Main article

View the most recent version.

Archived Content

Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.

Archived

This page has been archived on the Web.

1. Introduction
2. Data and definitions
3. A profile of new immigrants in Canada and the U.S.
4. Do changes in source country composition account for the changing labour market performance of university-educated immigrants?
5. Outcomes after 11 to 15 years in host country
6. Why did the wage gap between immigrants and the domestic-born increase in Canada and fall in the U.S. over the 1990s?
7. Conclusion

1   Introduction

While it is well known that entry earnings have been declining for successive cohorts of entering immigrants, perhaps less well known is the fact that these relative (to domestic-born) declines in entry earnings were larger for university graduates than for high school graduates between the early 1980s and mid-1990s (Green and Worswick 2010). Furthermore, relative and absolute low-income rates have been increasing faster among highly educated immigrants than the less educated (Picot and Hou, 2003). If economic outcomes of highly educated immigrants are poorer in Canada than in other major immigrant receiving countries, this could adversely affect the willingness of highly skilled workers to move to Canada. University-educated immigrants are becoming increasingly important to most immigrant-receiving, developed countries, including Canada. The goal of this study is to compare the labour market performance of university-educated new immigrants (relative to their domestic-born counterparts) in Canada and one of its main competitors for skilled labour, the United States.

Previous comparative studies have analyzed cross-country differences in language fluency, education, and labour market outcomes of immigrants in light of differences in immigration systems. For example, Duleep and Regets (1992) and Borjas (1993) compare Canada and the U.S., and Antecol, Cobb-Clark, Trejo (2003) compare Canada, Australia and the United States. These studies focus on whether an immigration system that selects newcomers based on skills alters the skill composition of immigrants from a given source country. The answer appears to be that immigrants from the same country of origin possess similar skills regardless of their destination, at least among immigrants who arrived before 1990. The study closest in spirit to ours, Antecol, Kuhn, and Trejo (2006), compares the change over time in the gap between the domestic-born and immigrant populations with respect to employment and wages in Canada, Australia, and the U.S. in light of the different labour market institutions in the three countries. Generally speaking, the study finds that new immigrants improve their economic outcomes over time more in terms of wage adjustment in the U.S. than in the other two countries, and more in terms of employment in Australia than in the other two countries, with Canada falling in the middle in both cases.

There is currently little evidence on economic outcomes of university-educated immigrants specifically. The exceptions include Picot and Hou (2003), Green and Worswick (2010), and Frenette and Morissette (2005) for Canada and Borjas and Friedberg (2007) for the U.S. To our knowledge there are no cross-country studies that compare the economic success of university-educated immigrants despite the fact that highly educated immigrants are becoming increasingly important to major immigrant receiving countries. This study fills this gap in the literature by comparing cross-cohort patterns in entry earnings, defined as earnings during the first five years after arrival, of university-educated immigrants in both Canada and the U.S. The analysis covers the period from 1980 to the year for which the most recent data are available, namely, the 2006 Census of Population in Canada and the 2005 American Community Survey in the U.S.

There are a number of reasons why immigrant entry earnings are an important metric. First, the significant decline in earnings at entry over the 1980s and early 1990s (likely the major topic addressed in the immigrant economics literature over the last two decades) was followed by a significant change in the characteristics of entering immigrants in Canada, notably an increase in the share with university degrees. It seems reasonable to focus on entry earnings to assess the effects of these changes in immigrant characteristics. Second, recent research on return migration (Aydemir and Robinson 2008) has suggested that the rate of out-migration of immigrants in Canada is large, and occurs primarily during the first couple of years following entry. Hence, economic outcomes during the first few years after migration may be an important factor in determining the extent to which Canada retains its immigrants. Third, most immigrants who fall into low-income status do so during their first full year in Canada (between 35% and 45% of immigrants are in low-income status during their first year in Canada), and this is followed by fairly high rates (around 20%) of remaining in low-income status for five or more years (Picot, Hou, and Coulombe 2008). Lower entry-level earnings have been followed, by and large, by less favourable economic outcomes during at least the first decade or so.

