This paper compares the occupational and earnings attainments of male immigrants to Canada to those of native-born men to test explanations based on prejudice/discrimination, competitive market, and national origin-related productivity theories using data from the Public Use Microdata File of the 1991 Census of Canada. Although there is no indication that the returns to human capital among immigrants educated in Canada are any less than those received by the native-born majority, I do find that immigrants educated abroad receive lower returns to education in the form of both occupational status and earnings than the native born, as well as no returns to foreign labour market experience. A macro-level model could detect no effect of a measure of prejudice on the level of returns in the form of status points or earnings. I did, however, find a substantial effect of level of development of country of origin on earnings returns to postsecondary education among immigrants educated abroad.
Dans cette communication l'auteur confronte les realisations professionnelles et de gains des immigrants masculins canadiens avec ceux des hommes de souche canadienne afin d'examiner les explications axees sur le prejuge et la discrimination, le marche concurrentiel et le rendement national selon l'origin du travailleur. Pour faire ceci il a eu recours aux donnees publiees dans les Fichiers de microdonnees a grande diffusion, Recensement du Canada, 1991. Quoiqu'il n'y aie pas de signe que le rendement au capital humain pour les immigrants formes au Canada est moins que celui pour les canadiens majoritaires de souche, l'auteur constate que les immigrants formes a l'etranger, quand ils obtiennent la meme annee de scolarite, sont moins recompenses en gains et statut professionnel par rapport a leurs homologues de souche canadienne. De plus, ces-premiers ne profitent pas de leur experience dans le marche etranger du travail. Un modele concu sur le microplan n'a pas decele un effet issu d'une mesure du prejuge sur la quantite du rendement en forme des gains et des points de statut. Neanmoins, l'auteur a bien constate un effet substantiel issu du niveau de developpement du pays d'origine en fonction de gains relatifs a l'education post-secondaire obtenue par les immigrants formes a l'etranger.
As a nation founded by immigrants, Canada has continued to welcome newcomers in substantial numbers, and is, along with Australia, Israel, and the United States, one of the main immigrant-receiving nations of the world. As of the 1996 Census, 17.4% of Canada's population was born in another country, the largest percentage in more than fifty years. Are the opportunities afforded these immigrants in Canada's labour market equivalent to the opportunities of the Canadian-born, given equivalent human capital endowments? Or are the foreign-born relegated systematically to lower levels of the occupational structure and earnings distribution than would be warranted by their training and experience? The term 'discrimination' is normally applied by sociologists to this situation, though it is conceptually complex and notoriously difficult to measure (Evans and Kelley, 1991). In contrast, economists studying the phenomenon allude to discrimination only as a last recourse. (2) Instead, most favour either a neoclassical p osition that assumes that "rational" employers will not practice discrimination because it means that they must forego hiring cheaper minority workers and hence reduce their profits, or introduce additional variables to explain earnings differences between immigrant and native-born workers, such as changes in country of origin composition overtime, the self-selection of immigrants, or country differences in immigration policy (Borjas, 1994).
Theories of Ethnic Inequality
Sociologists have typically assumed that, in the labour market, ethnic or racial prejudice on the part of employers, and sometimes workers, has produced discriminatory behaviour that restricts the opportunities of members of minority groups that are its object (see Evans and Kelley, 1991). Such discrimination might take the form of refusing to hire members of the minority group or, once hired, paying them wages lower than the prevailing wage or refusing to promote them regardless of training, skills, or experience. (3) The former constitutes exclusionary discrimination, while the latter two are defined as economic discrimination, measured in terms of the extent to which the returns to human capital endowments depart from returns among members of the majority group, although economists sometimes measure it as a residual difference in earnings between majority and minority groups after human capital and other variables with an effect on earnings are controlled (e.g. deSilva, 1992). Since the degree of prejudice is assumed to be a function of social and cultural distance, the theory would predict that economic discrimination would increase as a function of distance. Not only that, but we should observe little appreciable difference in the returns to education between those who immigrated as adults, obtaining their education in the source country, and those immigrating at school age, obtaining at least a portion of their education in Canada.
Despite the plausibility of this theory, a number of political and economic developments in postindustrial societies would imply that, despite the continued presence of prejudicial attitudes, discrimination would be on the decline. Indeed, Evans and Kelley (1991) could find little evidence of economic discrimination against immigrants to Australia. On the political side, Canada now has in place employment equity legislation that requires equal treatment in sectors of the labour market for women, the disabled, members of visible minority groups (i.e. racial minorities), and aboriginal people. As well, the public expression of extreme prejudice is barred by federal anti-hate legislation which prohibits the promotion of hatred against an identifiable group. On the economic side, an employer must possess the power and other resources necessary to translate prejudicial beliefs into discriminatory actions. Increasingly, "employers" are hired managers charged with maximizing the profitability of the enterprise by sh areholders, who have little interest in how that is accomplished. If the practice of ethnic discrimination reduces profits, managers have a strong inducement to eliminate it.
Competitive Market Theory
This last point states the essence of the neoclassical economic argument (Becker, 1971): given a free and competitive labour market and equivalent productivity, cheaper immigrant labour will inevitably be hired in preference to more expensive native-born labour because the practice boosts profitability. Employers who insist on discriminating against immigrants will pay the price in higher costs and lower profits. The "rational" employer, subject to competition with firms employing cheaper immigrant labour, will overcome his or her prejudicial attitudes and also hire immigrants, bringing the market back into equilibrium. The theory thus predicts that, even though immigrants and the native born may differ in the absolute quality of jobs and earnings, the returns to education, experience, and language ability they receive should be equivalent.
A serious problem with the neoclassical argument is that it must assume that all workers compete in a single, uniform labour market. Drawing on earlier work in institutional economics (Averitt, 1968; Doeringer and Piore, 1971) much sociological research on occupational and earnings attainment has been based on the assumption that market economies are not uniform systems (e.g. Beck et al., 1978; Wanner and Lewis, 1983). Instead, they are assumed to be segmented into two or more sectors in which both capital and labour markets operate differently (see Krahn and Lowe, 1998: 122-26). Labour market segmentation theory has been linked to ethnicity through Bonacich's (1979) split labour market hypothesis and Wilson and Portes' (1980) ethnic enclave hypothesis. Bonacich proposed that ethnicity is the crucial determinant of allocation to core or periphery sector employment, with visible minority groups overwhelmingly being consigned to the latter. In contrast, Wilson and Portes claimed that for some ethnic groups isol ation has facilitated small business development, permitting ethnic businesses to offer employment opportunities comparable to what is available in the broader labour market. These positions have stimulated a substantial amount of research on the question of immigrant success in the labour market, and have fundamentally reframed the issue of immigrant "success" compared to that of native-born participants in the labour force (see, for example, Nee et al., 1994; Portes and Jensen, 1989; Waldinger et al., 1990). Although I do not propose to test directly either segmented labour market or ethnic enclave hypotheses here, entailments of the former would likely be consistent with the predictions of prejudice/discrimination theory, while predictions from the latter would likely, with some modification (Evans, 1997), be consistent with the implications of neoclassical theory.
Quality of Skills and Self Selection
Borjas (1992, 1994) has been the leading advocate of an explanation for lower returns to skills among some immigrant groups based on the changing country of origin composition of the immigrant population over time combined with self-selection of the immigrant flow. Decomposing the skill decline in the U.S. between 1960 and 1980 into portions due to the national origin composition and change in skill level of immigrants from particular countries, Borjas (1992) demonstrates that the changing national origin mix accounts for over 90% of the observed decline in educational attainment and relative earnings across successive waves of immigrants. If it is the case, as Evans and Kelley (1991) demonstrated for Australia, that there is a greater return in the form of both earnings and occupational status to higher levels of education than to lower levels, declining levels of educational attainment among more recent immigrants would by itself account for a portion of any lower returns that are observed.
In addition, immigrants are not randomly selected members of the source countries. Depending upon the economic situation in the source country and the immigration policies of the host country, they will have on average higher or lower human capital endowments than non-immigrants from the same country. Making his argument using a formal model of equilibrium skill sorting, Borjas (1990) shows that rate of return to skills in the source country is correlated negatively with the earnings of immigrants in the host country. Since developing countries tend to be characterized by higher rates of return, and the proportion of immigrants from these countries has increased, Borjas's argument predicts that the less developed the source country, the lower the returns to education we should observe in the host country. Further, these lower returns should not apply to childhood immigrants educated in the host country. I will test these predictions using a multilevel model in which returns to education by country of birth ar e predicted from GNP per capita and average educational level of the source country, along with a measure of prejudice against immigrants from specific countries among the Canadian-born.
The Canadian Situation
Immigration Trends and Policy
Like the United States, Canada experienced an enormous surge of immigration, primarily from Europe, at the turn of the century. The opening of the Canadian West, combined with government-sponsored recruitment efforts, lured hundreds of thousands of immigrants per year, many of them from peasant backgrounds. These numbers dropped precipitously, first during World War I, then with the onset of the Great Depression of the 1930s, during which emigration exceeded immigration (McVey and Kalbach, 1995). In part these declines were due to policy changes that severely restricted the immigration of non-Europeans, particularly the Chinese who were singled out for exclusion by the Chinese Immigration Act of 1923. World War II further depressed immigrant numbers to record lows. Following the war, however, rapid economic development in Canada combined with difficult political and economic circumstances in Europe pushed immigration levels back to numbers not seen since the turn of the century. During the 1990s, Canada has b een receiving in excess of 200,000 immigrants annually, nearly 1% of its population.
