Immigration and ethnic change in low-fertility countries towards a new demographic transition?

Transcription

1 Immigration and ethnic change in low-fertility countries towards a new demographic transition? Paper presented to Population Association of America Annual Meeting, Philadelphia March 31 April , Session 98, I April. David Coleman, University of Oxford Sergei Scherbov, Vienna Institute of Demography Abstract This paper presents a comparative analysis of the likely future effects of immigration on the ethnic or foreign-origin composition of the developed world, using projections from various European countries, with new projections for the United Kingdom. On conservative assumptions, the foreign-origin proportions are projected to rise to between about 20% and 30% by mid-century with almost linear rates of change. Variant migration assumptions are more important than fertility assumptions in determining this outcome. Interethnic unions, incorporated only rarely in such projections, moderate the growth of immigrant populations of immigrant origin while generating complex rapidly growing heterogeneous mixed origin populations which may eventually defy categorisation. Given the dependence of this ethnic change upon immigration, it is not inevitable, but most official projections show no end to the process up to mid-century. Further continuation of ethnic transformation into the longer term might be regarded as a third demographic transition. Introduction While fluctuating and, in 2004, not generally increasing, immigration to Western Europe has persisted at a high level. It is likely to stay high for some time, and may increase further. Labour demand from some European countries will grow, especially from those unable or unwilling to mobilise their demographic reserves into their workforce (European Commission 2004). Some ethnic minority populations, which are increasing rapidly in numbers, show little sign of abandoning their preference for arranged marriage with spouses from the countries of origin (e.g. Lievens 1999). Asylum claims to Europe fell to 314,300 in 2004, the lowest since 1997 (UNHCR 2005) but still a substantial number. Natural increase, immigration and contemporary population growth Migration is becoming the driving force behind demographic change in many European countries. In a few, for example Norway and France, with robust natural increase and moderate migration, the current migration effect is relatively minor. In others, in Italy and Germany, where deaths exceed births, net immigration now accounts for all remaining population growth (Héran 2004, OECD 2003, Salt 2003). Immigration adds to population growth both directly and indirectly through the natural increase of populations of immigrant origin, over several generations. The natural increase of foreign-origin populations is often greater than that of the indigenous population, thanks to a more youthful age-structure, higher age-specific fertility rates, and transient distortions in family - building arising from the migration process itself (Thompson 1982, Toulemon 2004). It should be noted, however, that 1

2 immigrant groups may have lower, not higher, age-specific birth -rates compared with the national average. The overall effect, as in Germany, may be to create overall omnibus foreign-population fertility rates little higher than the national average, even if their more youthful age-structures generate a higher level of natural increase. In France between , TFR was raised from 1.65 among natives only, to 1.72 including immigrants in France (Héran 2004). Some examples are given in Figure 1 below Figure 1 TFR of native and foreign populations - selected European countries Sources: SOPEMI, national statistical offices. E&W New Commonwealth France all women France all foreign Sweden all foreigners Dutch Turks in the Netherlands Swedes England and Wales all birthplaces The foreign-origin population - a broader picture Evaluation of the number of people of foreign origin in European countries is complicated by problems of categorisation. Most European counties routinely define foreign origin populations on the criterion of citizenship, and births of foreign origin by the citizenship of the mother. That creates confusion. Recent high levels of naturalisation in many countries have exceeded the inflow of immigrants. That has diminished statistically, but not in reality, the numbers of people, and of mothers, of foreign origin in countries such as Belgium, France and the Netherlands. To present a more accurate picture, some European countries are adopting statistical definitions of foreign origin population that include the second immigrant generation, although not the third. For example, in the Netherlands the foreign background population now defined statistically through population registers, includes any person with a parent born abroad. (Alders 2001). Everyone else, including persons with third generation foreign ancestry, is assumed to be Dutch. No third generation or subsequent generation is included, leading to a progressive under-estimate of the population or foreign origin compared with more enduring ethnic or racial criteria used, for example in the US and UK population estimates and projections. This conservative approach 2

3 yielded a foreign origin population of 3.04 million out of the 16 million total population in the Netherlands in 2003 (19%), compared with the 700,000 persons of foreign citizenship. Furthermore, the foreign origin population has increased fast, unlike the foreign citizen population, which has declined since 1995 (Figure 2). Figure 2 Foreign citizen and 'foreign background' population, Netherlands (thousands). Source: CBS. For definitions see text Foreign citizens All foreign origin First generation, all foreign origin Second generation, all foreign origin f i i i Note: Persons of from the Dutch West Indies of Antillean origin, and of Surinamese origin before 1980, are of foreign origin but are not foreign citizens. Ethnic classifications Where the cultural characteristics of immigrants, and their self-identity, endure over generations an ethnic classification may give a more truthful picture of demographic and other consequences of the migration process, as long as identity and the official categories are stable and inter-ethnic unions are not too common. Such ethnic classifications are favoured in the countries of the English-speaking world, both for new immigrants from outside Europe and for old or indigenous minorities (Lee 1993, Coleman and Salt 1996, Statistics Canada 1993). In the UK, the 1991 and 2001 censuses and the Labour Force and General Household Surveys have requested selfidentification to an ethnic group. No other Western European country has done so apart from the UK. In France the concept has traditionally been considered to be fundamentally contrary to the principles of the equality of citizenship (Haut Conseil 1991), and the collection of statistics on ethnic criteria is forbidden. In some East European countries however, ethnic questions do feature in the census and in other statistics (e.g. Bulgaria, Romania) mostly for indigenous minorities such as Magyars, Turks, Gypsies, Germans and others (Courbage 1998). 3