The goal of this study is to determine whether highly educated new immigrants to Canada have fared as well economically as their counterparts entering the U.S., in light of the significant rise in the number of highly educated immigrants entering Canada. This study asks how economic outcomes at entry for the highly skilled have changed in the two countries over the last quarter-century (1980 to 2005). It further asks whether changes in the standard observable background characteristics of entering immigrants can account for any differences in the outcome trends observed between Canada and the U.S. Two economic outcome measures are used: the mean relative (to domestic-born) entry wages of highly educated new immigrants (i.e., the wage gap at entry) and the university wage premium, defined as the difference between the wages of the university-educated and those of the high-school-educated. Both unadjusted and adjusted (controlling for changes in observable characteristics across successive cohorts) estimates of these outcomes are produced.

This study finds that relative entry earnings of university-educated immigrants followed a significantly different path in Canada and the U.S., with generally superior outcomes in the U.S., particularly since 1990. This occurred despite the fact that significant declines in entry earnings for successive groups of entering immigrants as a whole (combining those with and without university education) were observed in both countries over the last quarter-century. Changes in the composition of new immigrants with respect to age, language spoken at home, English language ability (English or French in Canada), source country, and region of residence, which tended to be greater in Canada than in the U.S., accounted for most of the observed change in relative earnings of university-educated immigrants in Canada during the 1980s, but this is less true for more recent cohorts. Compositional shifts had a smaller negative effect on aggregate relative earnings of university-educated immigrants in the U.S. The university wage premium increased marginally among new immigrants in Canada between 1980 and 2000, but fell between 2000 and 2005, especially for men. In the U.S., the university wage premium rose quite rapidly over the 1980-2005 period among both new immigrants and domestic-born workers, both men and women.

2   Data and definitions

This paper uses data from the 1981, 1991, 2001, and 2006 censuses of Canada, each representing a 20% sample of the Canadian population. The U.S. data for comparable years come from the Integrated Public Use Microdata Sample (IPUMS) of the 1980, 1990, and 2000 U.S. censuses, each representing a 5% sample of the population, and from the 2005 American Community Survey (ACS), a 1% sample of the population (Ruggles et al. 2008). ¹  The sample consists of individuals aged 25 to 54 living in private dwellings and working in civilian occupations. ²  The sample excludes individuals with Aboriginal ancestry from the domestic-born sample of both countries. ³ 

Immigrants are defined as foreign-born individuals who are not citizens by birth of the host country. The sample excludes temporary residents in Canada because they were not enumerated in the 1981 Census. The immigrant sample is restricted to individuals who arrived at or after the age of 25. This is in order to exclude immigrants who arrived as children or youth, given that their labour market experiences differ markedly from those of individuals who arrived as adults and likely completed all or some of their education before arrival.

The aim is to focus on entry wages of immigrants, i.e., the wages of "recent" or "new" immigrants. To identify immigrants who have lived in the host country for no longer than five years, the question on the year in which a person became a landed immigrant on the Canadian census, and the year in which a person came to live or stay on the U.S. census is used. ⁴  These questions are traditionally used to estimate the number of years that an immigrant has lived in the host country. They do not capture exactly the same information in both countries, however. The question on the Canadian census refers to a concrete event - the year in which an individual became a landed immigrant. The equivalent event in the U.S. would be the year in which an individual becomes a permanent resident (i.e. obtains what is commonly called a "green card"), but the wording of the question on the U.S. census and ACS appears to be more ambiguous, so that some individuals who have lived in the U.S. before obtaining permanent legal status might report some earlier year as their "year of arrival." ⁵ 

Hence, the "years in Canada (or the U.S.)" variable may not be accurately reported in the two censuses, and not comparably reported between the two countries. This is important because there is a positive correlation between "years in the country" and earnings. It is known that some immigrants live in Canada before becoming landed immigrants. The Canadian census measures "years since becoming a landed immigrant," not "years in Canada." Therefore, "years in Canada" may be underestimated in the Canadian census, and hence earnings at any given number of "years in Canada" (which is really years since becoming a landed immigrant) overestimated in research based on census data (including this study).