Changes to immigration regulations in 1967 eliminated the system of "preferred nationalities" in favour of a point system that screens potential immigrants on the basis of possession of skills suitable for the labour market. As a consequence, the composition of immigrants on the basis of country of origin shifted dramatically. In the period 1971-75, nearly 55% of immigrants to Canada came from either Europe or the United States, with Britain being the single leading source of immigrants (McVey and Kalbach, 1995). By 1991, a majority of new immigrants came from Asia (52%), with just 24% coming from Europe and the U.S. (McVey and Kalbach, 1995).
Under current regulations, applicants for immigration to Canada are classified under one of three classes: independent immigrants subject to the points system, including business immigrants, who must either invest in Canadian enterprises or establish one themselves; family class immigrants who have a relative in Canada who has promised, and is willing and able, to look after their care and shelter; and refugees, admitted only under special circumstances. The first two classes are subject to the point system, under which they are screened on the basis of such factors as education, specific job skills, job experience, occupation, arranged employment, age, and ability to communicate in English or French.
As Borjas (1993) has shown, the introduction of the point system has had a substantial impact on the average educational attainment of immigrants to Canada since the 1970s. Nevertheless, as he points out, the effect of the points system has been to make it more likely that persons originating in "high education" countries will qualify, rather than to select persons with more schooling from specific countries. Another trend has also contributed to shifting the balance toward immigrants from Third World countries: between the early 1970s and late 1980s, the proportion of immigrants entering under the independent class has declined from 74% to less than 45% (deSilva, 1992). Over the same period refugees have increased from a negligible 1.3% to nearly 18% of new immigrants. Future research on the economic integration of immigrants to Canada would do well to control for class of immigrant, since refugees have on average lower earnings, higher unemployment rates, and are considerably more likely to be receiving wel fare benefits than are members of other immigrant classes (Citizenship and Immigration Canada, 1997). Nevertheless, refugees still constitute a small percentage of Canada's total immigrant population, and failure to control for immigrant class should not affect the results reported here.
Previous Research on Canada
A large research literature exists on the occupational and earnings attainment of Canada's immigrant population, but much of that research by sociologists is based on data collected in the 1970s or early 1980s and is now seriously dated in view of the shifting country of origin composition of immigrants to this country combined with dramatic changes in the Canadian economy over the past three decades which saw the decline of manual work, the rise of the "service economy" (Economic Council of Canada, 1991), and the entry of women into the labour market in unprecedented numbers. Over the same period, Canada's economy completed a transition from an industrial mode of organization in which a majority of workers are found in the primary (agricultural and extraction) and secondary (manufacturing and construction) industrial sectors to a postindustrial mode in which a vast majority of workers are found in the tertiary (service) sector. Indeed, by 1991 over 73% of Canada's labour force was employed in service industr ies (Statistics Canada, 1993).
The issue of the role of immigration and ethnic affiliation in social class formation has a long history in Canadian sociology, originating largely in John Porter's (1965) seminal book, The Vertical Mosaic. Basing his argument on the situation in Canada up to the 1950s, Porter contends that a Canadian immigration policy that favoured professionals and skilled labour, combined with a failure to develop a system of postsecondary education, produced a "mobility deprivation for Canadians" (Porter, 1965: 49). In contrast, he contends that both the United States and the United Kingdom encouraged immigration by less skilled individuals, which had the effect of moving the native-born population to higher occupational levels.
In most respects, subsequent research has not supported Porter's claims. (4) Using data from the 1973 Canadian Mobility Study, Boyd (1985) found that foreign-born males had a great deal more difficulty converting educational attainment into higher status occupations than did native-born males. Further, only men born in Great Britain were able to overcome the handicaps inherent in immigration to exceed the average occupational status of Canadian-born men. For other origin groups, she found foreign birth to be a distinct handicap in occupational attainment, this despite an immigration policy that favoured the well-educated or those with specific occupational skills in demand in the labour market (Wanner, 1986). Boyd's (1984) findings for women are similar: the occupational statuses of foreign-born women were on average lower than those of both native-born women and foreign-born men, even after adjusting for educational attainment and social background. As in the case of men, the sole exceptions were women born in either the U.K. or the U.S.
The more recent research on immigrant economic outcomes by sociologists either contrasts all immigrants to the native-born population without regard to country of origin differences or differences among those educated in Canada or abroad (Creese et al., 1991), studies only recent immigrants (Piche et al., 1998), or includes few controls (Richmond, 1992). In general these studies find that, despite coming from higher status social origins and attaining on average higher levels of schooling, immigrants are not as able as the Canadian-born to convert these advantages into higher levels of occupational status. In the case of earnings, it has repeatedly been found that disparities between native-born racial minorities and Canadians of European origin are smaller than those between the latter and foreign-born racial minorities (Boyd, 1992; Reitz and Breton, 1994). If what distinguishes foreign-born from native-born minorities is the amount of Canadian labour force experience and education, the present study, even t hough it focuses only on immigrants, must be careful to separate those who were educated abroad from those educated principally in Canada.
By estimating models that include both recent and established immigrants, those educated both in Canada and abroad, measures of country of origin, and extensive controls, economists studying immigrant integration frequently find no evidence of earnings discrimination against immigrants, at least among those educated in Canada (deSilva, 1992). While wage gaps between recent immigrants and the native born are frequently found, the existing evidence suggests that these tend to disappear as immigrants acquire Canadian experience (Bloom and Gunderson, 1991; Chiswick and Miller, 1988; Meng, 1987). However, in all of this research, the evidence for discrimination rests on the net earnings gap between immigrants and the native born after all other factors related to earnings have been controlled. In the present study, the emphasis is instead on differences in returns to education in the form of occupational status and earnings, an emphasis more closely related to the most frequent definition of economic discriminatio n as receiving lower returns than others with the same productivity (see Aigner and Cain, 1977).
Data and Methods
The analysis presented here is based on the 1991 Census of Canada Public Use Microdata File. This file contains a 3% random sample of individuals enumerated during the 1991 Census, over 800,000 cases in all, and includes a rich array of social and economic characteristics. Although some forms of analysis are inhibited by the highly aggregated form in which some variables are represented in the file, sufficient detail is available on such variables as education, age, year of immigration, skill in one of the official languages, country of origin, earnings, and occupation to make this analysis possible. The analysis focuses particularly on the 152,501 non-aboriginal men age twenty to sixty-four who were working mainly full-time in 1990 and employed by others.
The measures of social distance to various country of origin groups utilized in the multilevel model described later in the paper are from a national survey of 3,325 Canadians conducted in 1991 for the Multiculturalism and Citizenship Canada by the Angus Reid Group, a commercial public opinion research firm (Angus Reid Group, 1991). For purposes of the interviews, telephone numbers were generated randomly by census division making the sample proportionate to population distributions. To ensure that Canada's largest urban places were adequately represented, the sample was stratified to oversample residents of the Toronto, Montreal, and Vancouver Census Metropolitan Areas.
Sample Selection Criteria
Although restricting attention to men ignores a sizable proportion of Canada's labour force and makes it impossible for us to take account of the gendered nature of labour market structures, the census data do not include sufficient detail on actual labour market experience to represent adequately the more complex careers of women, which likely vary considerably across immigrant groups. In the case of men, it is not unreasonable to estimate experience as age at the time of the survey minus age at which education was completed. For women, this would provide an extremely biased estimate. (5) As well, in earnings and occupational status models for women, one must correct for sample selection bias to take account of the greater propensity of women who receive low wage offers to leave employment compared to men. As a result of this selection into employment, the earnings and occupational distributions of a sample of employed women may be truncated in a way that biases coefficients reporting the effects of variable s such as education or experience and requires a correction for such sample selection bias (see Winship and Mare, 1992).
Since the vast majority of Canadian men complete high school and nearly half of all high school graduates now go on to some form of post-secondary schooling, I have set the lower age limit to twenty to approximate the typical age at labour force entry. Men over age sixty-four were not included to avoid selectivity bias associated with higher mortality rates associated with lower socioeconomic status and with the timing of retirement.
Respondents whose ethnic background is aboriginal, including North American Indians, Metis, and Inuit, were excluded from the analysis. Not only do these groups experience extremely high unemployment rates and low labour force participation rates compared to other Canadians, but a substantial proportion of them live on rural reserves cut off from the mainstream economy.
Models and Measures
The individual-level analysis reported here is based on OLS regression models estimating effects of educational attainment, labour force experience in both Canada and abroad, and language proficiency on occupational status of first job and earnings, (6) with controls for marital status, the presence of other adults and children in the household, citizenship, and size of place. Since the main question to be addressed here concerns differences in returns to human capital across country of origin groups, my major focus is on interactions between country of origin and educational attainment. I first estimate a basic occupational or earnings attainment model in which the linear effects of the independent variables are estimated separately for each country of birth group:
[Y.sub.ij] = [[alpha].sup.j] + [SIGMA][[beta].sub.jk][X.sub.ijk] + [[epsilon].sub.ij] 
where Y is either status of respondent's job, measured as points on the International Standard Socia-Economic Index of Occupational Status, ISEI (Ganzeboom et al., 1992), or earnings, the Xs are the independent predictors, the [beta]s are the linear effects, and the subscripts i, j and k represent individuals, countries of birth, and independent variables, respectively. For the Canadian-born and immigrants educated in Canada, independent variables in the structural model are respondent's years of schooling, years of schooling (minus ten) squared, years of Canadian labour force experience, years of Canadian experience (minus ten) squared, and proficiency in English or French plus the controls indicated above and weeks worked in the earnings models (see Appendix, Table Al). For immigrants educated abroad, years of foreign labour force experience and years of foreign experience (minus ten) squared are added to the model. Although not reported here, reduced form models deleting official language proficienc y are also estimated to determine the degree to which education has indirect effects via this variable.
