The migration ^ immigration link in Canada's gateway cities: a comparative study of Toronto, Montreal, and Vancouver

Transcription

1 Environment and Planning A 2006, volume 38, pages 1505 ^ 1525 DOI: /a37246 The migration ^ immigration link in Canada's gateway cities: a comparative study of Toronto, Montreal, and Vancouver Feng Hou Business and Labour Market Analysis Division, Statistics Canada, Ottawa, ON K1A 0T6, Canada; feng.hou@statcan.ca Larry S Bourne Department of Geography and Program in Planning, University of Toronto, 100 St George Street, Toronto, ON M5S 3G3, Canada; bourne@geog.utoronto.ca Received 21 July 2004; in revised form 17 February 2005 Abstract. In this paper we explore the links between internal migration and international immigration in Canada's three largest metropolitan areas. In particular, we use a place-specific approach to test the displacement hypothesis that the migration behaviour of the less-well-educated native-born population is sensitive to the inflows of immigrants. Based on analyses of microdata from five consecutive censuses covering the period from 1981 to 2001, we find that the migration ^ immigration relationships are complex, often subtle, and inconsistent across the three cities. Growth in the immigrant population is correlated with an increased out-migration rate among the less-well-educated native-born population, but only in Toronto and Vancouver. This correlation, however, is not independent of changes in housing prices. We also find no consistent support for an alternative hypothesis that economic restructuring accounts for the net out-migration from immigrant gateway cities. 1 Introduction The 1990s has witnessed significant increases in immigration in almost all Western countries. This immigrant inflow has in many instances become the largest component of national population growth and a major factor in social and cultural change. The destinations of these immigrants are also highly concentrated geographically, adding to the uneven patterns of urban growth and change. The impacts of the concentration of new immigrants in a few large metropolitan areas have yet to be evaluated, but they are likely to be considerable. These effects can also be either mitigated or further augmented by the internal migration of the nativeborn and earlier immigrants to and from these major immigrant-gateway centers. The volume of net internal-migration flows not only impinges on the uneven geographical distribution of population growth, but selective out-migration and in-migration may also influence spatial differences in the sociodemographic and ethnic composition of the country's population. Recently, selective migration seems to be a common pattern in major immigrantgateway cities in the United States and Australia, and in some European countries: the domestic-born, the less-well-educated in particular, are moving away and leaving behind a markedly more heterogeneous social environment in metropolitan areas that stands in contrast to the rest of the nation (Champion, 1994; Frey, 2002a; Ley, 2003; The Economist 2003). Indeed, Frey (2002a) suggests that the differences in immigration and internal-migration flows in the United States are transforming what was a once `single melting pot' nation into three distinct Americas: a suburb-like `New Sunbelt' region, a socially diverse and economically vibrant `Melting Pot' region, and an aging, whiter, and slow-growing `Heartland' region. Similar concerns about increasing geographic divisions in levels of social diversity have also been raised in other major immigrant-receiving countries.

2 1506 F Hou, L S Bourne The literature, however, is not clear about to what extent and how the net out-migration of the domestic-born from major-gateway cities is related to immigrant growth. A particular concern is whether the concentration of recent immigrants leads to selective internal migration from and into major-gateway cities. Canada offers an interesting case study of the possible migration ^ immigration link. 1.1 The Canadian context Between 1991 and 2000, Canada admitted 2.2 million immigrants, the highest intake of newcomers in any decade in the past one hundred years, and one of the highest rates in the Western world. During the 1990s, the major source regions of immigrants shifted further from Europe to Asia and other Third World countries. These new sources of immigrants have led to the rapid increase in the level of sociocultural diversity. A salient indication of such change is a three-fold increase in the country's visible minority (non- Caucasian in race or nonwhite in skin colour, non-aboriginal) population between 1981 and The economic and sociodemographic impacts of changes in the size and pattern of immigration flows in Canada have been felt most strongly in three census metropolitan areas (CMAs): the `gateway' centers of Toronto, Vancouver, and Montreal. Immigrants who arrived in Canada in the 1990s were much more likely to settle in these areas than those who came in earlier decades. About 58% of immigrants who came to Canada in the previous ten years lived in the three CMAs in By 2001, this proportion had increased to 74% (Statistics Canada, 2003). The Toronto CMA has a higher proportion of immigrants in its population (44% in 2001) than other major-gateway centers in the United States (40% in Miami, 30% in Los Angeles, and 23% in New York in 2000), Australia (31% in Sydney in 2001), and Europe (15% in Paris in 1999 and 14% in London). Vancouver holds the second largest immigrant population, and the proportion of immigrants (40%) is also among the highest worldwide. Montreal's immigrant population ranks third in size in the country, accounting for 18% of the city's population. If the migration behaviours of the native-born and established immigrants are responsive to the concentration of recent immigrants, we should be more likely to observe a migration ^ immigration link for the major Canadian cities, particularly in Toronto and Vancouver, than in most other gateway cities in the developed world. 1.2 Objectives In this study we examine trends in the internal migration of the Canadian-born and long-term immigrants into and out of Canada's three largest CMAs. Specifically, we ask the following questions: (1) Has the population of Canadian-born and long-term immigrants become more likely to move away from and less likely to move into the three major immigrant-gateway cities in the last two decades, given the increased inflows of new immigrants? (2) Do these trends vary with education levels and visible minority status? (3) To what extent is the level of inflows of recent immigrants into the three metropolitan areas associated with the parallel trends in out-migration and in-migration? We conclude with a discussion of the implications of these trends and relationships. 2 The migration and immigration linköa literature review In the extensive literature on immigration and internal migration, there is a growing body of studies focused on the internal migration patterns into and out of major immigrant-gateway cities and regions. Most of these are based on US research. The most vocal and probably the most controversial in this field are a series of publications by Frey (1995a; 1995b; 2002a; Frey et al, 1996). He suggests that the concentration of