The situation with respect to the U.S. census is less clear. Since the U.S. census asks when a person came to live or stay in the U.S., it is not measuring "years since obtaining a green card." Nor is it necessarily measuring "years since first coming to the U.S."; rather, it is likely measuring something in between. Hence, the underestimation of the true "years in the U.S." and the resultant overestimation of earnings at any given number of years in the U.S. (as measured by the U.S. census) may be significantly less than is the case in Canada. The data suggest that this is indeed the case.

Text table 1 in the Appendix shows that 20% of university-educated immigrant men who report their "year of arrival" (i.e., year of becoming a landed immigrant) as 1997 on the 2001 Canadian Census in fact lived in Canada in 1996. In comparison, 10% of university-educated immigrant men who "came to live" in the U.S. in 1996 (as reported in the U.S. census) already lived in the U.S. in 1995. This information is obtained by using the "where did you live 'X' years ago" questions contained in the Canadian and U.S. censuses. A larger proportion of Canadian than American immigrants lived in the country before the immigrant "arrival" date reported in their respective censuses. Further data analysis suggests that it is more common for individuals to live in the host country for one or two years rather than for a longer period of time before the "arrival date" reported on the census (obtaining landed immigrant status in Canada and arriving "to live" in the U.S.), as one would expect. Data from an entirely different data source in Canada produce very similar estimates for that country. ⁶ 

Text table 1 also shows that university-educated immigrants who lived in Canada prior to obtaining landed immigrant status (i.e., before their reported "arrival" date in the census) generally report slightly higher wages during the first five years in Canada on the census than those who did not. In contrast, university-educated immigrants who lived in the U.S. before the "year of arrival" generally report lower wages. Therefore if anything, as a result of the "error" in reporting the true date of arrival, wages of immigrants in Canada during the first five years in Canada are likely overestimated and wages of immigrants in the U.S. underestimated, at least on the basis of evidence in Text table 1 on immigrants who arrived in the 1990s. ⁷  Hence, the gap in outcomes between Canadian and American highly educated immigrants is likely underestimated, as the research shows that by 2005 outcomes were superior in the U.S. to those in Canada.

The sample of new immigrants consists of individuals who, at the time of the survey, had lived in Canada between 1 and 5 years, while, for the U.S., it is between 0 and 5 years (on the basis of the "year of arrival" reported in the census for both countries). The definitions of new immigrants in Canada and the U.S. for purposes of this study are slightly different because information on year of arrival and immigrant earnings are reported differently in the two countries. ⁸  In Canadian data, there are no earnings reported for immigrants who arrived in the census year (i.e., year 0). In contrast, about half of the immigrant men in the 2000 U.S. census who report having arrived in 2000 report non-zero earnings for the previous calendar year; not all individuals who arrive in the census year will therefore drop out of the sample of workers with positive weekly earnings automatically. On the other hand, in the 1980 and 1990 U.S. censuses, year of arrival information is available in typically 3-to-5-year brackets, and thus it is not possible to identify and exclude those who arrived in the census year. The slight difference in the definition of new immigrants may result in a further underestimating of the relative performance of new immigrants in the U.S. relative to their Canadian counterparts since immigrant wages tend to grow with years since arrival.

The main outcome variable is weekly wages of paid workers. Paid workers are defined as individuals with positive earnings in the reference year who reported working a positive number of weeks in the reference year and who made more income from wages and salaries than from self-employment. Weekly wages reported in this study are in 2000 constant Canadian dollars for Canada and in U.S. constant dollars for the U.S. They are not adjusted for purchasing power parity, and hence their levels should not be compared across the two countries.