I extend these baseline models to test for differences in intercepts across country of origin groups, implying differences in mean status or earnings as follows:
[Y.sub.i] = [alpha] + [SIGMA][[beta].sub.k][X.sub.ik] + [[beta].sub.j][C.sub.j] + [[epsilon].sub.i] 
where the [C.sub.j] are dummy variables representing country of birth groups and the reflect differences in intercepts between immigrants from each source country and the Canadian-born. Finally, a model testing for differences in the effects of years of schooling and years of schooling squared between the immigrant groups and the Canadian-born is specified as:
[Y.sub.i] = [alpha] + [SIGMA][[beta].sub.k][X.sub.ik] + [SIGMA][[beta].sub.j][C.sub.j] + [SIGMA][[beta].sub.j]([C.sub.j][X.sub.1]) + [SIGMA][[beta].sub.j]([C.sub.j][X.sup.2.sub.1]) + [[epsilon].sub.i] 
where [X.sub.1] is years of schooling and [X.sup.2.sub.1] years of schooling squared.
Assessing Fit with a Large N
As noted above, my sample size exceeds 152,000 cases, despite the age, gender, non-aboriginal origin, and full-time work restrictions imposed. Therefore, the use of conventional p-values to test null hypotheses is likely to result in finding even negligible effects to be statistically significant. As a result, I supplement these tests with Baysian Information Criterion (BIC) values developed by Raftery (1995). The BIC statistic addresses the question: given the observed data, which model, [M.sub.1] or [M.sub.2], is more likely given the data? In the case of an OLS regression model where model, is being compared to the null model, Raftery proposes that a version of BIC based on comparisons to the null model, BIC', be defined as follows:
[BIC'.sub.k] = n log(1-[R.sup.2.sub.k]) + [p.sub.k] log n 
where n refers to sample size, [R.sup.2.sub.k] indicates the value of the coefficient of determination for some hypothesized model, [M.sub.k], and [P.sub.k] is the number of independent variables. In this form, negative values of [BIC.sup.1], indicate that the hypothesized model has a higher probability than the null model. As well, the more negative the value of [BIC.sup.1], the more likely the model. In assessing the significance of individual coefficients, I also use a BIC-based method involving the adjustment of the t-value to take account of sample size and grade of evidence. Given a sample size in excess of 100,000 and requiring what Raftery calls "positive" evidence, I will use a minimum t-value of 3.68 to reject the null hypothesis (see Raftery, 1995: Table 7).
Findings: Individual-Level Models
Table 1 (all Tables and Figures appear in the Appendix) reports means and, where appropriate, percentages for the main variables in the analysis by country of birth, as well as characteristics of those countries to be used in the macro-level models. The Canadian-born average 12.5 years of schooling and over 19 years of labour force experience. On the ISEI occupational status scale, occupational status averages 42.3, representing occupations at the midranks of white-collar jobs, such as real estate agents or insurance sales persons. This should not be surprising, since 73% of Canada's labour force was employed in the service sector by 1991. Approximately 90% of the mean annual total incomes of the native-born, $36,490, consisted of earnings from employment.
Immigrants from Anglophone or Francophone Countries
I have classified these immigrants separately, because their mother tongue is one of Canada's official languages, presumably giving them an advantage in the labour market. By our definition, only a small fraction of these immigrants have weak language skills, and regardless of country of birth, they exceed the mean years of schooling of the Canadian-born. This educational advantage translates into both higher occupational status and higher earnings than the Canadian-born. These advantages are likely related to Canada's policy of recruiting immigrants from these countries on the basis of the point system described above.
With the exception of immigrants originating in southern Europe, non-anglophone or francophone European immigrants also tend to be better educated than the Canadian-born and typically hold higher status occupations. In contrast, immigrants from southern Europe have the lowest levels of educational attainment of any immigrant group, which is in turn reflected in their lower occupational standing and lower average earnings. That immigration from European countries peaked in the 1950s and 1960s is reflected in the lengthy Canadian careers these immigrants have had on average. It is surprising to note that, given the lengthy period on average they have resided in Canada, all immigrant groups from Eastern and Southern Europe either continue to speak a nonofficial language at home or are unable to carry on a conversation in an official language, a trait not shared by immigrants from Northwestern Europe.
Asian and Middle Eastern Immigrants
Overall, immigrants from these countries have educational attainment levels equal to or greater than those of the Canadian-born, with the sole exception of those originating in Viet Nam. (7) Despite their relatively high educational levels and commensurate levels of occupational status, average earnings from employment are lower among Asian immigrants, though this may in part reflect their considerably lower amounts of labour force experience, particularly Canadian experience. Like immigrants from Eastern and Southern Europe, large proportions of those from the Middle East and Asia either cannot carry on a conversation in French or English, or continue to speak their native tongue at home.
Immigrants from Africa, the Caribbean, and Latin America
I have treated nonwhite immigrants originating in Africa and the Caribbean separately because they are likely to have had lower opportunity levels than whites originating in the same regions. The term 'nonwhites' is used instead of blacks, since it was not possible to distinguish blacks from other racial groups, particularly South Asians, immigrating from these regions lacking a race item in the Census questionnaire. In the case of African nonwhites, their educational attainment is considerably above that of the native born, as is their occupational status. However, their earnings are lower, in part perhaps because they have considerably less Canadian labour force experience. Immigrants from the Caribbean who are predominantly black (Ponting and Wanner, 1983), are characterized by an average schooling level equivalent to that of the Canadian-born, but are found in lower status occupations and earn significantly less despite their strong skills in one of the official languages compared to other immigrant group s.
Country of Birth Differences in Returns to Schooling
As indicated above, the analysis here will define economic discrimination primarily in terms of differences in returns to schooling by country of birth. In particular, prejudice/discrimination theory holds that prejudicial attitudes are directed against those who are somehow different culturally, racially or in terms of some other visible trait, such as an accent. Therefore, the theory would predict lower status and earnings returns to schooling among immigrants compared to the native born, regardless of whether an immigrant is educated abroad or in Canada, though the differences might be smaller for the latter group since they would be less "different" culturally and in terms of language usage.
Table 2 reports fit statistics and BIC values for the baseline models described above, a second model which permits the intercepts to differ across country of origin groups, a third model which permits linear education effects to differ across country of origin groups, and a fourth model permitting differences in quadratic education effects. Models are reported for both occupational status and earnings separately for immigrants educated in Canada and immigrants educated abroad.
Contrary to the predictions of prejudice/discrimination theory, the best fitting model contrasting the Canadian-born to immigrant groups educated in Canada, according to the BIC criterion appropriate to our extremely large sample, is the baseline model permitting no differences in intercepts or differences in returns to education. This is true whether returns in the form of status points or dollars of earnings are considered. In effect the substantial differences in mean earnings between immigrant groups observed in Table 1 are due entirely to differences in the human capital endowments of these groups, at least insofar as they were educated in Canada. As well, the earnings or status returns to an additional year of schooling, regardless of the level of that schooling, are equal across country of origin groups. Thus, there is no evidence of economic discrimination against these groups.
The picture is considerably different in the bottom panel of Table 2, where the Canadian-born are contrasted with immigrants educated abroad. For both occupational status and earnings, not only are the intercepts for immigrants different from that of the native-born, but the best fitting model in both cases also permits the linear effects of education to vary, again according to the BIC criterion. Is this evidence of economic discrimination against immigrants, or are these differences observed because immigrants educated abroad are less productive than persons educated in Canada, as the neoclassical theory would have it? I will leave this question until later, as I first examine detailed patterns in the separate models for country of origin groups presented in Tables 3 and 4.
Models for Occupational Status
Table 3 reports a model for each of the countries or regions listed in the tab, separately for immigrants educated in Canada and those educated abroad. A glance at the columns reporting returns to years of schooling and years of schooling squared among those educated in Canada confirms the results from Table 2: these columns are virtually devoid of daggers indicating significant differences from the corresponding Canadian-born coefficient. Among those educated in Canada, only immigrants from anglophone or francophone countries and those from South Asia have returns that differ from those of the native-born, though in the first instance returns are significantly greater. Among South Asian immigrants, it is the quadratic term that differs, implying that returns to higher levels of schooling are less than those to the Canadian-born.
The picture is dramatically different among those who immigrated as adults. Without exception, all ethnic groups receive significantly lower returns to schooling than the native-born, consistent with the results reported in Table 2. Among most immigrant groups, the effect of Canadian labour force experience is nonsignificant, implying that their occupational status changes little over the course of their careers, and is largely set by educational attainment. This is consistent with results reported by Creese et al. (1991). Weak language proficiency has a substantial negative effect on the occupational standing of seven of fourteen groups, but the effect is particularly marked in the cases of those from South Asia and the PRC. Based on a comparison of the schooling and schooling squared coefficients in the structural (reported in Table 3) and reduced form equations (not reported here), there is little evidence that the effect of schooling is transmitted indirectly via the language ability measure.