3 The migration ^ immigration link in Canada's gateway cities 1507 new immigrants and the selective out-migration of the native-born are sharpening the geographic differences in racial and sociodemographic characteristics within the United States. Frey (2002a) argues that much of this selective migration of the nativeborn was a response to the high cost of living and desire for a suburban lifestyle that can no longer be met in the original suburbs in major metropolitan areas. Although for the most part, the out-migration of the native-born population from immigrant-gateway areas may be lifestyle driven, Frey (2002b) suggests that the lowincome, less-educated native-born population are most sensitive to the inflows of immigrants as they are most likely to be in direct competition with new immigrants for low-skill, low-paying jobs. The perception of the higher social costs associated with increasing numbers of immigrants, including higher crime rates, reduced services, or increased taxes, combined with racial and ethnic prejudices, may also contribute to the selectivity of the outward movement of the native-born population (Frey, 2002b). Frey's displacement hypothesis is based on the observation that there is persistent out-migration of low-skilled whites during the up-turns and down-turns in local economies and out-migrants are prone to select nonmetropolitan destinations (Frey, 1995a; 1995b). In comparison, the internal migration of those with a college education, who may be less affected by competition with the large numbers of new immigrants, is more closely tied to the general economic condition of a region. Thus, although immigrantgateway areas in the United States sustained negligible or negative net internal migration over the last half of the 1980s and early 1990s, they still gained college graduates, upperincome households, and professionals from other parts of the country in periods of relatively strong economic growth. Frey's displacement hypothesis has been contested by other researchers (Ellis and Wright, 1998; Harrison, 2002). The disagreement centers around whether immigrant concentration leads to selective internal migration, which remains ``one of the most important unresolved questions about US immigration'' (Card, 2001, page 36). Some researchers have proposed alternative explanations for the net out-migration of the native-born from major immigrant-gateway centers. Walker et al (1992) interpret their observation of a positive association between immigration to a metropolitan area and a net loss of unskilled blue-collar workers as a result of economic restructuring in those global cities that also serve as immigrant gateways. They suggest that a shift from manufacturing to services and high-technology production leads to the out-migration of domestic blue-collar workers and the inflow of low-skilled immigrants. This explanation is further emphasized in a later study by Wright et al (1997). A third potential explanation is related to housing prices. From his study on internal migration flows in Toronto and Sydney, Ley (2003) found that housing-market cycles may have much stronger effects than labour-market conditions on the internalmigration flows of the domestic-born. Bourne (1999; 2000) also suggests that some of Toronto's out-movers are actually equity-migrants, taking advantage of the equity built up in their housing through urban growth and the demands posed by high levels of new immigrant inflows. In studies in the United States in which the connection between immigration concentration and internal migration have been examined, analysis has been primarily based on aggregate data at the state or metropolitan-area level. From a correlation analysis of the relationship between immigration and native-born migration rates among 272 metropolitan areas from the 1980 census, Filer (1992) found a strong negative effect of immigrant arrivals on the attractiveness of the local labour market to native-born workers. This effect was concentrated among less-skilled and lesseducated white workers. Borjas et al (1997) also found a negative correlation between the annual growth rate of the native-born and the growth rate of immigrants among

