reducton n nformatonal costs and to a spread of health norms (Putnam 1995, 2000). Therefore, n relaton to the mgrant populaton there s a postve and strong assocaton between access to psycho-socal resources, health condtons and access to health servces (Zambrana & al, 2004; Leclere & al, 1994; Campbell & Mclean, 2002). More recently, Gresenz, Rogowsk and Escarse (2007) have shown that a large socal network s benefcal to the health status of the mmgrant populaton resdng n the US and partcularly for those who are lvng n an area wth a hgh concentraton of mmgrants because t mproves access to health servces. Fnally, studes have shown that a mgrant populaton could be consdered as a specfc group snce they have ther own health characterstcs. Hence, t has been proved that among the 3

4 mmgrant populaton health status may be heterogeneous. The natve country, the length of stay n the host country and the language barrer are all relevant determnants of an ndvdual mgrant s health (Attas-Donfut & Tesser, 2005; Lert, Melchor & Vlle, 2007). The country of orgn has an mportant mplcaton on an ndvdual s health and especally through the nfluence of the economc or poltcal context and the country s customs. Jusot & al (2009) have noted that ndvduals who have emgrated from countres whose GDP per capta s low (that s second or thrd quartle of GDP) are more lkely to report a poor health status than ndvduals who have emgrated from countres wth a hgher GDP. Moreover, the results are smlar f they ntroduce the country human development ndcator nto ther analyss. In ths way there s a clear protector effect of a country s development level on health status. Hence, ths study suggests that there s a long term effect of the economc stuaton of a mgrant s natve country on ther ndvdual health. Cultural habts (such as food consumpton or medcne patterns) may also explan a mgrant s health. Fndngs of Khlat and Courbage s (1995) study have shown that ndvdual who have emgrated from Morocco are more lkely than French people to beneft from a lower death rate due to a healthy det and lower alcohol consumpton. More recently, Gee, Kobayask and Prus (2007) ndcated that ndvduals resdng n Canada, who had emgrated from Asa, have a much hgher rsk of reportng chronc dsease. To explan ths result, the named authors suggest that Asan people encounter dffcultes n understandng the health care system or preventon programs. Fnally, to explan the health dspartes wthn the mgrant populaton, some authors have shown that the length of stay n the host country and the language barrer are postvely assocated wth the lkelhood of reportng poor health (Attas-Donfut & Tesser, 2005; McDonal & Nely, 2007; Lert, Melchor & Vlle, 2007; Zambrana & al., 1994; Leclere, Jensen & Bddlecom, 1994). In fact, the mgrant populaton may suffer from language dffcultes and thus the nformaton assocated wth the heath care system or the preventve acton may be msunderstood whch n turn leads to a poorer level of health. As for the natve populaton, the mgrant health depends not only on soco-economc status and psycho-socal resources but also on specfc determnants attached wth the mgratory hstory such as the length of stay, the language barrer, the mgratory status or the country of orgn. Taken together, these factors may explan dfferences n the levels of ndvdual health between natve born people and the mmgrant populaton but also wthn the mgrant populaton. Usng a representatve survey of the French populaton, the Health, Health Care and Insurance Survey (ESPS: Enquête sur la santé et la protecton socale ) we ntend to analyse the lnks between mgratory status, country of brth and health status by supplementng the exstng lterature n several ways. Apart from comparng the health status of the mgrant populaton and natve French populaton, we propose to analyse more precsely the health dspartes accordng to mgratory status through dstngushng between frstgeneraton and second-generaton mgrants. Due to ths analyse t s then possble to consder the healthy mgrant effect hypothess and to assess whether people who are descendents of mmgrants have a smlar level of health n comparson to the natve born populaton. 4

5 Moreover, the analyss s further refned by ntegratng the country of orgn varable nto our study n order to determne health heterogenety wthn both groups of the mgrant populaton. Hence, we attempt to explore the health dfferences between ndvduals who have emgrated from South-East Medterranean countres and ndvduals who have emgrated from all other countres. Ths dstncton enables to apprecate the extent of socal health nequaltes caused by a mgrant s regon of orgn. Fnally, n order to confrm the health determnants proposed by prevous lterature we explore the nfluence of soco-economc condtons and psychosocal resources on health status. The next secton ntroduces the data and varables used n the regresson analyses. The methodology and the estmaton strategy are also presented n ths secton. The results are presented n secton 3, followed by a concluson n secton Data and Method: The analyss s based on a populaton survey, representatve of the French populaton, the Health, Health Care and Insurance Survey (ESPS: Enquête sur la santé et la protecton socale ), coordnated by the Insttute for Research and Informaton n Health Economcs (IRDES). We have used the 2006 survey whch ncluded a set of queston on natve country, country of brth and psycho-socal resources. The survey sample, whch comprsed of 8100 households and ndvduals, s based on a random draw from the admnstratve fles of the man health funds of France whch over 90% of the French populaton are members of. Indvduals drawn at random from the admnstratve fles are used to dentfy households. The soco-economc questonnare has been answered by one key respondent from each household (aged at least 18 years old), who s not necessarly the ndvdual who was selected at random. The questons on health status are collected through a self-admnstered questonnare completed ndvdually by each household member. Questons on psycho-socal resources and natonalty are answered by the key respondent. Snce our man objectve s to examne the health dfferences between mgrants and the natve populaton, we have restrcted our analyss of the populaton to ndvduals aged 18 years old and over, who have reported both ther health status and ther natonal orgn (7260 ndvduals) Mgraton status and country of brth: To buld a mgratory status varable, we have used nformaton relatng to natonalty and country of brth of ndvduals and those of ther parents. Through ntegratng these questons, we have dentfed three dstngushable mgratory statuses: ndvduals who were born French and whose parents were born n France, Frst-generaton mgrants and Secondgeneraton mgrants. 5