4 Population projections Immigration, then, contributes generally to national population growth and specifically to the growth of foreign-origin populations, through the immigration process and through natural increase. That contribution can be substantial. In Great Britain, for example, the non-white ethnic minority population (4.5 million in 2001), irrespective of nationality or birthplace, increased from 1991 to 2001 by about 5% per year. Some of its components grew even faster: the population of African origin has doubled every ten years, it was 108,000 in 1981, 212,000 in 1991 and 440,000 in 2000 growing at 7.9% per year in the 1990s. In the Netherlands, the foreign background population grew on average by 2.7% per year between 1995 and That is important for two reasons. First, because it promises to add substantial numbers to the total population size of various European countries where growth would otherwise tail off, or turn to decline: for example Belgium, Denmark, the Netherlands and Sweden (see e.g. Haug et al. 2002, Poulain 2003 pp 85-86, Nilsson 2004 p.117). Second, because most of the additional population comes from origins different from those of the original inhabitants, increasing their ethnic and racial diversity. The ethnic (cultural) diversity, often originally closely linked with visible (racial) differences, may remain permanent or it may become dissociated from it as time goes on. That will depend partly on national government policy, partly on the scale of inflow, partly on the preferences of the populations concerned: some may become assimilated (like the Poles and Italians in pre-war France) or remain encapsulated: like the Turks in Belgium (Lesthaeghe 2000) and the Bangladeshis in the UK (Eade et al. 1996). National policies promoting assimilation or multicultural diversity, and requiring or forbidding official ethnic labelling, may have strong effects here, although not always in the intended direction. The trend in numbers of ethnic populations is of considerable social and political interest both to the native population and to the ethnic populations themselves, especially in those countries that have developed group rights and entitlements. Sometimes welfare considerations are ostensibly paramount, as in the projections for Denmark, intended to foresee the likely impact of their relative size on the labour market, education, day-care and care of the elderly and in those forthcoming for the UK (Think Tank on Integration in Denmark 2002; Haskey 2002). Population projections to evaluate these future effects of immigration and minority growth upon future population and society have been prepared by a number of European statistical offices and by demographic researchers. Not all have done so, both for political and for technical reasons. European statistical systems do not make it easy to determine current levels of foreign origin population, fertility, mortality or migration, quite apart from assumptions about their future levels. None records the race or ethnic origin of vital events, unlike, for example, the United States. Usually, births and deaths are recorded by citizenship or (as in the UK) by the birthplace of the mother or of the deceased. Current and future levels of mortality do not pose many difficulties. Such evidence as exists suggests that foreign-origin mortality is not very different from that of the national population and in some cases is less heavy. Fertility levels present more difficulties. In general, the fertility of foreign populations in industrial countries has tended to converge somewhat on the national average, as would be expected from demographic transition theory. But only in a few cases is that process complete. Fertility differences may persist if immigrant groups do not achieve socio-economic equality, if they retain strong attachment to religious or other 4