4 1508 F Hou, L S Bourne US states during the period from 1970 to 1990 when the population growth pattern of a state, before large changes in the size of the immigrant population, was taken into account. In contrast, other studies found no support for the immigration ^ migration connection. White and Imai (1994) showed that changes in in-migration and out-migration rates of the native-born across US metropolitan areas between the late 1970s and late 1980s were not significantly affected by the initial level of immigration and growth of the number of immigrants. In order to overcome the limitation of other studies that treat new immigrants as a homogeneous group, Card (2001) and Card and DiNardo (2000) examined the relationship between population movements of native-born workers in different skill groups with the inflow rates of less-skilled immigrant labour. Using data from the 1970, 1980, and 1990 censuses, they found that the intercity migration flows of native-born and established immigrant workers (who previously lived in the city) were insensitive to immigrant inflows in the same skill groups. Research results based on microdata are also ambiguous. From their analysis of the 1990 Census, Frey et al (1996) found that persons classified as living in poverty were more likely to move away from states with high levels of immigration. This impact was particularly strong for the white population. With pooled cross-sectional data from the 1981, 1984, 1987, and 1990 Current Population Survey, White and Liang (1998) modelled the probability of interstate in-migration and out-migration as a function of state and individual characteristics. They found that non-hispanic white workers were more likely to move out from, and less likely to move into states with high levels of recent immigrants. However, Kritz and Gurak (2001) found that both native-born and immigrant men were less likely to leave states with high immigration than they were to leave other states in the late 1980s. Although native-born, non-hispanic white men tended to leave states with large numbers of Latin American and Caribbean immigrants, this relation became insignificant after controlling for differences in state economic situations. In an attempt to account for the conflicting results of previous studies, Wright et al (1997) conducted extensive testing of various model specifications. They found that the measurement of the size of the immigrant population, the sample chosen, and the formal educational attainment of the native-born workers, all affect the estimation of the impact of immigrants on native-born migration. Most importantly, they suggested that the few key immigrant-gateway centers, primarily New York and Los Angeles, stand out from other metropolitan areas in the immigration ^ migration linkage. Similarly, Butcher and Card (1991) found that native-born in-migration flows were positively correlated with inflows of recent immigrants to most major metropolitan areas in the first half of the 1980s. But this was not the case for New York, Los Angeles, and Miamiöthe three most immigrant-intensive cities. Kritz and Gurak (2001) also observed that the association between the out-migration of non-hispanic white men and recent immigrants from Latin America and the Caribbean resulted mostly from the outlier effects of California and Florida. These results indicate that the migration response of the native-born to immigration is likely a ``large metropolitan area phenomenon'' (Wright et al, 1997). As the linkage between immigrants and internal migrants may be limited to major-gateway centers, Wright et al (1997) suggest that a place-specific analysis would generate more meaningful interpretations of immigrant ^ native-born interactions in labour markets. Such analyses should focus on the socioeconomic characteristics of the in-migrants to major immigrant-gateway centers and how they fit together with immigrants and other native-born residents within the local labour market. It is also important to examine