These effects of schooling can best be observed in the upper panels of Figures 1 through 3, which report predicted values of occupational status for each year of schooling from the structural models in Table 3. (8) Only the Canadian-born and immigrants educated abroad are represented here, since our results in Table 2 confirm that there are no real differences in returns to schooling between the native-born and immigrants educated in Canada. In the upper panel of Figure 1, while the curves are fairly tightly clustered, those for Southern and Eastern Europe are clearly shallower than that for the Canadian-born. In Figure 2, the departure of the native-born curve from the others is particularly striking; it crosses the band created by the roughly parallel curves for the other countries of birth at a sharp upward angle. Also of note here is the clustering of the curves for South Asia, Vietnam, and the Philippines at the bottom, and those for the remaining Asian countries at the top, indicating that immigrants fr om those three countries occupy lower status positions than the others at all levels of education, as the significant differences in intercepts observed in the bottom panel of Table 2 testify. A similar pattern can be observed in Figure 3. Again the curves for African nonwhites, Caribbean nonwhites, and Latin Americans are approximately parallel, with the Canadian-born curve jutting upward across them, indicating the greater returns to schooling at all levels the native-born receive. In terms of absolute returns, nonwhites from Africa (a large proportion of whom I suspect are actually of South Asian ancestry) attain higher occupational status at all levels of education than do Caribbean nonwhites and Latin Americans.
Models for Earnings
As in the case of the occupational status models, the earnings models for those educated in Canada reported in the first panel of Table 4 are devoid of effects of schooling that differ significantly from the effects for the Canadian-born, as indicated by the daggers. Based on a comparison to reduced-form equations not reported here, the indirect effects of schooling via proficiency in an official language in the case of nearly all countries of origin are extremely weak. At the same time, there are strong indirect effects of schooling via occupation. These range in magnitude from 25% to 50% of the total effects, though there are several outliers and there appears to be no systematic pattern to the indirect effects.
As in the case of the models for occupational status, it is among immigrants educated abroad that substantial differences compared to the Canadian-born in returns to schooling are evident. With only the exception of anglophone or francophone countries, significantly lower net returns to the linear term can be observed. This is most easily seen in the lower panels of Figures 1 through 3. As in the case of occupational status, among European immigrants, those originating in the Eastern and Southern regions fare worst. Not only are their returns to schooling lower than those for the native-born, particularly at the highest levels among Eastern Europeans, but their mean earnings at all levels of schooling fall considerably below those of immigrants from anglophone or francophone countries and those from Northwestern Europe. Indeed, the net earnings of the latter two groups outstrip those of the native-born, though returns to graded schooling among those from Northwestern Europe are quite flat.
Figure 2 compares the schooling effects of the Canadian-born to those born and educated in Asian regions and countries, to the decided disadvantage of the latter. Not only are the returns to schooling lower across the entire range among the Asian-born, except for the least educated, the Canadian-born earn considerably more. The picture is similar in Figure 3, which compares the effects of schooling among the Canadian-born to those immigrating from Africa, the Caribbean (nonwhites only), and Latin America. Here, the slopes associated with graded schooling among these groups are essentially flat, though in the postsecondary years they are indistinguishable from that of the native-born. What is particularly obvious here is that the predicted earnings at all levels of education are so much higher for the Canadian-born, with differences in the $8,000 to $10,000 range.
It is also clear from Table 4 that few immigrant groups benefit from labour force experience acquired abroad. The only group for whom significant positive effects are evident are those from anglophone or francophone countries. Immigrants educated in the People's Republic of China and Viet Nam even suffer a significant penalty for their years of experience abroad. In addition, half the groups suffer a penalty for either speaking a nonofficial language at home or being unable to carry on a conversation in English or French, in most cases a deficit in the $3,000 to $4,000 range.
Findings: A Macro-Level Model
To this point our results are somewhat ambiguous with regard to adjudicating between prejudice/discrimination theory and competitive market theory. On the one hand, there is compelling evidence that for Canada, childhood immigrants educated in the host country are not subject to economic discrimination in the form of lower status or earnings returns to schooling. On the other hand, those educated abroad do suffer lower returns to their schooling, particularly postsecondary schooling. Does this actually constitute discrimination, or is some other mechanism at work? As described above, Borjas's argument predicts that the less developed the source country, the lower the returns to education observed in the host country. Further, these lower returns should not apply to childhood immigrants educated in the host country. In other words, if we predict returns to schooling, particularly postsecondary schooling, from level of development of the source countries, we should observe a positive effect of development on re turns to schooling among immigrants educated abroad. In the case of prejudice/discrimination theory, we would anticipate that immigrant groups with whom the native-born population feels less comfortable would be the object of discrimination in the form of lower status jobs and reduced earnings for a given level of educational attainment.
I test directly both of these theoretical predictions by means of a macro-level model in which returns to both graded schooling and postsecondary schooling for the detailed country of birth groups listed in Table 1 are regressed on gross national product per capita in the source country, percent enrolled in secondary schooling in the source country, and a measure of the "comfort" the Canadian born feel toward members of specific immigrant groups. (9) Values for these macro-level variables are shown in Table 1. The comfort measure can be regarded as at least an indirect measure of prejudicial attitudes toward these groups. Although the original Likerttype items in the questionnaire provided a seven-point rating scale, I convert this to the scale from 0 to 100 reported in Table 1.
The structure of the macro-level model is:
[[beta].sub.jk] = [[gamma].sub.0] + [SIGMA] [[gamma].sub.k][Z.sub.jk] + [[micro].sub.j] 
where [[beta].sub.jk] represents the effects of either graded or postsecondary schooling on status of occupation or earnings, with subscripts j and k representing countries of origin and type of schooling, respectively; [Z.sub.1j] is the GNP per capita (in thousands of U.S. dollars) of the source countries; [Z.sub.2j] is the percent of eligible young persons enrolled in secondary schools in the source countries; [Z.sub.3j] is a measure of the comfort level of the native-born with each immigrant group, recoded to a scale of 0 to 100; the [gamma]s are estimated parameters of the model; and [[micro].sub.j] are the macro-level errors. Both GNP per capita (reported in Table 1) and percent of eligible young persons enrolled in secondary schools are from World Bank (1991). Since the comfort measure is available for just twelve of the twenty-five detailed source countries, I also include a dummy variable in the model coded 1 if a comfort value is present, 0 otherwise. Means are then substituted for the missing values, y ielding an unbiased estimate of the effect of the comfort value net of the missing cases (Cohen and Cohen, 1975).
Instead of using the coefficients associated with schooling and schooling squared in previous models, I instead utilize a "spline function" in which all schooling from zero to twelve or more years is coded separately from years of postsecondary schooling, in effect yielding separate measures of the effects of the two levels of schooling. Thus, graded schooling is coded '1' to '11' for respondents whose highest level corresponds to those years and '12' for those who have twelve or more years of schooling. The postsecondary variable is coded '0' for those who never attended college or university and in increments of one unit for each year attended. In this way the information provided by the quadratic function used in the core models is preserved, but I obtain a more readily interpreted index of returns to the two main forms of schooling.
The results of this macro-level analysis are presented in Table 5. Most striking here is the lack of a significant effect of the comfort measure on occupational or earnings returns of either form of schooling, a finding inconsistent with the predictions of prejudice/discrimination theory. While the effects of GNP per capita and percent enroled in secondary school on the effect of graded schooling on occupational status are marginally significant, the proportion of variance explained in this model does not reach significance. In the model for effects of postsecondary schooling on occupational status we observe no significant coefficients. In fact we should have expected none, since Borjas's hypothesis deals only with returns in the form of earnings. While there is no effect of any of the predictor variables on returns to graded schooling in the form of earnings, there is a substantial net effect of GNP per capita on returns to postsecondary schooling. Indeed, this model accounts for over 48% of the variance in these returns. Although I do not report separate models with the comfort variable as sole predictor, other analysis shows that it has no total effect on earnings returns, and hence the lack of a net effect is not due to strong collinearity with level of development.
As Evans and Kelley (1991) concluded for Australia, I find little support for the argument that widespread prejudice against ethnic minorities who have immigrated to Canada has led to economic discrimination. In the case of immigrants educated in Canada, there is no indication that the returns to human capital they receive are any less than those received by the native-born majority. On the other hand, I do find that immigrants educated abroad receive lower returns to their investment in education in the form of both occupational status and earnings than the native-born, and that these returns vary systematically with country of birth. Nevertheless, our macrolevel model predicting returns could detect no effect of a measure of prejudice on the level of returns in the form of status points or earnings. We did, however, find a substantial effect of level of development of country of origin on earnings returns to postsecondary education among immigrants educated abroad, consistent with the theory developed by Bo rjas to supplement conventional competitive market theories.
This strong effect of level of development of country of origin on earnings returns to postsecondary education and the low value accorded to experience acquired abroad are somewhat difficult to interpret. On the one hand, neo-classical economists would be inclined to see these results as a consequence of the lower quality of both foreign education and experience, particularly in Third World countries, and hence the lower productivity of such workers. Our finding from the macro-level model is certainly consistent with that interpretation. On the other hand, sociological research into the question has tended to view the devaluing of foreign credentials in the Canadian labour market as just another form of exclusionary discrimination (McDade, 1988), particularly since it seems to be particularly directed at members of visible minority groups who immigrate from less developed countries. While the finding here that level of comfort with immigrant groups has no effect on returns to education is inconsistent with th is view, it does not rule out the possibility that some immigrants suffer a form of exclusionary discrimination when their foreign credentials are devalued or ignored. This seems to occur primarily in the licensed professions, particularly law, medicine, and engineering, where the governing bodies of these professions tend to preserve their monopoly in the labour market by recognizing almost no credentials earned outside their jurisdiction, whether held by immigrants or the Canadian-born. (10)
Another piece of evidence that tends to cast doubt on prejudice/discrimination theory is the repeated finding in the literature that even immigrants educated abroad eventually close the earnings gap with the native-born (Chiswick and Miller, 1988; Meng, 1987). If in fact employers are discriminating against some visible traits of immigrants and not responding to productivity considerations, one would anticipate that such discrimination should continue throughout the immigrant's career, and not be reduced as he or she acquires experience in the host country's labour market.