6 Frstly, the populaton of ndvduals who were born French and whose parents were born n France represents n our analyss the reference populaton and t gathers ndvduals wth French natonalty whether they were born n France or not and whose parents were born n France. Secondly, the populaton of Frst-generaton mgrants gathers foregn ndvduals who were born abroad, regardless of ther parents natonalty and country of brth. Lastly, the second-generaton mgrant group represents ndvduals who are not foregners born abroad and who have at least one parent who was born abroad. To analyse the socal health nequaltes of ndvduals who come from SEM countres, we have used an ndvdual s country of brth for frst-generaton mgrants and then for second-generaton mgrants, the parent s country of brth. Hence, we constructed an ndcator varable named orgn n order to dstngush ndvduals or parents who have emgrated from Turkey, Morocco, Tunsa, Algera, Egypt, Israel, Lbya and Lebanon from ndvduals or parents who have emgrated from all other countres. Indvduals who were born French and who have French parents, whch consttutes the reference populaton, represent 80.9% of the sample (Table 1). 9% of the sample s composed of frst-generaton mgrants. Wthn ths group, 31.2% have emgrated from South-East Medterranean countres and 68.8% have emgrated from all others countres. Secondgeneraton mgrants represent 10.2% of the sample. Wthn ths last category, 23% have parents who have emgrated from SEM countres and almost 77% have parents who have emgrated from all other countres. Note that the frst-generaton mgrant populaton s on average older than the second-generaton one (49.2 years versus 45.3 years old), whch s tself younger than the French reference populaton (48.8 years old). Table 2 below shows that the majorty of the mgrant populatons comng from SEM countres are natonals from Maghreb (Morocco, Algera and Tunsa). Actually, among frstgeneraton mgrants who have emgrated from SEM countres, nearly 88% are natonals from Maghreb and the composton of the second mgrant generaton group s smlar, wth 89% natve to these countres. Note that ndvduals comng from Turkey or Mddle East are not well represented n our sample. Only 16 and 14 ndvduals have come from Turkey for the frst and second generaton groups respectvely. Smlarly, only 9 and 5 ndvduals have come from the Mddle East, Lbya and Egypt for the frst and second generaton mgrants groups. In ths way, these small samples may consttute a statstcal problem nto econometrc analyss. 6

7 Table 1. Statstcs descrptves : Orgn and mgratory status of the sample Characterstcs N % Mgratory status and orgn Mgratory status: French ,88 Frst mgrant generaton From SEM countres 203 2,81 From all other countres 447 6,19 Second mgrant generaton From SEM countres 169 2,34 From all other countres 561 7,77 Table 2. Statstcs descrptves: Orgn country of mgrant populaton Characterstcs Country of Orgn SEM countres Frst mgrant generaton N 203 Second mgrant generaton N 169 Morocco Algera Tunsa Egypt Lebanon Israel Lbye Turkey From all other countres The Health Status Assessment: Health status s dffcult to represent as a unque ndcator due to ts multdmensonal character. Accordng to the WHO, a good health status means not only the absence of dsease or njury but also physcal, mental and socal well beng. Mortalty and morbdty ndcators are the most common measures for health status and the latter s used n our study. To assess ndvdual health status, we use the frst of three standardsed questons suggested by the WHO European Offce relatve to self-assessed health. Ths ndcator reles on the followng queston: Would you say that your health s: very good, good, far, bad or very bad? Ths self-assessed health queston s a subjectve ndcator of an ndvdual s overall health status whch refers to the percepton of a person s health n general. It has the advantage of reflectng aspects of health not captured n other measures, such as: ncpent dsease, dsease severty, aspects of postve health status, physologcal and psychologcal reserves and socal 7

8 and mental functon. Ths ndcator may however suffer from ndvdual reportng heterogenety (Bago d Uva & al. 2008) and ts comparablty among natve and mmgrant populatons may be questoned. Some studes have shown that health percepton dffers accordng to health norms and ndvdual aspratons, whch relate to culture. Despte the varable s subjectvty, several studes have valdated ts utlsaton among ethnc groups and have shown that across ethncs groups a poorer self-assessed health status s constantly assocated wth hgher dsease prevalence rate (Chandola & al., 2000; Molnes & al., 2000; Jenknson & al., 2001). Ths ndcator has also been found to be a good predctor of mortalty (Idler & Benyamn, 1997). To study ndvduals health we have constructed a bnary health descrptor. Ths descrptor places people who have reported a very good or good general health status opposte people reportng a far, bad, or very bad general health status. Nearly 25% of the sample declared that ther own self-assessed health was poor. The descrptve analyss shows some health dfferences accordng to mgratory status and country of orgn (Table 3). On average, frst-generaton mgrants are more numerous n the poorer health category than the natve French populaton. Among the French populaton, 26.1% report poor self-assessed health whle 43.8% and 40.2% of frst-generaton mgrants who emgrated from SEM countres and from all other countres (respectvely) report the same results. Frst-generaton mgrants are also more numerous n the poorer health category when compared to second-generaton mgrants. Ths latter group seems, on average, to be almost dentcal to French populaton wth regard to health status. Note that 23.1% of secondgeneraton mgrants comng from SEM countres and 29.1% comng from all other countres report a poor health status. Table 3. Statstcs descrptves: Health status of the sample accordng to orgn and mgratory status (% row) Characterstcs Mgratory status and orgn Mgratory status: French Frst mgrant generaton From SEM countres From all other countres Second mgrant generaton From SEM countres From all other countres Poor self assesed health 26,1 43,8 40,2 23,1 29,1 Note: * Wthn the french populaton, 26.1% report a poor self assesed health. 4 Data not reported. 8