5 The migration ^ immigration link in Canada's gateway cities 1509 the populations who tend to move out from these immigrant magnets, where they move to and how they fare after the move. No Canadian studies have adopted this place-specific approach to examining internal migration in major immigrant-gateway cities. In some studies it has been investigated whether the internal migration of immigrants leads to spatial dispersal or further concentration, how existing concentration affects immigrants' migration behaviour, and how immigrants are different from the native-born in the determinants of migration (Edmonston, 2002; Moore and Rosenberg, 1995; Newbold, 1996). In none of these studies, however, has special attention been devoted to the internal migration of long-term immigrants to and from major immigrant-gateway centers. Furthermore, in no studies has there been an examination of how the Canadian-born and immigrants differ in terms of their internal migration into and out of these gateways. In this paper, we take the approach of place-specific analyses suggested by Wright et al (1997) and focus on internal migration into and out of Canada's three largest CMAs. We extend US studies on the immigration ^ migration linkage by filling some of the gaps in the indices used and by adding insights from the Canadian literature and experience. This also has advantages in terms of data and methods over previous, US studies. Most of the US studies were based on the decennial census, which contains migration data only for the second half of each decade. In comparison, Canada conducts a census every five years. By pooling five censuses spanning the period from 1981 to 2001, we can examine the trends in internal-migration flows in the 1980s and 1990s by education level and visible-minority status. With access to the census 20%-sample microdata files, we can further examine how an increase in the immigrant population, variations in housing prices, and the rate of economic restructuring affect the trends in internal migration and how these effects differ among the three metropolitan areas. Whereas in earlier studies only net migration or out-migration from immigrant gateways have tended to be examined, in this paper we demonstrate how in-migration and out-migration of the native-born and long-term immigrants are associated differently with growth in the immigrant population. In particular, we examine the migration ^ immigration link in progressive steps. First, we examine the trends in net-migration flows of the Canadian-born and longterm immigrants (those who have lived in Canada more than five years) by education and visible minority status. If immigrants have a displacement effect, we should observe a net out-migration from gateway cities among the less well-educated and among the white population. Second, we examines changes in out-migration and in-migration rates. Given the increased level of immigration during the 1990s, the displacement hypothesis would predict an increase in the out-migration rate and a decrease in the in-migration rate, particularly among the less well-educated and among the white population. Third, we examine the statistical association between trends in migration rates on the one hand, and immigration level, housing-price variation, and economic restructuring on the other. The displacement hypothesis would predict that immigration growth is statistically associated with increased out-migration from and decreased in-migration into gateway centers, notably among the less well-educated and among the white population. And, this association should be independent of changes in economic restructuring and housing prices. Specifically, we anticipate that the effect of the three city-level factors will vary across cities. During the 1990s, all three cities had higher rates of immigration growth, higher average housing prices, and slower economic restructuring than in the 1980s (table 1, over). However, the changes in these factors are different across the three cities in several ways. The rate of immigration growth was the highest in Vancouver, followed by Toronto. We would therefore expect to find that the effect immigration growth had

6 1510 F Hou, L S Bourne Table 1. Immigration growth, housing-price variation, and economic restructuring in Toronto, Montreal, and Vancouver. Toronto Montreal Vancouver Immigrant growth (%) a 1976 ± ± ± ± ± Housing-price variation b 1976 ± ± ± ± ± Economic restructuring (%) c 1976 ± ± ± ± ± a Ratio of recent immigrants (arrived within five years) to other immigrants. b Five-year average of new housing-price index, 1992 ˆ 100. c Percentage increase in the proportion of managerial and professional occupations in the labour force. on changes in internal migration between the 1980s and 1990s was largest in Vancouver and smallest in Montreal. During the 1980s and 1990s, housing prices, after taking inflation into account, peaked in the late 1980s in Toronto, in the early 1990s in Vancouver, and in the late 1990s in Montreal. On average, Montreal had much lower average housing prices, but the largest increase in prices between the 1980s and 1990s, followed by Toronto. We would expect that the effect of the increases in housing prices on changes in internal migration is most likely to be manifest in Montreal and least likely to be manifest in Vancouver. In terms of economic restructuring, the largest five-year increase in the share of quaternary (managerial and professional) occupations in the labour force occurred in the 1976 ^ 81 period in Vancouver, in the 1981 ^ 86 period in Montreal, and in the 1986 ^ 91 period in Toronto. On average, Toronto had the highest rate of economic restructuring in the 1980s, followed by Montreal. We would therefore expect to observe that economic restructuring had the largest effect on changes in internal migration in Toronto and the smallest effect in Vancouver. 3 Data and methods We used microdata from the 20%-sample files from five censuses from 1981 to Internal-migration status is drawn from the census question on place of residence five years before the census date and pertains to movement over the five years prior to each census. Out-migrants from a CMA were those migrants who resided in the CMA five years prior to the census but did not live in the same CMA at the census date, whereas in-migrants included those migrants who resided outside of the CMA five years prior to the census.