5 elements of foreign culture, if they continue to be numerically and culturally reinforced by large-scale migration, especially through importing spouses from highfertility countries. Their actual minority status per se may make some groups resistant to change (e.g. Siegel 1970, Goldscheider 1999, MacQuillan 2004). For example, among Indians in the UK and people of Caribbean origin in the Netherlands and the UK fertility rates have fallen to about the same as the national average. Among Turks and some other Muslim populations where women are less in contact with the wider society and usually do not work, fertility has fallen much less (Coleman, 1994a). In Great Britain in 2001, the TFRs of women born in Pakistan and Bangladesh were 4.67 and 3.89 respectively (ONS 2004). Reinforcement of the unacculturated population by arranged marriage and other migratory processes may sometimes drive up immigrant fertility rates, as among the African population in Sweden and the UK. Examples of projections of the population of immigrant origin The best known examples are those made in respect of the United States by the US Bureau of the Census (1996, 2000) and by the US National Research Council (Smith and Edmonston 1997). These are too familiar to need repetition. Projections of the foreign-origin population in Denmark Recent Danish projections incorporated seven categories: the Danes themselves, immigrants and their descendants from developed countries, immigrants and their descendants from more developed, and from less developed, third world countries on the UN classification. (Think Tank on Integration in Denmark 2002). Coefficients derived from information from the population registers on origins and on naturalisations determines the distribution of the numbers of descendants and Danes among the offspring of immigrants born in Denmark in the projections. Children with one parent who is a Danish citizen born in Denmark are Danish. If both parents are foreign or born abroad, the child is a descendant. Children of two descendants who choose not to naturalise, or of one such with an immigrant, are also descendants, and, potentially, subsequent generations (Larsen, C., pers.comm.). In 1996, the TFR of foreigners from more advanced countries (like the Western foreign population defined in the Dutch projections) was 1.73, about the Danish average. This was assumed to increase, in step with that of the Danish population, to 1.85 by 2020, the limit of the projections at that time. Net immigration from such countries was assumed to be 6200 per year. The TFR of the less developed countries, equivalent to the Non-Western foreigners in the Dutch classification, was 3.12 in The principal projection assumed that would decline to 2.1 by 2020, a variant assumed no change. Net immigration from those countries was assumed to be 7000 per year. To put these modest figures in perspective with the larger European countries such as France, Italy or the UK, a multiplier of about 11 would be appropriate. On the relatively short time scale to 2020, those fertility variants make a surprisingly small difference, thanks to the youthful age-structure of the immigrant generation (Danmarks Statistik 1997). Overall, 6.1% growth in total population to 2020 was projected from 1998 (Figure 3). Without immigration, population would decline slightly. Later projections lack a zero-migration variant so the total demographic 5

6 effect of immigration cannot be estimated. But in the later, 2002-based projections the non-danish population amounted to 14% of the total by 2020, and in 2050, according to the 2004-based projections, to 18%. Figure 3 Denmark : Danish, non-danish and total populations. Source: Statistics Denmark 2002, Total 2002-based Danes 2004-based Non-Danish 2004-based Total 2004-based The zero immigration alternative assumption makes a much bigger difference to the projections of population foreign origin, compared with the fertility variants. With zero net migration, the immigrant population declines to 115,000 from 126,000 and the descendants even with unchanged fertility declined to 131,000 (Danmarks Statistik t. 17, page 21; Table 1 below). Danes themselves were projected to decline Table 1. Denmark: population of foreign origin as % of total population, , on two alternative projections. Net immigration 13,200 Zero migration year TFR of third-world populations TFR of third-world populations falling from 3.25 to constant Source: Danmarks Statistik 1997, tables 15, 17. slightly in absolute numbers (from 4.95 million to 4.93 million, to 87% of the population). The principal projection assumed declining fertility on the part of immigrants, falling by 2020 from a TFR of 3.25 to 2.1 (still higher than the projected Danish level of 1.85; Table 1). 6

7 Since those very detailed 1997 based projections were made, 2002-base projections made on only slightly different assumptions, have taken the projection to In these, net immigration, 10,000 in 2020, declines to 5,300 by Only two immigrant categories are recognised; from more and less developed countries, on the UN definition. As before, the categories into which subsequent generations fell were carefully specified: 46% of the children of mothers born in more developed countries becoming Danes in the next generation, compared with 12% of the children of immigrant mothers from less developed countries Since then further projections have been necessary following a change in immigration policy in That has substantially changed the actual trend, and the assumptions, relating to immigration to Denmark. A more restrictive policy has curtailed a number of avenues of immigration streams; restricting marriage migration to spouses aged 24 and over, and imposing further conditions on the criteria for marriage migration, asylum, and other channels. On the new assumptions, the proportion of the Danish population of foreign origin rises only to 11% in 2020 compared with the 2002-based projection estimate of 14%; in 2040 to 13.8% compared with 18.4% and finally to 14.8% by 2050 (Figure 4). No zero-migration variants are presented for 2002 or 2004 so the total effect ignoring naturalisation cannot be estimated. Figure Denmark - percent population of foreign origin ; three projections. Source: Statistics Denmark based 16-12k migration 2004 based 1997-based zero migration based 13.2k migration Projections of the foreign-origin population in Sweden In line with the practice in Denmark, the Netherlands and elsewhere, the Swedish 2000-based projections of the population of foreign origin define a population of foreign background ( Ultändsk bakgrund ) that comprises persons born outside Sweden together with persons born in Sweden with both parents born abroad. Persons born in Sweden with one parent born in Sweden are defined to be of Swedish background (Statistics Sweden 2003, p 17, 77). That is more conservative that the definitions used in the countries noted above. 7