9 2.3. Psycho-socal resources measures: Psycho-socal resources, whch represent a proxy ndcator of socal ntegraton, can be assessed through the three dmensons usually used n the lterature: socal captal, socal support and sense of control at work. Socal captal s often measured at the ndvdual level through cvc engagement, whch refers to partcpaton n collectve actvty such as assocatons, sportng clubs, a relgous communty, unons or poltcal partes. Therefore, n ths research the study of socal captal s part of Putnam s framework whch refers to features of socal organsaton, such as trust, norms and networks that can mprove the effcency of socety by facltatng coordnated acton (Putnam, 1995). For socal support, we used a queston whch addresses whether ndvduals have suffered from lonelness durng ther lfe. Lastly, sense of control at work whch refers to an ndvdual s percepton towards ther poston n socety s measured va ndvdual autonomy at work. The last ndcator of socal ntegraton that was used refers to the language spoken durng chldhood, for whch there were three possble responses: to have spoken n French, to have spoken n French and another language or to have spoken only n language other than French. Descrptve statstcs ndcate that the dstrbuton of psycho-socal resources s unevenly dstrbuted wthn the sample (Table 4). In ths way, frst and second generaton mgrants are more represented n the categores of not havng any cvc engagement, to have suffered from lonelness and not to have any sense of control at work n comparson to the French populaton. Note that more frst-generaton mgrants than second-generaton mgrants suffer from lack of psycho-socal resources. Wthn both groups of mgrants, ndvduals comng from SEM countres are more numerous not to have partcpated n a collectve actvty or not to have autonomy at work than ndvduals comng from all other countres. Fnally and not surprsngly, only 10.3% and 9.4% of frst-generaton mgrants comng from SEM countres and all other countres respectvely have spoken French durng chldhood. Ths s n contrast to the share of second-generaton mgrants where ths s the case (respectvely 40.2% and 56.9% for ndvduals whose parents have emgrated from SEM countres and all other countres). Descrptve statstcs also ndcated that ndvduals wth a poor access to psycho-socal resources presented on average a poorer health status (Table 5). Hence, ndvduals who do not partcpate n collectve actvtes are more lkely to report poor self-assessed health than ndvduals who partcpate n some form of collectve actvty (31.7% versus 20.5%). Smlarly, ndvduals who have suffered from lonelness contrary to people who dd not are more lkely to have a poor self-assessed health status (47.2% versus 25.7%). Fnally, ndvduals who have spoken French and another language, or solely n a language other than French, are more lkely to report poor health than those who have spoken only n French. 9

10 Table 4. Statstcs descrptves: Socal ntegraton ndcators accordng to mgratory status and orgn (% col) Characterstcs French Populaton 1st Mgrant Generaton SEM countres 1st Mgrant Generaton All other countres 2nd Mgrant Generaton SEM countres 2nd Mgrant Generaton All other countres Cvc engagement Partcpaton 38,1 21,2 24,6 31,4 35,7 No partcpaton 61,9* 78,8 75,4 68,6 64,4 Lonelness Yes 7,5 21,2 21,9 14,1 9,8 No 92,5 78,8 78,1 85,9 90,2 Sense of control at work Yes 62,1 48,6 55,5 52,4 65,4 No 38,0 51,4 44,5 47,7 34,6 Spoken language French language 84,2 10,3 9,4 40,2 56,9 French & other language 11,2 17,7 9,6 36,1 23,2 Other language only 4,6 71,9 81,0 23,7 20,0 Note: * Wthn the French populaton, 61,9% do not have any cvc engagement. Table 5. Statstcs descrptves: Health status of the sample accordng to psycho-socal ressources (% row) Characterstcs Cvc engagement Lonelness Sense of control at work Spoken language Partcpaton No partcpaton Yes No Yes No French language French & other language Other language only Poor self assesed health 20,5 31,7 47,2 25,7 21,7 33,2 25,2 30,0 40,2 Note: * Wthn ndvduals who do not have any cvc engagement, 31.7% report a poor self assesed health. 10

11 2.4. Soco-economc varables: To assess the nfluence of soco-economc status on an ndvdual s heath, educatonal level, occupaton, actvty status, ncome and household composton are used. Level of educatonal s measured as follows: wthout qualfcaton, prmary level, frst level of secondary school, second level of secondary school, post secondary educaton and other level of educaton whch ncludes mssng values, foregn dplomas, professonal tranng and other educaton. There are four occupatonal statuses: n employment, non-workng, retred and unemployed. For our analytcal framework we also used the famous French Soco Professonal Category n whch 8 actvty statuses are defned: executve (used as reference), agrcultural employee, self-employed, ntermedary occupatons, admnstratve employee, busness employee, sklled worker, unsklled worker, non-workng. Income s measured as household ncome (from all sources of ncome), dvded by the OECD equvalent scale (1 for the frst household composton, 0.5 for the second and 0.3 for the thrd and followng one). We created an ncome quntle and a last category was bult whch refers to those who dd not provde ncome nformaton. Fnally, to assess the household composton we constructed 5 categores: couple wth chld (used as reference), sngle, sngle-parent, chldless couple and other household compostons. As prevously stated, the descrptve analyss proves some dfferences accordng to mgratory statuses and country of orgn (Table 6). Frst-generaton mgrants have on average more unfavourable soco-economc condtons than the natve French born populaton and ths s confrmed when all ndcators are consdered (educatonal level, occupaton and actvty statuses, ncome or household composton). Frst-generaton mgrants are, for nstance, more lkely not to have any qualfcatons, to be unemployed or to have a lower ncome than the natve French populaton. However, soco-economc status s not homogenous wthn the frst mgrant generaton group. Indvduals who have emgrated from SEM countres have a poorer soco-economc stuaton than those who have emgrated from all other countres. In contrast, the soco-economc stuaton of second-generaton mgrants s not always poorer than the natve French populaton. For example, second-generaton mgrants are more lkely to have a post secondary educaton level than the natve French born populaton but at the same tme are more lkely to be unsklled workers, unemployed or to have low ncomes (1 st quntle of ncome). Note that wthn the second-generaton mgrant group, ndvduals whose parents have emgrated from SEM countres have generally a more unfavourable stuaton. 11