One bothersome finding remains, however. Although we have ruled out economic discrimination in the form of lower returns to immigrant human capital, our Figures clearly show that certain immigrant groups, particularly those from less developed countries, also receive lower status jobs and lower earnings at all levels of schooling than do the native-born. This issue can only be addressed convincingly by the use of proper panel data that follow immigrant cohorts through their careers (Borjas, 1994) to capture the different composition of successive cohorts as well as the varying legal and economic conditions under which they enter the host county. Although it has become a hopeless cliche, further research on these issues is surely needed to understand the dynamics of the "vertical mosaic" (Porter, 1965) that is the stratification system of immigrant-receiving countries.
(1.) An earlier version of this paper was presented at the XIV World Congress of Sociology, Montreal, Quebec, Canada, July 27-August 1, 1998. The research reported here was supported by a grant from the Social Sciences and Humanities Research Council of Canada through the Prairie Centre of Excellence for Research on Immigration and Integration. Portions of the data used in this paper were provided by the Department of Canadian Heritage and collected by the Angus Reid Group, federal collection number SSC/MUL-050-3343. I wish to thank Fernando Mata for his assistance in acquiring these data.
(2.) Indeed, in his extensive review of economic research on immigration, Borjas (1994) only uses the term in a single footnote, and then only to indicate that there is no evidence for it in the United States. But see Hoffman, 1991.
(3.) Although prejudice is normally conceived of as an individual-level attitude set and discriminatory behaviour attributed to individual actors, both have institutional and structural dimensions. Prejudice is frequently embedded in a culture that transmits it by means of the socialization process, while discrimination can be accomplished through institutionalized practices not controlled directly by individuals. Bonilla-Silva (1997) has recently used the phrase "racialized social systems" to refer to societies in which the economic, political, social, and ideological institutions are at least partially structured by the placement of actors into racial categories.
(4.) Another research literature has more generally addressed the association between ethnic affiliation and occupational attainment in Canada, typically ignoring the distinction between immigrants and the native-born and using simple differences in means or adjusted means (see Li, 1996). Most of this research has concluded that the degree of ethnic occupational inequality has declined since the 1931-1961 period Porter (1965) analyzed.
(5.) Meng (1987) demonstrated that some bias is introduced into earnings functions for men by using potential experience, as measured here, instead of actual experience, not available in the 1991 PUMF. The difference between potential and actual experience among native-born men was 3.13 years in his 1973 sample, but 4.11 years among the foreign-born. Needless to say, these gaps would be considerably larger among women, producing more consequential biases in coefficient estimates.
(6.) It is conventional in earnings models to use a semi-logarithmic specification in which the natural logarithm of earnings is regressed on the predictor variables to adjust for the typically skewed distribution of the dependent variable. I have chosen to use dollars of earnings instead, because under the semi-logarithmic specification, the coefficients of the model are interpreted as proportional effects on the dependent variable, at least for small values. Since these effects are relative to the mean and variance of earnings, like standardized coefficients, it is not meaningful to compare them across country of origin groups. The unstandardized coefficients of the dollar regressions are expressed in absolute values that may be meaningfully compared. While it would be appropriate to use a semi-log specification in the models reported in Table 2, parallel analyses with the dependent variable defined as dollars of earnings and log earnings resulted in identical conclusions, hence I report only the former.
(7.) A substantial proportion of immigrants from Viet Nam entered Canada as refugees, and were hence not subject to the point system that would have given preference to the better educated.
(8.) Confidence intervals around these predicted values are not reported, since they would duplicate the tests reported in Tables 2, 3 and 4.
(9.) The exact wording of the item in the Multiculturalism Attitude Survey described above is: "I would like you to think of recent immigrants to Canada. These are persons who were born and raised outside of Canada. How comfortable would you feel being around individuals from the following groups of immigrants..." (list of country of origin groups).
(10.) While a useful extension of this research would be to compare the effect of postsecondary schooling on the attainment of licensed professional occupations, the lack of a detailed occupational code in the Census PUMF file would make this impossible.
Aigner, Dennis J. and Glen G. Cain. 1977. "Statistical Theories of Discrimination in Labor Markets." Industrial and Labor Relations Review 30: 175-87.
Angus Reid Group. 1991. Multiculturalism and Canadians: Attitude Study 1991. Ottawa: Multiculturalism and Citizenship Canada.
Averitt, Robert T. 1968. The Dual Economy: The of American Industry Structure. New York: W. W. Norton.
Beck, E. M., Patrick M. Horan, and Charles M. Tolbert II. 1978. "Stratification in a Dual Economy: A Sectoral Model of Earnings Determination." American Sociological Review 43: 704-720.
Becker, Gary. 1971. The Economics of Discrimination, 2nd Edition. Chicago: University of Chicago Press.
Bloom, David E. and Morley Gunderson. 1991. "An Analysis of the Earnings of Canadian Immigrants." Pp. 321-367 in John M. Abowd and Richard B. Freeman (Eds.), Immigration, Trade and the Labor Market. Chicago: University of Chicago Press.
Bonacich, Edna. 1979. "The Past, Present, and Future of Split Labor Market Theory." Pp. 1764 in C.B. Marrett and C. Leggon (Eds.), Research in Race and Ethnic Relations. Greenwich, Connecticut: JAI Press.
Bonilla-Silva, Eduardo. 1997. "Rethinking Racism: Toward A Structural Interpretation." American Sociological Review 62: 465-480.
Borjas, George J. 1990. Fri ends or Strangers: The Impact of Immigrants on the U.S. Economy. New York: Basic Books.
_____. 1992. "National Origin and the Skills of Immigrants in the Postwar Period." Pp. 17-47 in George J. Borjas and Richard B. Freeman (Eds.), Immigration and the Work Force: Economic Consequences for the United States and Source Areas. Chicago: University of Chicago Press.
_____. 1993. "Immigration Policy, National Origin, and Immigrant Skills: A Comparison of Canada and the United States." Pp. 21-43 in David Card and Richard B. Freeman (Eds.), Small Differences that Matter: Labor Markets and Income Maintenance in Canada and the United States. Chicago: University of Chicago Press.
_____. 1994. "The Economics of Immigration." Journal of Economic Literature 32: 1667-1717.
Boyd, Monica. 1984. "At a Disadvantage: the Occupational Attainments of Foreign-Born Women in Canada." International Migration Review 18: 1091-1119.
_____. 1985. "Immigration and Occupational Attainment in Canada." Pp. 393-445 in Monica Boyd, John Goyder, Frank E. Jones, Hugh A. McRoberts, Peter C. Pineo and John Porter. Ascription and Achievement: Studies in Mobility and Status Attainment in Canada. Ottawa: Carleton University Press.
_____. 1992. "Gender, Visible Minority, and Immigrant Earnings Inequality: Reassessing an Employment Equity Promise." Pp. 279-321 in Vic Satzewich (Ed.), Deconstructing a Nation: Immigration, Multiculturalism, and Racism in '90s Canada. Halifax: Fernwood Publishing.
Chiswick, Barry and Paul Miller. 1988. "Earnings in Canada: The Roles of Immigrant Generation, French Ethnicity, and Language." Research in Population Economics 6: 183-228.
Citizenship and Immigration Canada. 1997. "Preliminary Profile of Immigrants Landed in Canada over the 1980 to 1994 Period." Paper presented at the Conference on Immigration, Employment, and the Economy sponsored by the Canadian Employment Research Forum and Citizenship and Immigration Canada, Richmond, B.C., October 17-18.
Cohen, Jacob and Patricia Cohen. 1975. Applied Multiple Regression/Correlation Analysis for the Behavioural Sciences. New York: John Wiley & Sons.
Creese, Gillian, Neil Guppy, and Martin Meissner. 1991. Ups and Downs on the Ladder of Success: Social Mobility in Canada (General Social Survey Analysis Series No. 5). Ottawa: Statistics Canada.
deSilva, Arnold. 1992. Earnings of Immigrants: A Comparative Analysis. Ottawa: Economic Council of Canada.
Doeringer, Peter B. and Michael J. Piore. 1971. Internal Labor Markets and Manpower Analysis. Lexington, Massachusetts: D.C. Heath.
Economic Council of Canada. 1991. Employment in the Service Economy. Ottawa: Minister of Supply and Services.
Evans, M. D. R. 1997. "Do Ethnic Enclaves Benefit or Harm Linguistically Isolated Employees?" Working Paper: Institute for Advanced Studies, Australian National University.
Evans, M. D. R. and Jonathan Kelley. 1991. "Prejudice, Discrimination, and the Labor Market: Attainments of Immigrants in Australia." American Journal of Sociology 97: 721-759.
Ganzeboom, Harry B. G., Paul M. De Graaf, Donald J. Treiman, and Jan de Leeuw. 1992. "A Standard International Socio-Economic Index of Occupational Status." Social Science Research 21:1-56.
Ganzeboom, Harry B. G. and Donald J. Treiman. 1996. "Internationally Comparable Measures of Occupational Status for the 1988 International Standard Classification of Occupations." Social Science Research 25:201-39.