8 The projections are complex, made separate assumptions about fertility, mortality and migration for six populations: of Swedes, populations from the Norden countries (excluding Sweden, of course), the EU 25 (excluding Norden), and separately for countries with high (excluding the EU25), middle and low Human Development Index. Countries of the Balkans and former Soviet Union, not in the EU25, are classed as middle HDI. That procedure resembles somewhat the western and nonwestern distinctions made in the Dutch projections and the more developed and less developed distinction in those for Denmark. Net immigration to Sweden in 2002 was 31,100, including a net loss of 2,500 people born in Sweden and a net gain of 33,600 others, including 20,100 from countries of middle or low HDI. Those are projected to fluctuate somewhat but generally decline slightly to 31,100 by 2020, of which 20,400 are projected to be from countries of middle or low HDI (Table 2). Immigration and emigration are projected separately, Table 2. Assumptions behind Swedish 2002-based projections. Total Fertility Rate Norden EU25 Other countries according to HDI Sweden excl. Sw excl. N Higher Middle Low All Standardised Mortality Ratio (ages 0-100) males females Net Immigration (thousands) according to birthplace. born in Norden EU25 According to HDI All Sweden excl Sw excl N Higher Middle Low countries Statistics Sweden 2003, table 9, table 19, table 20. 8

9 the former reaching a plateau by 2025, the latter increasing slightly. These unusually detailed time trends of observed data, and projections from 2002, are sufficiently interesting to present in detail in Table 2 below. There is just one principal projection, with no variant projections. The projection of the population by background is a combination of those projected by birthplace and those, born in?sweden, with foreign or Swedish background. The outline is given in Figure 5 below. The population of Swedish background is kept almost constant because the relatively high projected fertility (1.78) is supplemented by constant recruitment from the foreign population, among whom all the third generation, and those with a Swedish born parent, are classified as Swedish background. By 2020, the limit of the 2002-based projections, the number of overseas-born residents is projected to rise to 1.49 million and the children of immigrants to 0.55 million, giving a total population of foreign-background of 21% of the total Swedish population (Figure 5, Statistics Sweden p 18). Further projections based on 2003, which extend to 2050 (Statistics Sweden 2004), enable these proportions to be taken forward on a different basis. The new projections do not subdivide the population by background as above, although they do project the population born abroad expected Figure Projection, Sweden by background. Source: Statistics Sweden Middle and low HDI background Swedish background All All foreign background to rise to 15.3% by 2020 and 18.0 % by A zero-migration projection, however, provided by Mr Å. Nilsson of Statistics Sweden (pers. comm), enables the additional foreign-origin population arising from immigration from 2004 to be determined, by simple subtraction from the principal projection, using the foreign-origin population in 2004 as a basis. That process, inevitably, assumes zero naturalisation among the post-2004 foreign origin population. It also enables the increment of population due to immigration to be calculated. This estimate by subtraction is a more strict definition 9

10 of further foreign-origin increments, in that all the additional population is deemed to be non-swedish, while on the other hand all the future natural increase of the population as it existed in 2004 will be regarded as Swedish even though a (diminishing) proportion will be of foreign background. By interpolation in the previous projection, the proportion of foreign background in 2004 was 1.44 million or Figure Foreign -born and foreign origin population in Sweden, , percent of total population. Source: Statistics Sweden 2003, 2004, Akers pers. comm Foreign birthplace 2000-based projection Foreign background 2000-based projection Foreign background 2003-based projection by subtraction Addition of all population differences between the zero migration and principal projections gives the proportions in Figure 6, indicating that on those assumptions, the foreign-background would reach 32.3% by The projection on this simple assumption from is identical with that on the previous basis, giving some Figure 7. Population projections, Sweden , standard and zero migration. Source: Statistics Sweden Zero-migration projection (millions) Standard projection (millions)

11 confidence in the soundness of the procedure adopted. The assumptions behind the new projections are very similar to those in the previous ones. That is; TFR rising to 1.85, expectation of life for males rising to 80.8 years by 2020 and to 83.6 by 2050; and for females to 84.2 and 86.2 respectively; net immigration to 25,000 by 2020 (the same, for all birthplaces as in the previous projection) and 24,000 by 2050 (Statistics Sweden 2004 t 9.1). The effect on the total population size is also fairly potent. In the principal projection, population rises from 9.01 million in 2003 to million in 2050, compared with 8.64 million for the zero migration assumption an increase of 1.6 million or 19% (Figure 7). Projections of the foreign-origin population in The Netherlands The relatively larger foreign origin population of the Netherlands has been projected to have a substantial impact on the size and the composition of the national population up to 2050 (Alders 2001b, Statistics Netherlands 2001, 2003). Overall, on the based medium variant projection the Dutch population was expected to increase from 15.9 million in 2000 to 17.6 million by 2050; slightly fewer than the 2000-based projection (18.0 millions). A zero-migration variant projected a total of 14.9 million, so according to that 2.7 million additional population is added, directly and indirectly (down from 3.3 million projected in 2000), by post-2003 net immigration. That would increase the population by 18% (Figure 8). Figure 8. Netherlands projections, medium variant and zeromigration (millions). Source: Statistics Netherlands 2003-based projections total-medium variant total-zero migration native - zero migration Two major foreign groups are distinguished: Western and non-western foreigners. The latter include Turks, Moroccans, Surinamese, Antilleans and Arubans, and others. Only two generations are recognised; First Generation immigrants and the Second Generation. As noted earlier, the First Generation is defined as a person born outside the Netherlands with at least one parent born outside the Netherlands. The Second Generation is a person born in the Netherlands with at least one parent born outside it. These are the same definitions as those noted above in the definition of the 11