12 Table 6. Statstcs descrptves: Soco-economc condtons accordng to mgratory status and orgn (% col) French Populaton 1st Mgrant Generaton SEM countres 1st Mgrant Generaton All other countres 2nd Mgrant Generaton SEM countres 2nd Mgrant Generaton All other countres Characterstcs Sex Male 40,5 54,6 40,7 37,3 41,9 Female 59,5 45,3 59,3 62,7 58,1 Age Age<30 14,4 12,3 8,3 37,3 14,1 30<=age<40 18,7 30,1 21,0 30,2 21,2 40<=age<50 21,5 23,2 21,0 16,0 18,4 50<=age<65 25,5 23,7 28,6 14,2 27,8 65<=age<75 10,7 5,9 10,1 1,2 11,9 age>=75 9,2 4,9 11,0 1,2 6,6 Educaton level Wthout qualfcaton 1,4 11,3 10,5 0,6 0,5 Prmary 18,6 16,3 22,2 5,3 16,2 1st level of secondary school 33,5 33,5 27,3 39,1 38,7 2nd level of secondary school 16,7 15,8 16,1 24,3 15,0 Post secondary educaton 29,8 23,2 23,9 30,8 29,6 Actvty Status Agrcultural employee 4,9 0,0 1,3 0,0 1,8 Self-employed 5,7 4,9 5,8 3,0 4,5 Executve 12,4 5,9 9,6 6,5 12,7 Intermedary occupatons 20,8 12,8 13,7 10,7 18,5 Admnstratve employee 17,8 8,9 11,4 18,3 21,8 Busness employee 12,4 13,8 22,4 17,8 13,0 Sklled worker 13,6 21,2 17,9 11,8 16,0 Unsklled worker 9,2 24,1 14,8 16,6 8,9 Non-workng 3,3 8,4 3,1 15,4 2,9 12

13 Table 6. Contnued Characterstcs French Populaton 1st Mgrant Generaton SEM countres 1st Mgrant Generaton All other countres 2nd Mgrant Generaton SEM countres 2nd Mgrant Generaton All other countres Occupaton status In employment 57,4 49,3 54,1 59,8 59,4 Non-workng 9,8 20,7 11,9 22,5 9,1 Retred 25,0 13,3 22,4 3,6 24,1 Unemployed 7,8 16,8 11,6 14,2 7,5 Income 1st Quntle 15,2 40,4 11,9 14,8 13,9 2nd Quntle 16,4 19,2 25,7 30,2 15,5 3rd Quntle 16,6 12,3 20,6 20,1 17,7 4th Quntle 18,1 7,4 17,2 13,6 17,5 5th Quntle 20,0 7,4 11,2 8,9 15,9 us 13,7 13,3 13,4 12,4 19,6 Household composton Sngle 19,4 20,7 20,1 10,1 18,0 Sngle-parent 7,8 7,9 11,0 13,6 7,8 Chldless couple 29,9 11,3 24,6 17,2 31,0 Couple wth chld 40,4 53,2 37,8 54,4 39,0 Other household composton 2,6 6,9 6,5 4,7 4,1 Note : * Among French populaton, 57,4% are actve. 13

14 2.5. Analytc Strategy: To analyse the lnk between mgraton, country of orgn and health status, we have run several bnary probt regressons wth margnal effect amed at studyng at the same tme the nfluence of mgratory status and country of orgn on the rsk of reportng a poor health status. Suppose that the bnary health varable H s the result of a contnuous latent health varable H, representng health status n a contnuous way. The observed dummy varable Η s defned by: Η =1 f Η >0 Η =0 otherwse Frst we ran a baselne probt analyss to assess the nfluence of orgn ( δ ) on the rsk of reportng a poor health status ( H ), controlled only by bologcal dmensons such age and gender ( D ). The average of health status n the sample s represented by the constant β 0 and the standard error e s assumed to follow a normal dstrbuton. Η = β 0 + αd + ρδ + e (Model 1) All soco-economc ndcators ( X ) were then ntroduced smultaneously n a second model to analyse the assocaton ceters parbus between self assessed health and orgn. Η = β 0 + αd + σχ + ρδ + e (Model 2) These two frst models enabled a dstncton between the drect effect of mgraton and the country of orgn on health status, from the ndrect effect whch passes through socoeconomc condtons. Through to these models, t s possble to assess the share of socal health nequaltes that s explaned on one hand by bologcal and materal factors and on the other hand by mgraton and country of orgn. To test further the nfluence of psycho-socal resources on health status, lastly we ntroduced n a thrd model ndcator representng socal ntegraton ( ψ ) that s: cvc engagement; socal support; sense of control at work and language spoken durng chldhood. Ths last analyss attempts to assess the share of socal health nequaltes that s explaned by a lack of psychosocal resources and more generally by a lack of socal ntegraton n France. Η = β 0 + αd + σχ + µψ + ρδ + e (Model 3) 14