7 The migration ^ immigration link in Canada's gateway cities 1511 We focus on the working-age population between ages 25 years and 64 years. For the Canadian-born and long-term immigrants (those who had lived in the country more than five years) separately, the following analyses first describe net-migration flows for each of the three largest CMAs over the period from 1976 to Then, multivariate-logistic-regression techniques are used to estimate changes in out-migration and in-migration rates. In the models for out-migration, the dependent variable is coded as out-migrants ˆ 1, and those who lived in the CMA five years prior to the census and still lived there at the current census ˆ 0. Similarly, in the models for in-migration, the dependent variable is coded as in-migrants to a CMA ˆ 1, and those who lived outside of the CMA over the five-year census period ˆ 0. For both out-migration and in-migration, the following four models are estimated separately for the Canadian-born population and long-term immigrants in each of the three CMAs. The first logistic model is used to estimate changes in migration rates between the 1990s and earlier decades, assuming the population composition remains the same over the entire study period. The independent variables include sex, age, family structure, home language, education, visible minority status, and a dummy variable (nineties) comparing the 1990s with earlier decades. The coding of these variables is presented in the appendix. We further include the interaction terms between the variable `nineties' and the two focal-independent variablesöeducation and visible minority status. These interactions allow us to examine whether internal migration flows were more selective by education and visible minority status during the 1990s than in the 1980s. We do not include individuals' labour-market characteristics, such as employment status, occupation, and income in the models as they were measured after migration and thus could be the outcomes rather than determinants of that migration. With models 2 to 4, we examine the statistical associations of immigrant inflows, housing-market conditions, and economic restructuring in a CMA with changes in out-migration or in-migration rates. In model 2, we add the variable representing immigration growth to model 1. This variable is measured by the percentage of recent immigrants (in other words, those living in Canada five years or less) in the total immigrant population in a CMA in each census. In model 3, we add the variable representing changes in housing-market conditions to model 1. This variable is measured by the five-year average of the new housing price index from Statistics Canada. In model 4, we add the variable representing the rate of economic restructuring to model 1. This variable is measured by the increase in the percentage of the labour force in managerial and professional occupations over the five-year census period in a metropolitan area. Comparing each of our models 2, 3, and 4 with model 1, we can examine how the three city-level variables affect the changes in the in-migration and out-migration rates between the 1980s and 1990s. The detailed specifications and sample size of these models are presented in the appendix. Due to space limitations, we do not present the model coefficients (available from the authors upon request). Rather, we will focus on estimated migration rates derived from these models. One limitation of our analyses is that we cannot put the three city-level variables in the same model. This is because there are only five observations for each of the three city-level variables derived from five consecutive censuses. Estimating the coefficients for the three variables plus the nineties variable in the same model would rely therefore on only one degree of freedom. Another possible complication is that immigration growth, housing prices, and economic restructuring may not be exogenous. Thus, we need to be cautious in interpreting the associations between these city-level variables and individual migration behaviours.

8 1512 F Hou, L S Bourne 4 Empirical results 4.1 Internal migration of the Canadian-born Overall, all three cities experienced a net out-migration of the Canadian-born in the 1990s and, confirming our initial hypothesis, the net loss of migrants was concentrated among the less well-educated and among the white population. In Toronto and Vancouver out-migration rates increased among those with less than high-school education between the 1980s and 1990s. Furthermore, in-migration rates decreased over the period in all three cities. These results also seem to be consistent with the expectation of the displacement hypothesis. However, the general decline in the out-migration rate in Toronto and Montreal in the 1990s is not consistent with this expectation. In this section we examine these trends in detail. The Toronto CMA, on balance, has lost Canadian-born population through domestic-migration exchanges since the second half of the 1980s (table 2). Against this general trend, Toronto gained university-educated migrants in both the 1980s and 1990s. Meanwhile, Toronto lost Canadian-born migrants with less than university education in both the 1980s and 1990s. This may reflect a tendency, noted in several US studies, for the displacement of lower-skilled workers through competition in the labour market from immigrants. The net loss in internal migrants during the 1990s was also concentrated among the white population. Table 2. Net internal migrants among the Canadian-born by education and visible minority status, population aged 25 years to 64 years [data source: 1981 ^ 2001 Canadian Census 20%-sample micro files (Statistics Canada, 1982; 1987; 1992; 1999; 2002)] ± ± ± ± ± 2001 Toronto Total education <high school high-school graduation some postsecondary university degree visible minority white Montreal Total education <high school high-school graduation some postsecondary university degree visible minority white Vancouver Total education <high school high-school graduation some postsecondary university degree visible minority white Note: all numbers were rounded to the nearest 10.