12 foreign origin population. Subsequent generations are assumed to be Dutch. As with most other projections, this definition therefore progressively understates the demographic impact as time goes. That is similar to the convention used (e.g.) in the Danish projections but not in all others. That should be kept in mind when making comparisons with other projections, made on a different basis. By 2050, 5.7 million people, 32% of the total population, would be of foreign origin on the limited definition noted above: two thirds of non-western origin and one-third of western origin (Figure 9). By this time the western foreign origin population (2.2 million) was projected to have ceased to grow and to be in slight decline. That of non-western origin, having increased from 1.41 million in 2000 (8.9% of total population) to 3.46 m in 2050 (19.6 %), was projected still to be increasing in The Dutch population itself would have declined slowly after 2007; falling eventually from 13.1 million in 2000 to 11.9 million in 2050 (and from 82.5% to 67.8% of the total population.). Figure 9. Netherlands Percent of population foreign, medium variant and zero-migration projections. Source: Statistics Netherlands 2003-based projections Percent population foreign - medium variant percent population foreign - zero migration The latter figures include as Dutch most of the third and subsequent immigrant generations, although the second generation total includes a mixed second/third generation because it includes Dutch-born children with a Dutch parent and foreign parent. The relative increases in the Western foreign populations according to the 2000 based projections are shown in Table 3, and in the non-western populations in Table 11 (de Jong 2001, Alders 2001 b,c). Table 3. Projection of Western foreign population in the Netherlands, 2000 and 2050 (thousands) Year Total First generation Second generation Total Western

13 European Economic Area Other Europe Indonesia Other outside Europe Source: de Jong 2001 page 25 This projection assumes considerable further immigration from Eastern Europe. The Indonesian population is included in the Western category as a highly-assimilated immigrant population, of mixed European-Asian origin. Table 4. Projection of non-western foreign population in the Netherlands, 2000 and 2050 (thousands) Foreign origin Year First Second Total generation generation Total non-western Turkey Morocco Surinam Netherlands Antilles, Aruba Africa Asia Latin America Source: Alders 2001 page 30. On these projections, the Surinamese population increases only to a modest extent, the Turkish population more substantially. Together with the Antillean and Moroccan populations, these currently dominate the Dutch immigrant scene. However the projections assume a substantial increase in new foreign immigrant populations, at present not so numerous in the Netherlands. The biggest increases are among Africans (from 151,000 to 596,000), and Asians; projected to increase over five-fold from 226,000 to 1,311,000. This follows high projected levels of future asylum-claiming (Alders 2001c). Latin American populations also projected to increase five-fold from a modest 51,000 today, to 238,000. This is not the only European projection to assume that much of Europe s ethnic growth may come from relatively new 13

14 populations, especially from asylum claiming, rather than from older established if still imperfectly assimilated ones. Revisions to the Dutch projection based on 2003 yielded a lower projected non- Western foreign origin total than in the tables above (3.46 million rather than 3.89 million) and rather more Western foreign origin population (2.22 million compared with 2.01 million). The latter data were used in Figure 9. In December 2004 the projections were further revised downwards (especially of Africans and Asians) to reflect the decline of asylum claims following the introduction of more restrictive policies. The non-western population for 2050 is now projected to be 2.80 million out of a foreign origin total of 5.03 million (29.7%). Tables and figures will be adjusted accordingly in future revisions of this paper. Projections of the German foreign-origin population The German projections of foreign-origin population, which also run to 2050 (Ulrich 2001), provide three scenarios together with a control scenario of zero immigration from 2000 onwards. Overall, the medium scenario projects a 17% drop in the population of Germany from 82.2 million in 2000 to 68.3 million in The zeromigration control scenario projects a total population of 59.7 million in 2050; 22.5 million or 27% fewer people than in 2000 (Table 5). Distinctions are made between German citizens on the one hand and Turkish, (former) Yugoslav, EU foreign and other foreign citizens. The familiar problems of the citizenship criterion are sidestepped by taking parentage taken into consideration, by calculating the growth in the number of persons of German nationality with two foreign parents, and thus allowing in part for the effect of naturalisation. That allows categories to be projected of second generation persons of foreign origin, even if they are naturalised Germans. However, as with most of the other projections, the third generation is assumed to be German. Within the total the foreign origin population is expected to double from 8.1 million (9.9%) in 2000 to 16.1 million or 23.6%. In the zero-migration scenario, the foreign origin population declines from 8.1 million to 7.7 million. The importance of naturalisation in determining the scale of the foreign contribution to future population is revealed in Table 4. On the simple criterion of citizenship, the foreign population in the medium variant scenario merely increases from 7.4 million to 11 million, not to the 16.1 million on the criterion of foreign origin. The difference of 5.1 million is accounted for by people of foreign origin (that is, with two foreign parents) being naturalised into German citizenship. As well as those 5.1 million naturalised foreign persons, a further 1.9 million of the German citizen population in 2050 are projected to be of aussiedler origin (there were 2.5m in 2000). As in the Dutch projections, the expected increase in the more established foreign populations (e.g. Turkish) is relatively modest, while the greater part of the future growth of foreign population comes from other non-eu foreigners. 14