15 Note that coeffcents α, σ, µ and ρ are estmated by the maxmum lkelhood methods under the assumpton that the resdual term e s uncorrelated wth the exogenous varable. These three analyses have been reproduced separately among men and women to test a dfferent mpact of orgn on a person s health status accordng to gender. In fact, the determnants of mgraton are dfferent for men and women. Consderng that men more commonly mgrate n search of new employment opportuntes than women, we may expect stronger health selectve mgraton n males than n females. Lastly, we performed an analyss n whch the two mgrant generaton groups comng from SEM countres are broken down nto two sub-groups: ndvduals comng from Turkey versus ndvduals comng from North Afrca or the Mddle East. Smlarly to the frst analyss, we replcated the three models (that s a baselne model n whch only sex, age and orgn are entered, followed by a second model n whch soco-economc condtons are then ntroduced and fnally a thrd model n whch socal ntegraton proxes are entered next to other control varables). The goal of ths exploratory study was to try to fnd any relatonshp between orgn and health status and more specally to assess the share of socal health nequaltes that can be explaned by the fact that a mgrant emgrated from South-East Medterranean countres. Is there any assocaton between health and mgratory status? Among mgratory status, s there any dfference n health accordng to a person s regon of orgn? Note that all models have been estmated usng ordered probt analyses snce our health varable (namely the self-assessed health) was orgnally a multnomal ordered and dependant varable. In theory, ths model enabled us to refne the analyss and to estmate more accurately the nfluence of mgraton and country of orgn on health across the transton between bad health status and good health status. However, we dd not retan ths estmaton strategy snce the consstency assumpton test of effects across dfferent categores was rejected. The lkelhood rato test whch was performed ndcated that slopes were not equal across our fve health categores. Therefore, we decded to mantan our dependant varable n a bnary health descrptor. 15

16 3. Fndngs: Table 6 presents the results of a probt analyss amed at studyng the ndvdual determnants of health status and mgrant health heterogenety accordng to ther mgratory status and ther orgn. Model 1, whch contans only a control varable of bologcal factors (age and sex), shows that mgratory status has a sgnfcant effect on the rsk of reportng a poor self-assessed health status (column 1, table 7). Frst and second generaton mgrants have a sgnfcant hgher rsk than the natve French born populaton to report a poor health status and ths rsk s dssmlar among the mgrant populaton. Indvduals comng from SEM countres are more lkely to be n the poorer health category, whether they belong to the frst generaton or second group. Ths result s consstent wth prevous French studes that show the poor health condtons of the mgrant populaton n France, whch contrasts wth the healthy mgrant effect hypothess (Attas-Donfut & Tesser, 2005; Lert & al., 2007; Jusot & al., 2009). As expected, the probablty of reportng a poor health status s hgher for women and also ncreases wth age. The control varable for soco-economc condtons provdes dfferent results (model 2). Frst, the decrease n margnal effects assocated wth mgratory status between model 1 and 2 shows that the poor health status of mgrants s partly explaned by ther more unfavourable soco-economc condtons. In spte of ths decrease, frst-generaton mgrants stll have a sgnfcantly hgher rsk than the natve French populaton to report poor health status. Even after control for soco-economc condtons, the rsk of beng n the poorer health category s hgher for people who have emgrated from SEM countres (margnal effect equals 0.14 and s sgnfcant to the 1% level) than people who have emgrated from all other countres (margnal effect equals 0.07 and s sgnfcant to the 5% level) 5. Ths result suggests that mgraton and orgn have a detrmental effect that s ndependent of soco-economc condtons of mmgrants n the host country. Among second-generaton mgrants, only ndvduals whose parents have emgrated from all other countres but not SEM countres have a sgnfcant hgher rsk than the natve French born populaton to report, ceters parbus, a poor self-assessed health status. Thus, the detrmental effect of mgraton ndependently of soco-economc condtons s agan observed for ths sub-group of second-generaton mgrants. However, ths effect s not verfed for second-generaton mgrants whose parents have emgrated from SEM countres and thus t seems that ther poor health status s entrely explaned by poor materal condtons. Furthermore, the fndngs of model 2 confrm the nfluence of soco-economc condtons on health status proved by prevous studes relatng to socal health nequaltes. All soco- 5 Note that the equalty test of margnal effect has been accepted. 16

17 economc varables have a sgnfcant effect on health and n the expected way (Goldberg & al., 2002; Cutler, Lleras-Muney & Vogl, 2008). Indvduals wthout any qualfcatons and those wth a prmary level of educaton are more lkely to report poorer health status compared to ndvduals wth a post-secondary educaton level and sklled or unsklled workers compared to executves. Indvduals also are more lkely to report a poor health status when they are nactve, unemployed or a sngle parent. Household ncome has a strongly sgnfcant effect on self-assessed health status snce t reduces the rsk of reportng a poor health status. Model 3 provdes the results of the thrd analyss where socal ntegraton ndcators (.e. psycho-socal resources) are ntroduced nto the regresson n addton to orgn, socoeconomc condtons, age and sex. Introducng psycho-socal resources nto the model substantally modfes the results and especally the nfluence of a person s country of orgn n the explanng health status. Wthn the frst-generaton mgrants group, only people who had emgrated from SEM countres had a sgnfcant hgher rsk of reportng a poor health status than the French born populaton. Hence the natve country, and especally beng born n SEM countres, generates a detrmental effect on self-assessed health ndependently of economc condtons and socal ntegraton. Ths result suggests that apart from the effect of soco-economc condtons and socal ntegraton, there are stll some hdden factors whch have a detrmental mpact on ther health status. Conversely, the effect of havng emgrated from countres other than those of the SEM dd not reman sgnfcant on the rsk of reportng a poor health status after control for psycho-socal resources. Thus, there s no more sgnfcant dfference n health status between ths sub-group of frst generaton mgrants and the natve French born populaton. Ths suggests that the poor self-assessed health of mgrants from those countres s manly explaned by ther poor access to psycho-socal resources, n addton to ther dsadvantaged soco-economc condtons. Despte a slght decrease n margnal effects, the results concernng the second-generaton mgrants dd not change after the ntroducton of psycho-socal resources. The self-assessed health status of the secondgeneraton mgrants group whose parents had emgrated from SEM countres s not sgnfcantly dfferent from the natve French born populaton whereas those whose parents have emgrated from other countres have a hgher rsk of beng n a poorer health status than the French born populaton. All socal ntegraton ndcators are strongly assocated wth the rsk of reportng a poor selfassessed health status. Havng spoken n French and another language durng chldhood n comparson to havng spoken only n French ncreases the rsk of reportng a poor health status. We dd not fnd any sgnfcant effect of havng spoken only n another language. Indeed, there s no sgnfcant dfference between ndvduals who have spoken only n another language and those who have spoken n French. In addton there s a clear assocaton between the three psycho-socal resources and the probablty of an ndvdual declarng a poor self-assessed health status. Hence, ndvduals who do not have any collectve 17