Hoffman, Emily P. (Ed.). 1991. Essays on the Economics of Discrimination. Kalamazoo, Mich.: W. E. Upjohn Institute for Employment Research.
Krahn, Harvey J. and Graham S. Lowe. 1998. Work, Industry, and Canadian Society. Toronto: ITP Nelson.
Li, PeterS. 1996. State of the Art Literature Review on Immigration: Sociological Perspectives. Ottawa: Citizenship and Immigration Canada.
McDade, Kathryn. 1988. Barriers to Recognition of the Credentials of Immigrants in Canada. Ottawa: Institute for Research on Public Policy.
McVey, Wayne W. Jr. and Warren E. Kalbach. 1995. Canadian Population. Toronto: Nelson Canada.
Meng, Ronald. 1987. "The Earnings of Canadian Immigrant and Native-Born Males." Applied Economics 19: 1107-1119.
Model, Suzanne. Forthcoming. "Immigrants' Social Class in Three Global Cities." in Malcolm Cross and Robert Moore (Eds.), Migration, Social Exclusion and the Globalisation of Urban Populations.
Nee, Victor, and Jimy Sanders and Scott Sernau. 1994. "Job Transitions in an Immigrant Metropolis: Ethnic Boundaries and the Mixed Economy." American Sociological Review 59: 849-72.
Piche, Victor, Jean Renaud, and Lucie Gingras. 1998. "Comparative Immigrant Economic Integration." in Shiva Halli and Leo Driedger (Eds.), Immigrant Canada. Toronto: University of Toronto Press.
Ponting, J. Rick and Richard A. Wanner. 1983. "Blacks in Calgary: A Social and Attitudinal Profile." Canadian Ethnic Studies 15: 57-76.
Portes, Alejandro and Leif Jensen. 1987. "The Enclave and the Entrants." American Sociological Review 54: 929-949.
Porter, John. 1965. The Vertical Mosaic: An Analysis of Social Class and Power in Canada. Toronto: University of Toronto Press.
Raftery, Adrian E. 1995. "Baysian Model Selection in Social Research." Sociological Methodology 25: 111-163.
Reitz, Jeffrey G. and Raymond Breton. 1994. The Illusion of Difference: Realities of Ethnicity in Canada and the United States. Toronto: C.D. Howe Institute.
Richmond, Anthony H. 1992. "Immigration and Structural Change: the Canadian Experience, 1971-1986." International Migration Review 24:1200-1221.
Statistics Canada. 1993.1991 Census of Canada, Industry and Class of Worker. Cat. 93-326. Ottawa: Minister of Industry, Science and Technology Canada.
Waldinger, Roger, Howard E. Aldrich, and Robin Ward. 1990. Ethnic Entrepreneurs: Immigrant Business in Industrial Society. Beverly Hills: Sage.
Wanner, Richard A. 1986. "Educational Inequality: Trends in Canada and the United States." Comparative Social Research 9: 47-66.
Wanner, Richard A. and Lionel S. Lewis. 1983. "Economic Segmentation and the Course of the Occupational Career." Work and Occupations 10: 307-324.
Wilson, Kenneth and Alejandro Portes. 1980. "Immigrant Enclaves: An Analysis of the Labor Market Experiences of Cubans in Miami." American Sociological Review 86:295-319.
Winship, Christopher and Robert D. Mare. 1992. "Models for Sample Selection Bias." Annual Review of Sociology 18: 327-50.
World Bank. 1991. World Development Report 1991: The Challenge of Development. New York: Oxford University Press.
[Figure 1 omitted]
[Figure 2 omitted]
[Figure 3 omitted]
Table 1 Descriptive Statistics for Income, Occupational Status, and Predictor Variables, Canadian Men Age 20-64 Employed Full-Time, 1991 Educated in Labour ForceExperience (a) Educated (a) Canada (a) In Canada Place of Birth (Mean Years) (Percent) (Mean Years) Canada 12.5 100.0 19.3 Anglophone or Francophone Countries United States 14.6 37.2 15.7 United Kingdom 13.6 40.2 19.5 France 14.3 41.2 18.2 African Countries 14.2 26.6 16.0 (Whites) Northwestern Europe Germany 12.8 48.0 25.0 Netherlands 12.2 18.9 27.7 Eastern Europe Poland 13.1 35.2 13.3 U.S.S.R. 13.0 24.3 23.9 Hungary 12.9 31.9 22.5 Southern Europe Italy 9.7 25.0 25.5 Portugal 8.4 20.1 15.8 Balkans 11.5 19.2 19.8 Greece 10.0 15.2 21.6 Middle East and West Asia Arab Countries 13.2 10.5 8.3 West Asia 14.1 27.4 7.8 Israel 14.5 15.5 11.4 South and East Asia South Asia 13.2 26.7 11.1 Hong Kong 14.7 14.6 9.0 P. Republic of China 12.4 16.2 14.8 Philippines 13.2 16.6 10.3 Viet Nam 11.6 18.3 8.3 Other Asia 13.5 15.5 10.1 African Countries 14.1 31.2 9.9 (Nonwhites) Caribbean (Nonwhites) 12.4 20.6 13.2 Latin America 12.6 21.4 11.8 Labour Weak Language Occupational ForceExperience (a) Abroad Proficiency (a) Status Score (a) Place of Birth (Mean Years) (Percent) (Mean) Canada 0.0 0.8 42.3 Anglophone or Francophone Countries United States 5.0 0.6 49.1 United Kingdom 5.3 0.6 46.8 France 5.9 0.0 48.5 African Countries 7.5 6.2 49.0 (Whites) Northwestern Europe Germany 3.9 12.2 43.9 Netherlands 3.3 4.2 43.7 Eastern Europe Poland 9.8 62.9 39.4 U.S.S.R. 8.5 45.1 46.1 Hungary 6.7 33.2 42.1 Southern Europe Italy 6.8 49.2 38.0 Portugal 11.3 65.0 34.1 Balkans 8.7 55.0 38.2 Greece 8.8 65.6 39.7 Middle East and West Asia Arab Countries 9.2 63.0 43.2 West Asia 10.5 63.3 44.1 Israel 6.6 37.1 50.4 South and East Asia South Asia 9.3 64.0 41.4 Hong Kong 7.2 78.7 50.6 P. Republic of China 12.7 83.8 44.3 Philippines 9.2 59.9 41.1 Viet Nam 9.2 86.4 37.6 Other Asia 9.2 64.8 44.9 African Countries 8.8 36.6 46.0 (Nonwhites) Caribbean (Nonwhites) 8.7 4.8 39.7 Latin America 8.8 35.2 40.4 GNP/Capita Percent in Annual Earnings (a) of Country Secondary Place of Birth (Mean $CDN) of Birth (b) Education (b) Canada 33168 19030 105 Anglophone or Francophone Countries United States 40611 20910 98 United Kingdom 43544 14610 83 France 37507 17820 94 African Countries 40978 2470 94 (Whites) Northwestern Europe Germany 38404 20440 104 Netherlands 39748 15920 81 Eastern Europe Poland 28774 1790 78 U.S.S.R. 37928 2614 71 Hungary 37660 2590 76 Southern Europe Italy 34034 15120 59 Portugal 28701 4250 80 Balkans 32388 2920 95 Greece 26955 5350 47 Middle East and West Asia Arab Countries 21021 7012 36 West Asia 23755 1690 83 Israel 41245 9790 37 South and East Asia South Asia 28317 320 74 Hong Kong 31608 10350 44 P. Republic of China 26043 350 71 Philippines 25871 710 48 Viet Nam 21985 500 46 Other Asia 31016 540 18 African Countries 28112 340 27 (Nonwhites) Caribbean (Nonwhites) 26677 2245 73 Latin America 26498 1950 48 Mean Comfort Number Place of Birth Score (c) of Cases (a) Canada 121938 Anglophone or Francophone Countries United States 1477 United Kingdom 91 5163 France 87 413 African Countries 337 (Whites) Northwestern Europe Germany 85 1472 Netherlands 928 Eastern Europe Poland 968 U.S.S.R. 87 293 Hungary 383 Southern Europe Italy 88 3030 Portugal 85 1590 Balkans 673 Greece 625 Middle East and West Asia Arab Countries 75 473 West Asia 73 354 Israel 86 62 South and East Asia South Asia 70 2325 Hong Kong 84 1053 P. Republic of China 84 1096 Philippines 875 Viet Nam 924 Other Asia 911 African Countries 1238 (Nonwhites) Caribbean (Nonwhites) 80 1366 Latin America 2442 (a)Source: 1991 Census of Canada Public Use Microdata File. (b)Characteristics of country of birth. Source: World Bank (1991). (c)Source: Multiculturalism and Canadians: Attitude Study 1991 microdata file. Table 2 Model Selection for OLS Regressions of Occupational Status and Earnings Canadian Born vs. Immigrants Educated in Canada Contrast Model Predictor Variables Contrast d.f. Models for Occupational Status  Baseline  vs. null 11   plus country differences in  vs.  14 intercepts   plus country differences in  vs.  14 linear education effects   plus country differences in  vs.  14 in curvilinear education effect Models for Income  Baseline  vs. null 12   plus country differences in  vs.  14 intercepts   plus country differences in  vs.  14 linear education effects   plus country differences in  vs.  14 curvilinear education effects Model [R.sup.2] for F-ratio for BIC for Model d.f. Model Comparison Model  11 .32658 5737 (*) -51321  25 .32697 5.33 (*) -51232  39 .32735 5.17 (*) -51140  53 .32795 8.23 (*) -51092  12 .38953 6919 (*) -64080  26 .38984 4.67 (*) -63981  40 .39003 2.94 (*) -63857  54 .39016 1.96 -63720 Canadian Born vs. Immigrants Educated Abroad