15 Table 5. Population and citizenship scenarios, Germany (millions) Citizenship Ancestry Medium Zero Medium Zero variant Migration variant Migration German Turkish Yugoslav Other EU states Other foreign All foreign All Source: Ulrich (2001) tables Projections of the Austrian foreign-origin population The projections for Austria (Lebhart and Münz 2003) differ from those above in dealing only with citizens and foreigners, with no broader foreign-origin population. There is no medium variant projection, although one has been issued separately by Statistics Austria (2005). Instead three scenarios are presented. The first is a Compensatory scenario with migration managed to keep the population of working age approximately constant ( 1/KZ in the original paper). In the second restricted scenario ( 2/RZ ), immigration is restricted to a lower ceiling, with a quota on non-eu immigrants, and in the third ( OZ ) all migration is stopped. The TFR overall is assumed to rise to only 1.5 by 2050 and that of the foreign population does not exceed 2.0. The TFR of Austrian citizens was 1.32 in 1995/8 compared with 1.95 for foreigners, and these are assumed to increase modestly to 1.40 and 2.0 by the end of the projection period. The unusually large TFR difference arises because almost 90% of Austria s foreign population are from non-european countries. Those low fertility rates are the fundamental reason why the population fails to increase in any of the scenarios. Instead, as Figure 10 shows, it is almost maintained in the first scenario and falls in the other two. The compensatory scenario is closest to the Statistics Austria main projection, which leaves the population slightly higher in 2050 than in 2000, primarily because of relatively high migration assumptions. With restricted or 15

16 with zero migration, Austrian population would fall by 1.1 and 1.5 million respectively by 2050 in the other two scenarios. Figure 10 9 Austria effect of three migration scenarios on total population (millions). Source: Lebhart and Munz Compensatory migration ('1/KZ') Restricted immigration ('2/RZ') Zero immigration ('OZ') In the three main scenarios, naturalisation is assumed to proceed at the same rather high rate as in Austria in 2000 (about 20,600 in 2000). Under that assumption, even under the highest immigration flows projected, the foreign-citizen population does not rise above 1.03 million or 13.1% of the population by 2050 (Figure 11). However, two of the scenarios are presented without naturalisation. Without naturalisation, not only do immigrants retain their foreign citizenship but so also do their children. So in Figure Austria : the effect of immigration on percent of population with foreign citizenship. Source: Lebhart and Munz (2003). Zero immigration with no naturalisation Zero immigration: percent foreign, with naturalisation. Restricted immigration without naturalisation 20 Compensatory immigration without naturalisation