For comparing relative wages of university-educated immigrants and the university wage premium, i.e., the wage differentials between university and high school graduates, it would be ideal to have education categories that are both consistent across time within each country and fairly comparable between the two countries. This proves to be challenging as a result of differences in the education question contained in the Canadian and U.S. censuses, and a change in the question over time in each country (between 1980 and 1990 in the U.S., and between 2001 and 2006 in Canada). The education classification chosen for this study allows for the most consistency across time and within a country. ⁹  For each country, four education groups are defined for denoting a person's highest educational attainment: less than high school; high school graduate; some post-secondary; and university graduate. ¹⁰  The university graduate and high school graduate categories are consistent within each country. The university graduate category is also reasonably comparable between the two countries (although there may be further differences in the relative distribution of undergraduate degrees and graduate degrees for example), but the high school graduate category is not. ¹¹ 

3   A profile of new immigrants in Canada and the U.S.

3.1  Socio-demographic characteristics of new immigrants of all education levels

New immigrants to Canada (Table 1) and the U.S. (Table 2) are quite dissimilar. Notably, in recent years, new immigrants to Canada tend to be older, more highly educated, and originate from quite different source regions than their American counterparts. The following observations refer to recent immigrants (in the host country five years or less) aged 25 to 54, with or without earnings in the reference year.

New immigrants in Canada became increasingly older than their U.S. counterparts. The average age of new immigrants remained stable in the U.S. at 36.5 years, but increased in Canada from around 36 to 38. The Canadian points system of immigration assigns the maximum amount of points for age to individuals who are between 21 and 49. It also assigns points for foreign work experience. This may be one reason for the change in the age structure across cohorts of immigrants entering Canada (recall though that the sample of recent immigrants includes only individuals who arrived at age 25 or older). In both countries, the average age of new immigrants is lower than the average age of the domestic-born population. Between 1981 and 2006 the average age of the domestic-born increased similarly in both countries, reaching about 40 years by 2006.

There was also a dramatic increase in the proportion of new immigrants to Canada who held a university degree; this was not the case in the U.S. In 1981, new immigrants in the U.S. had higher university completion rates than new immigrants in Canada; by 2006, the opposite was true. The Canadian government altered its selection system in the earlier 1990s to increase immigrants' educational levels and the share of economic immigrants. The educational level of new immigrants to Canada rose dramatically in the 1990s. Almost 60% of prime-age immigrant men (51% of women) who entered Canada between 2001 and 2005 had a university degree, compared to 35% of immigrant men (36% of women) who entered the U.S. during that period. This stands in stark contrast to the cohort of immigrants from 25 years earlier: 26.2% of immigrant men (16.4% of women) who entered Canada at that time had a university education. In the U.S., there was little change in the share of male immigrants with degrees (although the share of female immigrants who had a degree rose).

The source country composition of legal immigrants is also both quite different, and has changed in different ways, in both countries. Over the last two-and-a-half decades, the shift in source regions in Canada was characterized by a decline in the share of immigrants from Europe and the U.S. and a large increase in the share of immigrants from Asia in particular (from where 56% now come). Put differently, new immigrants from developed economies and countries with cultures similar to that of Canada have been replaced by immigrants from developing countries. In the U.S., the shift in source regions was marked by the large increase in the share of immigrants from South and Central America (from which almost half of legal immigrants now come) and a decrease in the share of immigrants from Asia. Changes in source country composition of immigrants have been linked in the literature to the declining entry earnings of successive immigrant cohorts in Canada (e.g., Green and Worswick, 2010; Aydemir and Skuterud 2005).

Differences in source country composition will also contribute to differences in language proficiency of new immigrants in the two countries. Using available data, one can construct a comparable measure of the proportion of people who speak a non-official language at home (i.e., not English in the U.S. and neither English nor French in Canada) ¹²  for the two countries. That proportion has risen from around 50% among new immigrants to Canada in the late 1970s to just over 80% among new immigrants who have arrived since 2001. This is lower than the proportion in the U.S., which remained above 80% across all cohorts, and was about 89% in 2006.