18 partcpaton, who dsagree that they have autonomy n ther work and who have suffered from lonelness, have a hgher rsk to be n the poorer self-assessed health category (margnal effects sgnfcantly dfferent from 0 at the 1% level). These results confrm prevous lterature on socal health nequaltes (Srven, 2006; Folland, 2007; Islam, 2007; Jusot & al., 2008). The analyss of the determnants of poor access to psycho-socal resources 6 confrms also the contrbuton of ths dmenson to socal health nequaltes related to mgratory status. Actually, soco-economc condtons, mgratory status and orgn play a consderable role n socal ntegraton. Hence, ndvduals wth lower levels of educaton, ncome and more generally unfavourable soco-economc condtons suffer on average from less access to psycho-socal resources. Smlarly, the mgrant populaton tends to have less access to these resources and s less lkely to be socally ntegrated nto socety than the natve French born populaton. For example, mgrants partcpate sgnfcantly less often and frequently have less emotonal and socal support. 6 Fndngs not reported. 18

19 Table 7. Influence of mgratory status, orgn and socal ntegraton on the rsk to report a poor health status Poor self assesed health Poor self assesed health Poor self assesed health Characterstcs Model 1 Model 2 Model 3 Mfx p-value Mfx p-value Mfx p-value Male Female 0,03 0,00 ** 0,00 0,91 0,00 0,95 Age Age 0,01 0,00 *** 0,01 0,00 *** 0,01 0,00 *** Mgratory status: French Frst mgrant generaton From SEM countres 0,25 0,00 *** 0,14 0,00 *** 0,10 0,01 ** From all other countres 0,12 0,00 *** 0,07 0,01 ** 0,03 0,29 Second mgrant generaton From SEM countres 0,11 0,02 ** 0,05 0,25 0,02 0,56 From all other countres 0,04 0,05 * 0,05 0,03 ** 0,04 0,08 * Post-secondary educaton Wthout certfcate 0,19 0,00 ** 0,16 0,00 ** Prmary 0,09 0,00 *** 0,07 0,00 ** 1st level of secondary school 0,05 0,01 ** 0,03 0,06 * 2nd level of secondary school 0,01 0,46 0,01 0,80 Other level of educaton 0,05 0,54 0,03 0,72 SES: Executve Agrcultural employee 0,01 0,72 0,02 0,52 Self-employed 0,00 0,99 0,01 0,84 Intermedary occupatons 0,01 0,70 0,01 0,77 Admnstratve employee 0,07 0,01 ** 0,05 0,05 ** Busness employee 0,08 0,01 ** 0,06 0,04 ** Sklled worker 0,07 0,02 ** 0,05 0,06 * Unsklled worker 0,09 0,01 ** 0,06 0,05 * Non-workng 0,00 0,93 0,00 0,92 Occupaton : Actf Non-workng 0,19 0,00 *** 0,17 0,00 *** Retred 0,02 0,41 0,02 0,37 Unemployed 0,16 0,00 *** 0,14 0,00 *** Income: 5th quntle 1th quntle 0,18 0,00 *** 0,16 0,00 *** 2nd quntle 0,11 0,00 *** 0,10 0,00 *** 3rd quntle 0,05 0,03 ** 0,04 0,06 * 4th quntle 0,05 0,03 ** 0,04 0,06 * Unknown 0,06 0,01 ** 0,05 0,04 ** Household composton: Couple wth chld To be alone 0,06 0,00 ** 0,05 0,00 ** Sngle-parent 0,07 0,00 ** 0,06 0,01 ** Chldless couple 0,04 0,02 ** 0,04 0,02 ** Other household composton 0,06 0,10 * 0,04 0,23 French language French and other language 0,03 0,07 * Other language 0,02 0,33 Collectve Pratcpaton No collectve partcpaton 0,05 0,00 *** To have autonomy at work To have no autonomy at work 0,04 0,00 *** Not applcable 0,03 0,19 To not have suffered from lonelness To have suffered from lonelness 0,14 0,00 *** No answer 0,00 0,89 N Pseudo R² (Mc Faden) 0,12 0,00 *** 0,19 0,00 *** 0,20 0,00 *** Log L -3383,7-3132,7-3092,5 Legend :* p<0,1; ** p<0,05; *** p<0,01 19