17 effect the no naturalisation scenarios are projecting what in the UK or US would be described as ethnic minority populations that remain permanently distinct. In the Compensatory variant 1/KZ, immigration is managed at a variable net inflow increasing from the current level of 6000 per year in 2000 to a maximum of 18,000 per year, to keep the population of working age approximately constant. Naturalisation in Austria at the moment is running ahead of net immigration, at about 20,600 per year. Without naturalisation the Compensatory variant would lead to a foreign-origin population of 28% of the total by 2050 compared with 8.9% in 2000 (that was calculated by subtracting the no immigration and no naturalisation ( OZ/OE ) scenario from the former; it is not provided in the published data). In the restrictive migration variant ( 2/RZ ), with net immigration limited to a maximum of 11,000 per year, and with non-eu immigrants subject to a quota, population falls slightly further to 7.4 million while in the zero-migration scenario ( OZ ) population fall to 6.3 million, 19% fewer people than under the compensatory scenario. The proportion foreign falls to 5.1% with naturalisation. In the variant without naturalisation ( OZ/OE ) the proportion foreign rises by 2050, through natural increase only, from 9% to 13.1%. It will be recalled that the foreign population is assumed to retain a higher fertility than average (2.0 compared with 1.5) to the end of the projection. Projection of the ethnic minority population of the United Kingdom In 1979 the UK Office of Population Censuses and Surveys (now Office for National Statistics, ONS) published simple projections of the New Commonwealth ethnic minority populations that is, non-white populations of recent immigrant origin, born in born in the New Commonwealth of Asian, West Indian and African countries (Immigrant Statistics Unit 1979) or with parents born there. That category roughly corresponded to non-western foreign origin populations, or middle and low HDI foreign background populations discussed above, except that the categories, being ethnic, were potentially permanent and did not disappear after the second generation. Those projections extended only to 1991 and were not too inaccurate up to that date, although they under-stated the out-turn of actual events. Political considerations have inhibited any repetition until recently. Following the 2001 Census, however, the Office for National Statistics is developing new projections (Haskey 2002) on a more subtle methodology and a more complex set of ethnic categories, as yet unpublished. Other projections have been made of the New Commonwealth (NC) population on various assumptions, but these assumed the diminution of the then modest levels of immigration and have been eclipsed by events (Brass 1982, Coleman 1995). Some preliminary work on new simple projections of the UK ethnic minority populations is presented here. Without a population register, and without detailed data on the nationality or other attributes of net migrant flows, all estimates of vital rates must be made indirectly. For example, the TFR has been estimated by the own-child method from the Labour Force Surveys (LFS), and the net immigrant flows by adjusting broad-brush estimates of net immigration from the International Passenger Survey by the ethnic distribution of recent arrivals in the UK derived from the LFS. Work is in progress on the mortality estimates; for the time being, it is assumed that they are the same as the national level and trend (Coleman and Smith 2005). Total 17

18 fertility rates averaged from the trends illustrated in Figure 12 below were used as the starting point. Figure 12. TFR (Births per 1000 women) TFR Estimates by Ethnic Group: United Kingdom, Source LFS by own-child method. Data points are 7-year averages White Chinese Black-Caribbean Indian Black-African Pakistani Bangladeshi Source: Labour Force Survey 1979, 1983, ; Quarterly Labour Force Survey (Autumn) Two level of convergence of fertility were used in most groups, according to curves fitted to the rate of change; one converging to the national average, the other to a slightly higher asymptote. In some cases, e.g. the Indian and the Chinese populations, mean TFR has already fallen to the national average or below. To begin with, a projection of the total non-white ethnic minority population is given below, on some simple assumptions. The projections are for England and Wales, not for the UK (most immigrants to the UK settle in England and Wales). Fertility is assumed to fall to an overall level of 1.9, slightly higher than the projected national total (1.75). On that assumption, immigration makes a powerful difference. Assuming a net annual inflow of 70,000, probably an under-estimate, the minority population rises to about 13 million by 2050, increasing in a linear fashion (Figure 13). 50% higher Projection of UK non-white population to 2051 (1000s) at three levels of immigration mortality constant EW 1998, TFR declining from Zero migration High migration 108k/yr Medium migration 70k/yr

19 immigration would take it to almost 16 million; with zero immigration it tails off to stabilise at about 7 million, from 4.5 million in To make for more comparability with other estimates, a projection is also made of white immigrants who are not Irish or British. At present their net immigration is very high estimated at 108k per year. That would generate considerable growth, as Figure 14 shows, in the white non- British population, to over 9 million by mid century if migration remained constant. That seems unlikely; neither asylum inflows from Eastern Europe nor those from the recent EU accession countries are likely to be enduring. But the result is presented below for illustrative purposes. The white British population is assumed to lose population by migration by about 53,000 per year. On those assumptions, the population of England and Wales would rise from 52 million to 62 million (more than GAD projects). The white British and Irish population would decline from 88.7% to 64.3% by mid century, a non-british proportion therefore of 35%, similar to that projected for (e.g.) the Netherlands as the same date. Of that, the non-white proportion would have risen from 8.7% to 19.9% and the white non-british from 2.7% to 15.8%. Figure Sample projection, ethnic composition trends, England and Wales (thousands) Non-white population British and Irish pop White non-british population Total England and Wales pop For the non-european populations, separate projections are being prepared to take into account their different fertility and migration characteristics. An example relating to the Pakistani population, one showing the most vigorous growth, is shown below Here the contrast is made between four scenarios. These combine two alternatives in migration and in fertility. Net immigration is assumed either to end, or to continue at the rate current around 2001, conservatively estimated to be 12,000 per year. Period TFR is assumed either to decline from the current 3 to 1.75 by 2016, or to decline more slowly, to 2.05, by 2051 (Figure 15). Given these alternatives it is clear that the alternative migration assumptions have the more powerful effect than the two very different assumptions about (partial) convergence to lower fertility levels. In that 19