9 The migration ^ immigration link in Canada's gateway cities 1513 Similarly, Vancouver gained Canadian-born migrants with a university education and visible-minority migrants in both the 1980s and 1990s (table 2). This city, however, started to experience a net loss of Canadian-born migrants in the early 1990s. The net loss in the 1990s was concentrated among people with less than university education and among the white population. In comparison, Montreal never had a significant gain in migrants with university education, although this city also tended to lose working-age population with less than high-school education. Whereas the other two cities more or less gained visible minorities, Montreal lost them. As net-migration flows do not reveal the migration propensity of population groups and the different contributions resulting from in-migration and out-migration, we now turn to an examination of migration rates. The first two columns corresponding to model 1 in table 3 (over) present average five-year out-migration and in-migration rates, estimated from our first logistic regression model, for the Canadian-born in the three metropolitan areas. In terms of changes over time, Canadian-born migrants were much less likely to move into the three CMAs during the 1990s than in the 1980s (model 1, bottom half of table 3). This decline in the in-migration rate was widespread across education levels and racial groups. Only in Toronto did the in-migration rate among visible minorities increase slightly between these two periods. In comparison, the trend in out-migration was much less uniform across CMAs and population groups. The overall out-migration rate decreased between the 1980s and 1990s in Toronto and Montreal, but increased in Vancouver. The only change common to the three CMAs was the decrease in out-migration rates among the university-educated. The outmigration rate of those with less than high-school education, in contrast, increased in both Toronto and Vancouver. The decline in out-migration was smaller among the white population than among visible minorities in Toronto. In Montreal the out-migration rate increased among visible minorities, but decreased among the white population. The opposite holds in Vancouver. To what extent are the observed trends in out-migration and in-migration among the Canadian-born associated with the three city-level variables: immigration growth, housing price variation, and economic restructuring? The effects of the three city-level factors are summarized in table 3 under the headings of model 2, model 3, and model 4. Overall, none of these factors can consistently explain the changes in the internalmigration patterns among the Canadian-born between the 1980s and 1990s in the three cities. These factors tended to have stronger associations with changes in out-migration rates than with changes in in-migration, and the directions of these associations were not consistent across cities. We first discuss the associations of the three city-level factors with changes in outmigration rates among the Canadian-born. In Toronto and Vancouver immigration growth was clearly associated with the rise in out-migration rates among those with less than high-school education. Before controlling for immigration growth, the outmigration rate among those with less than high-school education rose between the 1980s and 1990s [as in model 1, table 3(a)]. After the control, the rate fell [model 2, table 3(a)]. In Montreal, controlling for immigration growth has a much smaller effect, but in the same direction. In all three cities, the change in the out-migration rate among those with university education was the least affected by controlling for immigration growth. These findings suggest that the large increase in immigrants in the 1990s was primarily correlated with the rise in the out-migration rate among the less-well-educated Canadian-born population. The level of immigration growth had similar associations with changes in out-migration rates for visible minorities and the white population in all three cities.

10 Table 3. Estimated average five-year (a) out-migration and (b) in-migration rates for the Canadian-born population aged 25 years to 64 years, with and without controlling for immigration growth, housing-price variation, and economic restructuring. Model 1: no controls for Model 2: controlling for Model 3: controlling for Model 4: controlling for city-level factors immigration growth housing-price variation economic restructuring 1976 ± ± ± ± ± ± ± ± 2001 (a) Estimated out-migration rate (%) Toronto Total education <high school high-school graduation some postsecondary university degree visible minority white Montreal Total education * <high school * * high-school graduation * some postsecondary * university degree * visible minority * white * Vancouver Total education <high school * high-school graduation * some postsecondary * university degree visible minority white F Hou, L S Bourne

11 Table 3 (continued). Model 1: no controls Model 2: controlling for Model 3: controlling for Model 4: controlling for for city-level factors immigration growth housing-price variation economic restructuring 1976 ± ± ± ± ± ± ± ± 2001 (b) Estimated in-migration rate (%) Toronto Total education <high school high-school graduation some postsecondary university degree visible minority * * white Montreal Total education <high school * high-school graduation * some postsecondary university degree visible minority white Vancouver Total education <high school high-school graduation some postsecondary university degree visible minority * white Note: The estimated average five-year migration rates were calculated by holding constant age, family structure, education, home language, and visible minority status with the average of the Canadian born who were included in the model. * The difference in the rates for 1976 ± 1991 and 1991 ± 2001 was not significant at p < 0:05. The migration ^ immigration link in Canada's gateway cities 1515