20 Conductng separate analyses for both sexes we found dfferent assocatons between mgraton, orgn and health status (Table 8 & 9). The frst column of both tables shows that mgratory status does not work dentcally for men and women after control for bologcal factors. Among women, frst and second generaton mgrants have a hgher rsk to be n a poorer health category than the natve French born populaton (Table 8) whereas ths rsk s only hgher for frst-generaton mgrants among men (Table 9). In both genders, the rsk s hgher for people comng from SEM countres than for those comng from all other countres, whch s consstent wth the prevous analyss (Table 7). The control for soco-economc condtons reveals some nterestng patterns, especally for men (column 2 of Tables 8 & 9). Whle all mgratory status remans sgnfcantly assocated wth a poor health status for women, ths s not observed for men. Indeed, there s no sgnfcant dfference n health status between the mgrant populaton (whether they belong to the frst or second generaton) and the natve French born populaton. These results suggest that the poor health status of male mmgrant populaton s entrely explaned by ther more unfavourable soco-economc condtons n France. For women however, margnal effects assocated wth mgratory status are stll sgnfcant whch suggests a detrmental effect of mgraton on health ndependently of soco-economc condtons. Once agan, for frst and second generaton mgrants the rsk of reportng a poor health status seems hgher for women comng from SEM countres (margnal effects equal to 0.22 and 0.10 and s sgnfcant at the 1% and 10% level respectvely). Note that the effect of soco-economc condtons on health status s qute smlar among men and women. The only dfferences are that educaton s more strongly assocated wth a women health status than to men, whereas soco-economc poston s not assocated wth health status n women but s n men. Apart from the occupaton s modalty retred whch s negatvely assocated wth a poor health status for men, we found assocatons of the same sgn and magntude. Introducng socal ntegraton ndcators nto the model (column 3 of Table 8 & 9) consderably affected the results concernng the nfluence of orgn n the explanaton of health status but only for women. Among women, only frst-generaton mgrants from SEM countres had a hgher rsk than the natve French born populaton to be n the poorer health category (margnal effect equals to 0.15 and sgnfcant at 5%). Ths suggests that f socoeconomc condtons and socal ntegraton largely explan the poor health status of the female mgrant populaton, there s stll a hazardous effect on health of beng born n a SEM country. However, wth regard to men there are stll no sgnfcant dfferences between mgrant populaton and French populaton, even f socal ntegraton s strongly assocated to health status n male populaton. The nfluence of soco-economc condtons on health s, n general terms, qute smlar even after control for socal ntegraton ndcators and for both analyses we fnd agan the same assocatons between socal ntegraton ndcators and health status. 20

21 Table 8. Influence of mgratory status, orgn and socal ntegraton on the rsk to report a poor health status (Women only) Poor self assesed health Poor self assesed health Poor self assesed health Characterstcs Model 1 Model 2 Model 3 Mfx p-value Mfx p-value Mfx p-value Age Age 0,01 0,00 *** 0,01 0,00 *** 0,01 0,00 *** Mgratory status: French Frst mgrant generaton From SEM countres 0,30 0,00 *** 0,22 0,00 *** 0,15 0,01 ** From all other countres 0,15 0,00 *** 0,11 0,00 ** 0,05 0,18 Second mgrant generaton From SEM countres 0,15 0,00 ** 0,10 0,05 * 0,06 0,23 From all other countres 0,06 0,05 ** 0,06 0,03 ** 0,05 0,10 Post-secondary educaton Wthout certfcate 0,21 0,00 ** 0,17 0,02 ** Prmary 0,15 0,00 *** 0,12 0,00 *** 1st level of secondary school 0,07 0,01 ** 0,05 0,03 ** 2nd level of secondary school 0,04 0,16 0,02 0,36 Other level of educaton 0,10 0,33 0,08 0,45 SES: Executve Agrcultural employee 0,00 0,96 0,01 0,88 Self-employed -0,01 0,84 0,00 0,98 Intermedary occupatons 0,00 0,96 0,00 0,96 Admnstratve employee 0,06 0,13 0,04 0,27 Busness employee 0,07 0,09 * 0,05 0,21 Sklled worker 0,06 0,20 0,05 0,29 Unsklled worker 0,07 0,12 0,05 0,29 Non-workng 0,02 0,69 0,02 0,70 Occupaton : Actf Non-workng 0,15 0,00 *** 0,14 0,00 *** Retred 0,05 0,06 * 0,05 0,05 * Unemployed 0,17 0,00 *** 0,15 0,00 *** Income: 5th quntle 1th quntle 0,14 0,00 *** 0,13 0,00 *** 2nd quntle 0,08 0,01 ** 0,06 0,03 ** 3rd quntle 0,03 0,36 0,02 0,54 4th quntle 0,06 0,02 ** 0,06 0,05 ** Unknown 0,03 0,34 0,02 0,51 Household composton: Couple wth chld To be alone 0,07 0,00 ** 0,07 0,01 ** Sngle-parent 0,05 0,04 ** 0,05 0,09 * Chldless couple 0,02 0,46 0,01 0,50 Other household composton 0,05 0,22 0,03 0,42 French language French and other language 0,05 0,03 ** Other language 0,03 0,32 Collectve Pratcpaton No collectve partcpaton 0,06 0,00 *** To have autonomy at work To have no autonomy at work 0,04 0,02 ** Not applcable 0,01 0,86 To not have suffered from lonelness To have suffered from lonelness 0,16 0,00 *** No answer 0,01 0,73 N Pseudo R² (Mc Faden) 0,12 0,00 *** 0,18 0,00 *** 0,19 *** Log L -2046,3-1900,3-1870,4 Legend :* p<0,1; ** p<0,05; *** p<0,01 21