Contrast Model Predictor Variables Contrast d.f. Models for Occupational Status  Baseline  vs. null 11   plus country differences  vs.  14 in intercepts   plus country differences  vs.  14 in linear education effects   plus country differences in  vs.  14 curvilinear education effects Models for Income  Baseline  vs. null 12   plus country differences  vs.  14 in intercepts   plus country differences  vs.  14 in linear education effects   plus country differences in  vs.  14 Model [R.sup.2] for F-ratio for BIC for Model d.f. Model Comparison Model  11 .32233 6117 (*) -54919  25 .32476 36.4 (*) -55261  39 .32826 52.6 (*) -55830  53 .32856 4.57 (*) -55727  12 .37330 7022 (*) -65969  26 .37562 37.5 (*) -66328  40 .37752 30.8 (*) -66593  54 .37777 4.15 (*) -66484 Note: Baseline include predictor curvilinear education effects variables capturing education, labour force experience, and controls. See Appendix Table A1. Preferred models are listed in italics. (*)p<0.01 Table 3 Unstandardized OLS Regressions of Occupational Status on Education, Labour Force Experience, and Official Language Proficiency by Place of Birth, Canadian Men Age 20-64 Employed Full-Time, 1991 Years of Years of Place of Birth Schooling [Schooling.sup.2] Canada 2.07 (*) 0.120 (*) Immigrants Educated in Canada Anglo- or Francophone Countries 3.14 (*,+) -0.032 (+) Northwestern Europe 1.66 (*) 0.103 (*) Eastern Europe 2.36 (*) 0.052 Southern Europe 1.69 (*) 0.146 (*) Middle East and West Asia 3.49 (*) -0.074 South Asia 3.65 (*) -0.098 (+) Hong Kong 3.95 (*) -0.114 P. Republica of China 2.18 (*) 0.033 Phillipines 4.10 -0.185 Viet Nam 3.33 (*) 0.095 Other S/SE Asia 5.83 (*) -0.278 African Countries (Nonwhites) 1.42 0.073 Carribean (Nonwhites) 3.09 (*) -0.016 Latin America 2.02 (*) 0.082 Canada 2.07 (*) 0.120 (*) Immigrants Educated Abroad Anglo- or Francophone Countries 2.01 (*,+) 0.057 (*,+) Northwestern Europe 1.30 (*,+) 0.126 (*) Eastern Europe 1.29 (*,+) 0.084 (*) Southern Europe 1.16 (*,+) 0.123 (*) Middle East and West Asia 1.35 (*,+) 0.104 (*) South Asia 1.23 (*,+) 0.119 (*) Hong Kong 1.48 (*,+) 0.071 P. Republic of China 1.02 (*,+) 0.156 (*) Philippines 1.45 (*,+) 0.039 Viet Nam 0.86 (*,+) 0.154 (*) Other S/SE Asia 1.57 (*,+) 0.096 (*) African Countries (Nonwhites) 1.78 (*,+) 0.057 Carribean (Nonwhites) 1.17 (*,+) 0.212 (*,+) Latin America 1.30 (*,+) 0.144 (*) Canadian L.F. Canadian L.F. Place of Birth Experience [Experience.sup.2] Canada 0.283 (*) -0.003 (*) Immigrants Educated in Canada Anglo- or Francophone Countries 0.233 (*) -0.002 Northwestern Europe 0.206 -0.006 Eastern Europe 0.174 -0.003 Southern Europe 0.204 (*) -0.003 Middle East and West Asia 0.377 -0.003 South Asia 0.055 -0.001 Hong Kong -0.216 0.024 P. Republica of China 0.136 0.003 Phillipines 0.094 -0.035 Viet Nam -0.254 -0.066 Other S/SE Asia 0.434 -0.024 African Countries (Nonwhites) 0.087 0.017 Carribean (Nonwhites) 0.301 -0.033 Latin America 0.207 -0.003 Canada 0.283 (*) -0.003 (*) Immigrants Educated Abroad Anglo- or Francophone Countries 0.082 (*) -0.001 Northwestern Europe 0.133 (*) -0.003 Eastern Europe 0.227 -0.01 Southern Europe 0.011 (*) 0.002 Middle East and West Asia 0.292 -0.003 South Asia 0.113 0.001 Hong Kong 0.012 -0.009 P. Republic of China 0.085 -0.002 Philippines -0.020 0.013 Viet Nam 0.040 0.007 Other S/SE Asia 0.085 0.006 African Countries (Nonwhites) 0.131 0.001 Carribean (Nonwhites) 0.066 0.009 Latin America 0.156 (*) -0.002 Foreign L.F. Foreign L.F. Place of Birth Experience [Experience.sup.2] Canada Immigrants Educated in Canada Anglo- or Francophone Countries Northwestern Europe Eastern Europe Southern Europe Middle East and West Asia South Asia Hong Kong P. Republica of China Phillipines Viet Nam Other S/SE Asia African Countries (Nonwhites) Carribean (Nonwhites) Latin America Canada Immigrants Educated Abroad Anglo- or Francophone Countries 0.057 0.003 Northwestern Europe 0.026 0.002 Eastern Europe -0.009 0.004 Southern Europe -0.064 0.000 Middle East and West Asia 0.067 0.000 South Asia 0.130 (*) -0.006 (*) Hong Kong 0.035 0.000 P. Republic of China -0.065 0.000 Philippines 0.001 0.001 Viet Nam -0.196 (*) 0.005 Other S/SE Asia 0.054 0.003 African Countries (Nonwhites) -0.042 0.009 Carribean (Nonwhites) -0.055 -0.001 Latin America -0.041 0.001 Weak Language Place of Birth Proficience Intercept [R.sup.2] Canada 0.657 9.36 0.327 Immigrants Educated in Canada Anglo- or Francophone Countries 3.39 -0.576 0.365 Northwestern Europe -0.267 15.98 0.332 Eastern Europe -0.496 1.00 0.404 Southern Europe -2.06 (*) 14.30 0.323 Middle East and West Asia -2.72 -4.17 0.438 South Asia -5.96 (*) -0.577 0.401 Hong Kong -0.574 -11.48 0.384 P. Republica of China 1.18 23.58 0.352 Phillipines -1.61 -4.13 0.205 Viet Nam -2.282 -1.35 0.544 Other S/SE Asia 0.817 -29.53 0.407 African Countries (Nonwhites) 0.922 17.06 0.281 Carribean (Nonwhites) -7.41 -0.919 0.336 Latin America -1.90 12.53 0.306 Canada 0.657 9.36 0.327 Immigrants Educated Abroad Anglo- or Francophone Countries -0.04 16.11 0.298 Northwestern Europe -1.71 22.70 0.272 Eastern Europe -1.53 18.01 0.242 Southern Europe -1.32 (*) 23.57 0.174 Middle East and West Asia -1.38 20.60 0.332 South Asia -4.80 (*) 23.39 0.420 Hong Kong -2.30 26.28 0.262 P. Republic of China -4.13 (*) 31.95 0.389 Philippines -1.54 19.95 0.188 Viet Nam 0.34 24.26 0.213 Other S/SE Asia -2.22 (*) 19.39 0.352 African Countries (Nonwhites) -2.52 (*) 16.14 0.333 Carribean (Nonwhites) -3.00 (*) 19.46 0.291 Latin America -2.94 (*) 19.04 0.322 Source: 1991 Census of Canada Public Use Microdata File Note: All Models also include controls for marital status, children present, other adults present, other adults present, citizenship, and size of place. (*)p<0.001 (+)Coefficient significantly different from corresponding Canadian-born coefficient using BIC-based minimum t=3.68 (see Raftery, 1995: 140) on coefficients of models reported in Table 2. Table 4 Unstandardized OLS Regression of Earnings on Education, Labour Force Experience, and Official Language Proficiency by Place of Birth, Canadian Men Age 20-64 Employed Full-Time, 1991 Years of Years of Place of Birth Schooling Schooling (2) Canada 2094 (*) 105 (*) Immigrants Educated in Canada Anglo- or Francophone Countries 2759 (*) 36.4 Northwestern Europe 2424 (*) 3.09 Eastern Europe 1143 186 Southern Europe 1934 (*) 72.8 Middle East and West Asia 3313 -59.7 South Asia -1154 320 (*) Hong Kong 5818 (*) -201 P. Republic of China 3583 (*) 13.1 Phillipines 1427 65.3 Viet Nam 1431 94.7 Other S/SE Asia -343 394 African Countries (Nonwhites) -944 427 Carribean (Nonwhites) 371 235 Latin America 2005 (*) 14.3 Canada 2094 (*) 105 (*) Immigrants Educated Abroad Anglo- or Francophone Countries 2444 (*,+) 89.6 (*,+) Northwestern Europe 741 (+) 221 (*) Eastern Europe 1322 (*,+) -4.01 Southern Europe 845 (*,+) 136 (*) Middle East and West Asia 912 (*,+) 122 (*) South Asia 821 (*,+) 129 (*) Hong Kong 989 (+) 177 (*) P. Republic of China 912 (*,+) 122 (*) Philippines 660 (+) 36.0 Viet Nam 419 (*,+) 126 (*) Other S/SE Asia 1561 (*,+) 69.5 African Countries (Nonwhites) 874 (+) 170 (*) Carribean (Nonwhites) 1267 (*,+) 192 (*,+) Latin America 721 (*,+) 217 (*) Canadian L.F. Canadian L.F. Place of Birth Experience Experience (2) Canada 1043 (*) -19.9 (*) Immigrants Educated in Canada Anglo- or Francophone Countries 1269 (*) -23.9 (*) Northwestern Europe 1178 (*) -24.9 (*) Eastern Europe 943 (*) -24.2 (*) Southern Europe 1023 (*) -18.8 (*) Middle East and West Asia 1408 (*) -44.3 South Asia 889 (*) 1.48 Hong Kong 1476 (*) -64.5 P. Republic of China 823 (*) -22.9 Phillipines 807 (*) -43.1 Viet Nam 783 (*) -63.8 Other S/SE Asia 1218 (*) 25.4 African Countries (Nonwhites) 1445 (*) 55.0 Carribean (Nonwhites) 925 (*) -18.9 Latin America 1259 (*) -29.4 (*) Canada 1043 (*) -19.9 (*) Immigrants Educated Abroad Anglo- or Francophone Countries 381 (*) -4.33 Northwestern Europe 660 (*) -18.4 (*) Eastern Europe 726 (*) -14.1 (*) Southern Europe 314 (*) -6.12 (*) Middle East and West Asia 816 (*) -25.5 South Asia 749 (*) -11.2 Hong Kong 842 (*) -20.2 P. Republic of China 624 (*) -25.7 (*) Philippines 610 (*) 6.67 Viet Nam 518 (*) 51.5 Other S/SE Asia 608 (*) 3.36 African Countries (Nonwhites) 853 (*) -12.5 Carribean (Nonwhites) 438 (*) 7.90 Latin America 735 (*) -16.3 (*) Foreign L.F. Foreign L.F. Place of Birth Experience Experience (2) Canada Immigrants Educated in Canada Anglo- or Francophone Countries Northwestern Europe Eastern Europe Southern Europe Middle East and West Asia South Asia Hong Kong P. Republic of China Phillipines Viet Nam Other S/SE Asia African Countries (Nonwhites) Carribean (Nonwhites) Latin America Canada Immigrants Educated Abroad Anglo- or Francophone Countries 212 (*) -1.50 Northwestern Europe -164 4.17 Eastern Europe -16.3 1.19 Southern Europe -98.2 0.00 Middle East and West Asia -67.0 -1.86 South Asia 65.6 -1.90 Hong Kong 337 -20.3 P. Republic of China -362 (*) 9.89 Philippines 185 -5.28 Viet Nam -264 (*) 2.48 Other S/SE Asia 40.6 8.07 African Countries (Nonwhites) -217 -3.11 Carribean (Nonwhites) 136 -4.36 Latin America 67.5 -2.34 Weak Language Place of Birth Proficiency Intercept [R.sup.2] Canada -2762 -34023 0.363 Immigrants Educated in Canada Anglo- or Francophone Countries 842 -46159 0.417 Northwestern Europe 5017 -49353 0.352 Eastern Europe -3856 -32873 0.363 Southern Europe -3928 (*) -34669 0.332 Middle East and West Asia -5169 (*) -56157 0.462 South Asia -509 42322 0.543 Hong Kong -2417 -85961 0.353 P. Republic of China -4791 -31021 0.489 Phillipines -1549 -27172 0.547 Viet Nam 956 -27539 0.677 Other S/SE Asia -2872 67270 .490 African Countries (Nonwhites) 2366 -7781 0.533 Carribean (Nonwhites) -7225 -15598 0.547 Latin America -1059 -36471 0.417 Canada -2762 -34023 0.363 Immigrants Educated Abroad Anglo- or Francophone Countries -6813 -38975 0.228 Northwestern Europe -561 -14705 0.216 Eastern Europe -1783 -18207 0.368 Southern Europe -1832 (*) -5712 0.184 Middle East and West Asia -2988 -17669 0.408 South Asia -3856 (*) -13663 0.422 Hong Kong -3549 -20320 0.334 P. Republic of China -10892 (*) -112 0.337 Philippines -2311 (*) -9416 0.389 Viet Nam -67 -7374 0.298 Other S/SE Asia -3442 -24338 0.276 African Countries (Nonwhites) -4662 (*) -16603 0.416 Carribean (Nonwhites) -5648 (*) -20806 0.382 Latin America -4473 (*) -15140 0.389 Source: 1991 Census of Canada Public Use Microdata File. Note: All models also include controls for marital status, children present, other adults present, citizenship, size of place and weeks worked. (*)a<0.01 (+)Coefficient significantly different from corresponding Canadian-born coefficient using BIC-based minimum t=3.68 (see Raftery, 1995: 140) for models reported in Table 2. Table 5 Macro Level Models Predicting Effects of Graded and Postsecondary Education Acquired Abroad on Occupational Status and Earnings from Charcteristics of Country of Birth and Comfort of Native-Born with Immigrant Group, Canada, 1991 Effects on Effects on Occupational Status Earnings Graded Postsecondary Graded GNP Per Capita 0.050 (*) -0.013 40.5 (US$000s) Percent Enrolled -0.015 (*) 0.005 -13.41 in Secondary Education Comfort with -0.014 0.019 18.9 Immigrant Group Comfort Value 0.494 -0.233 645 Missing Constant 2.38 1.10 -581 [R.sup.2] 0.356 0.118 0.152 Effects on Earnings Postsecondary GNP Per Capita 125 (**) (US$000s) Percent Enrolled -21.4 in Secondary Education Comfort with 55.9 Immigrant Group Comfort Value -81.1 Missing Constant -1306 [R.sup.2] 0.484 (**) Note: Coefficients are OLS estimates. N equals 25 places of birth for all models (see Table 1) (*)p<005; (**)p<001
RELATED ARTICLE: APPENDIX
Measurement of Variables in Individual-Level Models
1. Earnings: Gross wages and salaries from employment earned in 1990 to the nearest dollar expressed in Canadian dollars.
2. Occupational Status: Socioeconomic status of respondent's occupation expressed as score on the International Standard Socio-Economic Index of Occupational Status (Ganzeboom et al., 1992; Ganzeboom and Treiman, 1996). Since the 1991 PUMF reports only a collapsed version of the 1991 National Occupational Classification, following Model (1998), scores were assigned to these categories based on weighted mean ISEI scores for the detailed occupations comprising the categories as follows: Senior Managers, 68.000; Middle Managers, 53.479; Professionals, 65.850; Semi-professional and Technical, 50.463; Supervisors, 49.939; Foremen/women, 45.211; Administrative and Senior Clerical, 53.828; Sales and Service (Skill Level III), 41.740; Skilled Crafts/Trades, 34.121; Clerical Workers (Skill Level II), 45.000; Sales and Service (Skill Level II), 38.976; Semi-skilled Manual, 31.311; Sales and Service (Skill Level I), 30.504; Other Manual Workers (Skill Level I), 19.539.
Human Capital Endowments:
3. Years of Schooling. Measured by summing an item indicating highest grade of primary or secondary schooling and items indicating highest level of postsecondary education, for which categories reflecting some nonuniversity postsecondary, some university, a bachelor's degree, some postgraduate education, a master's degree, or an earned doctorate were recoded into years.
4. Years of Schooling, Squared. Equals years of schooling minus ten squared, to reduce computational accuracies due to rounding error. Captures the additional effect of higher levels of schooling.
5. Canadian Labour Force Experience. Number of years a respondent could potentially have been in the Canadian labour force, measured as age minus number of years of schooling plus five for men educated in Canada, and age at time of survey minus age at immigration to Canada for men educated abroad.
6. Canadian Labour Force Experience, Squared. Equals Canadian labour force experience minus ten squared, to reduce computational accuracies due to rounding error. Indexes declining marginal returns to Canadian experience.
7. Foreign Labour Force Experience. Number of years a respondent could potentially have been in the labour force of country of origin or an intermediate country, set to zero for those born in Canada and immigrants educated in Canada. For immigrants educated abroad, it equals age of immigration minus age at completion of education.
8. Foreign Labour Force Experience, Squared. Equals years of foreign labour force experience minus ten squared, to reduce computational inaccuracies due to rounding error. Indexes declining marginal returns to foreign experience.
9. Weak Official Language Proficiency. Since both French and English are official languages in Canada, with French being predominant in Quebec, we consider proficiency in either or both to constitute language capital, though French is of less use in the labour market outside Quebec. Coded '1' if the respondent either cannot carry on a conversation in either official language or uses a language other than English or French at home, coded '0' if English or French are the respondent's mother tongue or if an official language is used at home.
1. Weeks Worked in 1990. Number of weeks during 1990 during which respondent worked, even if only a few hours.
2. Citizenship. A dummy variable coded '1' if respondent is a Canadian citizen, '0' otherwise.
3. Size of Place. Two dummy variables, one coded '1' if respondent resides in a medium-sized Census Metropolitan Area (CMA), '0' otherwise, and another coded '1' if respondent does not live in a CMA, '0' otherwise. Residing in a large CMA (i.e. Montreal, Toronto, or Vancouver) represents the baseline category.
4. Marital Status. A dummy variable coded '1' if respondent is married, '0' otherwise.
5. Children Present. A dummy variable coded '1' if at least one never-married child lives in the household, '0' otherwise.
6. Other Adults Present. A dummy variable coded '1' if at least one adult other than the respondent and wife lives in the household, '0' otherwise.…