20 respect the general conclusion is similar to that derived from projections in other European countries. Figure British Pakistani population projections (thousands) S 6 No mig, TFR S 8 Mig 12k, TFR S 9 No mig, TFR S 10 Mig 12k, TFR

21 Probabilistic projections of the UK population by major ethnic group. As well as these incomplete projections made in the conventional manner, some experimental projections of the UK population have been made using the LSSI probabilistic basis (the random scenario approach). This has various advantages. First, the probabilistic nature makes explicit the uncertainty which inevitably attends projections and especially projections of ethnic minority populations, the demographic responses of which are even less well understood than those of ordinary populations. The justification for probabilistic projections had been made elsewhere (Lutz, Sanderson and Scherbov 1997, 2001, Keilman, Dinh et al 2002). The probabilistic population projections are based on the multi-state cohort component model of population projections. This model applies a set of given age specific fertility, mortality and migration rates to the age and sex distribution of the starting population along cohort lines. The limits of likely variability must be set a priori for each variable. They cannot be known in advance. Usually, advice is sought from a panel of experts to define the likely maximum and minimum values of the variables at various points into the future, to cover (say) 90 per cent of all future paths of TFR, life expectancy at birth, and net migration (Lutz and Scherbov 1998). Because the resulting distributions of assumed values in previous exercises have turned out to be symmetrical, normal distributions are fitted to those ranges. In this case, symmetrical limits were imposed on future variability by the authors. They were guided, where appropriate, by the limits imposed by the variant projections for the UK provided by the UK Government Actuary s Department (2004). For each of the three variables, a single draw from a standard normal distribution determined its relative position within its range of future values at selected dates. The values at intermediate dates were determined by piece-wise linear interpolation. The results of 1000 of these random simulations were analysed. These projections also incorporate populations of mixed origin (that feature, of itself, does not require the method to be probabilistic, of course). The demographic attention given to mixed unions has not been commensurate with their prevalence in many populations. Most projections of foreign-origin or of ethnic minority populations noted above have ignored mixed unions and populations of mixed origin, generally assuming that any offspring of a mother of a given population sub-group will themselves also be considered to be part of that sub-group. In some societies, where individuals with some foreign background normally identify themselves with the minority group and are so treated by the rest of the population, that will not be unreasonable. However, given the opportunity in censuses and surveys, large numbers of people in the UK and the US actively identify themselves as having multiple ethnic origins, from the parental or even earlier generations (e.g. Goldstein and Morning 2002). However the rules governing the attribution of foreign origin to the secondgeneration in some of the projections (e.g. those of Denmark) incorporate a set of coefficients to determine the status of children of different combinations of parental origin. The projections of the US ethnic groups by Smith and Edmonston (1997) explicitly incorporate both mixed unions and change of ethnicity by individuals. The latter is not done here. The projections here are made for a maximum of 100 years, although the results for 2025 and 2050 are clearly the most important. Although it is obvious that much can 21

22 and must change over the longer period, there is at least some merit in showing the long-term consequences of current trends and their (at least plausible) future evolution, and the range of possible likely outcomes. The policy of the UK GAD is to give most attention to its projections to 2051 although detailed projections are made up to 2071 (GAD 2004) and some have been published to the end of the century (Shaw 2001). The UN now routinely produces longer-term projection, and even to 2300 (UN 2004) although not without some scepticism as to their value. Assumptions behind the projections The probabilistic projections presented here are based approximately upon the demographic experience of the ethnic minority populations of the United Kingdom, but using a very simplified scheme of ethnic classification. That was done in order to gain experience of this type of simulation before attempting a more complex operation. The ethnic groups used in the collection of official statistics in the UK have changed in a number of ways since the practice first began in the 1971 census and in the 1979 Labour Force Survey (see Bulmer 1996). The scheme used in recent Labour Force Surveys and in the 2001 census is given below (Table 6) together with the drastic amalgamations made necessary by the requirements of the current probabilistic projection program. These amalgamations undoubtedly do serious violence to reality. Black and Asian are the two major clusters of ethnic minority groups recognised in the UK ethnic taxonomy. However, the demographic characteristics of the two main Black groups (Caribbean and African) are rather different; the latter having higher fertility and higher immigration, especially from asylum-claiming. But they do share relatively high levels of inter-ethnic unions. Likewise the Asian group brigades together Indians, of relatively high status and low birth rate, with Pakistanis and Bangladeshis, of lower average status and higher birth rate. All share, however, in having high migration rates driven substantially by arranged marriages as well as by Table 6. United Kingdom ethnic population distribution 2001 (thousands) ONS category E&W Scotland Northern United United Final groups Ireland Kingdom Kingdom Bangladeshi Indian ( Asian ) Pakistani Chinese Other Asian Other African Caribbean ( Black ) Other Black White/Caribbean White/African ( Mixed ) White/Asian Other Mixed Source: ONS 2001 Census 22