22 Table 9. Influence of mgratory status, orgn and socal ntegraton on the rsk to report a poor health status (Men only) Poor self assesed health Poor self assesed health Poor self assesed health Characterstcs Model 1 Model 2 Model 3 Mfx p-value Mfx p-value Mfx p-value Age Age 0,01 0,00 *** 0,01 0,00 *** 0,01 0,00 *** Mgratory status: French Frst mgrant generaton From SEM countres 0,20 0,00 *** 0,07 0,13 0,05 0,34 From all other countres 0,08 0,04 ** 0,02 0,69 0,00 0,99 Second mgrant generaton From SEM countres 0,00 0,97-0,05 0,44-0,06 0,39 From all other countres 0,02 0,56 0,02 0,44 0,02 0,52 Post-secondary educaton Wthout certfcate 0,20 0,02 ** 0,17 0,05 ** Prmary 0,03 0,34 0,02 0,53 1st level of secondary school 0,03 0,30 0,02 0,45 2nd level of secondary school -0,01 0,71-0,02 0,55 Other level of educaton 0,02 0,88 0,01 0,95 SES: Executve Agrcultural employee 0,02 0,67 0,03 0,49 Self-employed 0,01 0,85 0,02 0,70 Intermedary occupatons 0,02 0,61 0,02 0,63 Admnstratve employee 0,10 0,04 ** 0,08 0,08 * Busness employee 0,06 0,37 0,04 0,50 Sklled worker 0,08 0,02 ** 0,07 0,06 * Unsklled worker 0,11 0,01 ** 0,08 0,07 * Non-workng -0,11 0,11-0,14 0,03 ** Occupaton : Actf Non-workng 0,38 0,00 *** 0,36 0,00 *** Retred -0,06 0,06 * -0,06 0,07 * Unemployed 0,13 0,00 *** 0,11 0,00 ** Income: 5th quntle 1th quntle 0,22 0,00 *** 0,20 0,00 *** 2nd quntle 0,14 0,00 *** 0,13 0,00 *** 3rd quntle 0,08 0,01 ** 0,07 0,03 ** 4th quntle 0,01 0,64 0,01 0,80 Unknown 0,10 0,01 ** 0,09 0,02 ** Household composton: Couple wth chld To be alone 0,06 0,02 ** 0,05 0,07 * Sngle-parent 0,12 0,01 ** 0,10 0,03 ** Chldless couple 0,07 0,00 ** 0,07 0,01 ** Other household composton 0,08 0,19 0,07 0,23 French language French and other language 0,01 0,82 Other language 0,02 0,63 Collectve Pratcpaton No collectve partcpaton 0,04 0,04 ** To have autonomy at work To have no autonomy at work 0,05 0,01 ** Not applcable 0,17 0,00 ** To not have suffered from lonelness To have suffered from lonelness 0,09 0,01 ** No answer -0,04 0,32 N Pseudo R² (Mc Faden) 0,14 0,00 *** 0,23 0,00 *** 0,23 0,00 *** Log L -1332,8-1199,6-1185,6 Legend :* p<0,1; ** p<0,05; *** p<0,01 22

23 The only dfference s that the language spoken durng chldhood nfluences only women s health status. Otherwse, the lack of cvc engagement, no sense of control at work and no socal support ncrease sgnfcantly the rsk to be n the poorer health category for both women and men. The last analyss (Table 10) was performed by breakng down the two mgrant generaton groups comng from SEM countres nto two sub-groups: ndvduals comng from Turkey versus those comng from North Afrca or the Mddle East. Model 1 s consstent wth results prevously shown snce frst and second generaton mgrants have a hgher rsk of beng n poorer health status after control for bologcal factors. Ths rsk s agan dssmlar among the mgrant populaton. Wthn both mgrant generaton groups, the rsk to report a poorer health status s hgher for ndvduals comng from SEM countres. Moreover, among both groups the effect s strongest for those comng from Turkey. After control for soco-economc condtons, frst-generaton mgrants stll have a hgher rsk than the natve French born populaton to report poorer health status. Wthn ths group, the estmated rsk of poorer health status s hgher for people who have emgrated from Turkey than for people who had emgrated from North Afrca or the Mddle East and fnally for those who had emgrated from all other countres. The effect assocated wth beng born n Turkey s therefore the strongest one (margnal effect equal to 0.24), even though t s only sgnfcant at the 10% level. 7 Among the second-generaton mgrants group, only people whose parents have emgrated from Turkey and from all other countres (that s not SEM countres) stll have a hgher rsk of beng n the poorer health status category. The decrease n margnal effects assocated wth mgratory status between model 1 and 2 shows that soco-economc condtons partly explan the poorer health status of ths group of the mgrant populaton. Unlke the poor health status of second-generaton mgrants whose parents have emgrated from North Afrca or the Mddle East, ths seems to be entrely explaned by soco-economc factors. When socal ntegraton ndcators are ntroduced nto the model (model 3), the results are partly altered. Among the frst-generaton mgrants, only people who have emgrated from North Afrca and Mddle East stll have a hgher rsk of reportng a poorer health status than the natve French born populaton. Therefore there are no more dfferences between ndvduals who have emgrated from Turkey or from all other countres and the French populaton regardng self-assessed health. These results confrm that a lack of socal ntegraton and poor soco-economc condtons largely explan the poor health status of ndvduals who have emgrated from Turkey or from all other countres. Even so, there s a detrmental effect of mgraton for those who have emgrated from North Afrca or the Mddle East whch s ndependent of soco-economc condtons or socal ntegraton. Fndngs concernng the second mgrant generaton are smlar to those of model 2, even after control for psycho-socal resources. Hence, only people whose parents have emgrated from Turkey 7 Due to the sample sze of ths mgrant populaton, we suspect a statstcal robustness problem. 23