12 1516 F Hou, L S Bourne Similar to the effect of immigration growth, higher housing prices in the 1990s tended to increase out-migration rates among the less-well-educated Canadian-born in both Toronto and Vancouver [model 3, table 3(a)]. Some of these may have been equity migrants who moved out in order to secure less-expensive housing elsewhere. Controlling for housing-price variations had smaller effects than controlling for immigration growth in both cities. In comparison, in Montreal, higher housing prices in the 1990s tended to reduce the out-migration rate among those with university degrees. Housing-price variations had similar associations with changes in out-migration rates for visible minorities and the white population in Toronto and Vancouver. In comparison, higher housing prices during the 1990s in Montreal were associated with an increased out-migration rate among visible minorities, but decreased out-migration rates among the white population. The effects of economic restructuring on out-migration rates also varied considerably across cities. When economic restructuring was controlled for, Toronto's out-migration rate among the Canadian-born increased rather than decreased between the 1980s and 1990s [model 1 versus model 4, table 3(a)]. This effect suggests that the slowing down of occupational restructuring during the 1990s was associatedölikely as both cause and effectöwith the decline in the out-migration rate. Controlling for economic restructuring had a particularly strong impact on the out-migration rate among the less well-educated. Without the slowdown in economic restructuring in the 1990s, those with less than university education would have had much higher out-migration rates. This finding is consistent with the argument that economic restructuring increases the out-migration of blue-collar workers. The effect of controlling for economic restructuring was primarily evident among the white population. The association between economic restructuring and the out-migration rate among the Canadian-born was in the opposite direction in Montreal. Without the slowdown in economic restructuring, Montreal would have had even fewer migrants moving out (model 4 versus model 1). The effect of economic restructuring was similar across different education levels, but was manifested mostly among the white population. Different from the situations in Toronto and Montreal, the slowdown in economic restructuring had no clear association with changes in internal-migration rates among the Canadian-born in Vancouver. This holds for all education levels and racial groups. Clearly, the flow of populations into Vancouver is influenced as much by noneconomic factors, such as amenity migration and retirement, as by economic incentives. In terms of changes in in-migration rates, the growth of the immigrant population in the 1990s tended to reduce the in-migration rate for Toronto and Montreal for all educational levels and mostly among the white population [table 3(b)]. In comparison, a growth in immigration tended to increase in-migration in Vancouver. The effect of housing-price variations was similar to that of immigration growth. The effect of controlling for economic restructuring was small among the Canadian-born in Vancouver, but noticeable in Toronto and Montreal. In Toronto the slowdown in economic restructuring in the 1990s tended to increase the in-migration rate among the less well-educated and among the white population, but to reduce the in-migration rate among visible minorities. In Montreal, the slowdown in economic restructuring in the 1990s tended to reduce the in-migration rates for all educational levels and racial groups. 4.2 Internal migration of long-term immigrants Among long-term immigrants, the general decline in in-migration rates in all three cities and the increase in out-migration rates among those with less than high-school education in Toronto and Vancouver, seem to be in line with the expectation of the displacement hypothesis. However, the general decline in out-migration rates in all

13 The migration ^ immigration link in Canada's gateway cities 1517 three cities and the net gain of long-term immigrants in Vancouver are not consistent with that hypothesis. Toronto gained immigrants with a university degree in both the 1980s and 1990s (table 4). Toronto's net loss of immigrant migrants in the 1990s was concentrated among those with less than university education and among the white population. Vancouver gained immigrants in both the 1980s and 1990s, and the gain was concentrated among those with university education. In contrast, Montreal lost long-term immigrants in both the 1980s and 1990s. The net loss was heavily concentrated among those with university or some postsecondary education and among visible minorities. As pointed out earlier, human capital is highly mobile, and in this case Montreal appears to be losing in the competition for talented immigrants. Long-term immigrants tended to have much lower out-migration rates but higher in-migration rates than the Canadian-born for both Toronto and Vancouver [model 1 in table 5 (over) versus model 1 in table 3]. This suggests that internal migration increases the concentration of immigrants relative to the Canadian-born in these two metropolitan areas. In comparison, immigrants on average were much less likely to move into Montreal than were the Canadian-born. Although immigrants also had lower out-migration rates than the Canadian-born in Montreal, the relative gap was much larger in in-migration rates. Table 4. Net migrants among long-term immigrants by education and visible minority status, population aged 25 years to 64 years [data source: 1981 ^ 2001 Canadian Census 20%-sample micro files (Statistics Canada, 1982; 1987; 1992; 1999; 2002)] ± ± ± ± ± 2001 Toronto Total education <high school high-school graduation some postsecondary university degree visible minority white Montreal Total education <high school high-school graduation some postsecondary university degree visible minority white Vancouver Total education <high school high-school graduation some postsecondary university degree visible minority white Note: all numbers were rounded to the nearest 10.

14 Table 5. Estimated average five-year (a) out-migration and (b) in-migration rates for long-term immigrants aged 25 years to 64 years, with and without controlling for immigration growth, housing-price variation, and economic restructuring. Model 1: no controls for Model 2: controlling for Model 3: controlling for Model 4: controlling for city-level factors immigration growth housing-price variation economic restructuring 1976 ± ± ± ± ± ± ± ± 2001 (a) Estimated out-migration rate (%) Toronto Total education <high school * high-school graduation some postsecondary university degree visible minority white Montreal Total education <high school * * high-school graduation some postsecondary university degree visible minority * * white Vancouver Total education * <high school high-school graduation some postsecondary * university degree visible minority * white * F Hou, L S Bourne

15 Table 5 (continued). Model 1: no controls Model 2: controlling for Model 3: controlling for Model 4: controlling for for city-level factors immigration growth housing-price variation economic restructuring 1976 ± ± ± ± ± ± ± ± 2001 (b) Estimated in-migration rate (%) Toronto Total education <high school high-school graduation some postsecondary university degree visible minority white Montreal Total education <high school high-school graduation some postsecondary university degree visible minority white Vancouver Total education <high school high-school graduation some postsecondary university degree visible minority * white Note: The estimated average five-year migration rates were calculated by holding constant age, family structure, education, home language, and visible minority status with the average of long-term immigrants who were included in the model. * The difference in the rates for 1976 ± 1991 and 1991 ± 2001 was not significant at p < 0:05. The migration ^ immigration link in Canada's gateway cities 1519

16 1520 F Hou, L S Bourne These differences likely reflect the role of cultural and linguistic differences between Montreal and the other gateway cities, and the differences in recent immigration flows. Estimates of migration rates by home language show that Montreal's net loss of immigrants was almost exclusively among non-francophones. Non-Francophone immigrants had an out-migration rate that was over three times higher than that of Francophones, whereas Francophone immigrants had an in-migration rate over twenty times higher than non-francophones. Due to Quebec's immigration agreement with the federal government, Quebec's immigrant selection criteria are less stringent than the rest of the country. It is possible that some immigrants may use Quebec as a springboard to get into the rest of Canada (DeVoretz et al, 2003). However, our data show that only among immigrants with a university degree did Montreal have a significantly higher out-migration rate than the other two cities. As university-educated immigrants may not benefit that much from Quebec's selection criteria, their high mobility may also reflect the effects of language and economic factors. In terms of change over time, both out-migration and in-migration rates among long-term immigrants decreased during the 1990s in all three cities. However, this decline was not uniform across cities or population characteristics. In particular, the out-migration rate among immigrants with less than high-school education either remained stable (in Toronto and Montreal) or increased (in Vancouver). Changes in both out-migration and in-migration rates among long-term immigrants were weakly and inconsistently associated with immigration growth, housing-price variations, and economic restructuring. Immigration growth tended to increase outmigration and reduce in-migration in the 1990s in Toronto, mostly among white immigrants (model 2 versus model 1, table 5). In Montreal immigration growth tended to reduce out-migration in the 1990s but had a much smaller effect on in-migration. In Vancouver immigration growth in the 1990s was associated with increased in-migration but had little effect on out-migration rates. Housing-price variations were not significantly associated with changes in outmigration and in-migration among long-term immigrants in Toronto (model 3 versus model 1, table 5). Higher housing prices in the 1990s were related to the decline in outmigration rates among immigrants in both Montreal and Vancouver. Higher housing prices in the 1990s were also related to increased in-migration rates among long-term immigrants in Vancouver. Finally, the slowdown in the rate of economic restructuring during the 1990s tended to reduce the out-migration rate among long-term immigrants in Toronto and Vancouver [model 4 versus model 1, top panel, table 5(a)]. In contrast, this slowdown tended to increase the out-migration rate but reduce the in-migration rate among long-term immigrants for Montreal. This suggests that the migration behaviour of immigrants is more responsive to local-labour market conditions in Montreal. 5 Conclusions From analyses of microdata from five consecutive censuses covering the period between 1981 and 2001, we found that the migration ^ immigration regimes in Canada's three major gateway centers are distinct, and the relationships are often subtle and inconsistent. The only evidence supporting the displacement hypothesis is the association between growth in the recent immigrant population and the increased out-migration rate among the lesswell-educated Canadian-born in Toronto and Vancouver. To a degree this confirms the displacement hypothesis with respect to low-skilled and less-well-educated populations. In Montreal the association between immigration growth and out-migration was much weaker, although in the same direction. This is expected as Montreal had the lowest rate of