3.2  Socio-demographic characteristics of university-educated new immigrants aged 25 to 54

The increased focus on attracting university graduates to Canada resulted in a significant rise in the supply of such immigrants to Canada, with a significantly smaller corresponding increase in the U.S., particularly during the 1990s. Between 1991 and 2006, the number of university-educated new immigrants aged 25 to 54 rose from around 84,350 to 298,000 in Canada (a 253% increase), while rising from 531,390 to 1,230,300 in the U.S. (a 132% increase).

In addition, Canada experienced a major shift away from European countries in favour of Asian countries as a source of university-educated immigrants. For example, in 1981 Canada, 39% of new (i.e., entering between 1976 and 1980) university-educated men came from Asia; this figure had increased to 61% by 2006 (Table 3). In contrast, there was essentially no such shift in the U.S. The corresponding shares in the U.S. actually declined marginally, from 52% in 1980 to 50% in 2005. The share of entering male university-educated immigrants from Europe changed little in the U.S., falling from 22% to 18%, while in Canada falling much more from 30% to 18%. Hence, among university-educated new immigrants, there have been much larger changes in both volume and source region distribution in Canada than there have been in the U.S. over the last quarter-century.

The more dramatic change in source region in Canada is reflected in the change in language spoken at home. The relatively stable source region distribution in the U.S. produced little change in the share of new immigrants speaking a language other than English at home; it remained at a very high mid-80s percent over the entire quarter-century. In Canada, a much smaller share of university-educated new immigrants spoke a language other than English or French at home in 1981 (around 40%); by 2006 this had risen to around 85%. And since this paper is concerned with the change in labour market outcomes of university-educated new immigrants, this more dramatic change in language profiles in Canada may be significant. Interestingly, the measure of the ability to speak English (English or French in Canada) does not follow the same pattern. This variable probably does not fully capture individual variation in English (English or French) proficiency. ¹³ 

The differences between Canada and the U.S. in the change in the distribution of background characteristics can obviously influence labour market outcomes. This issue is addressed in a later section, following a description of the unadjusted labour market outcomes.

3.3  Labour market outcomes of university-educated new immigrants

3.3.1  Relative wages

Relative entry earnings of university-educated immigrants have followed significantly different paths in the two countries over the period 1990 to 2005. The unadjusted (i.e., based on raw data) wage gap between university-educated new immigrants and university-educated domestic-born men widened much faster in Canada than in the U.S. (Chart 1); this is consistent with findings in Green and Worswick (2010) and Borjas and Friedberg (2007). While relative immigrant entry wages of university graduates were fairly similar among men in Canada and the U.S. in 1980 and 1990, the 1990s brought a dramatic divergence between the two countries. In 1980, the gap was -0.25 log points (i.e., entering immigrants earned roughly 25% less than the domestic-born) ¹⁴  in both Canada and the U.S. This gap expanded rapidly to -0.67 log points in Canada by the year 2005. In the U.S., in contrast, the gap grew until 1990 but had returned to the 1980 level by the year 2000. In 2005, it rose again, to roughly the 1990 level. ¹⁵ 

Immigrant women in Canada started out with a considerably larger disadvantage relative to their domestic-born counterparts in 1980, but the size of the gap was comparable to that among men for the remaining years (Chart 2). In the U.S., immigrant women's relative wages did not show as significant an improvement between 1990 and 2000 as they did for men.

Previous studies (Beaudry and Green 2000; Green and Worswick 2010) have found that young, Canadian-born workers have been experiencing declining labour market entry wages over the 1980s and 1990s. If labour market entry wages of domestic-born university-educated workers have followed a different path in Canada than in the U.S., it may be that much of the Canada-U.S. difference in wages observed in Charts 1 and 2 has more to do with conditions for university-educated new labour market entrants in general and not immigrants specifically. In the data used for this study, weekly wages of university-educated domestic-born workers aged 25 to 30 indeed have declined among Canadian-born men but have risen (at least between 1980 and 2000) for American men (results not reported). However, when wages of new immigrants are compared to those of domestic-born new labour market entrants (Charts 3 to 4), a cross-cohort, cross-country pattern resembling that in Charts 1 and 2 results, where domestic-born workers aged 25 to 54 were the comparison group. ¹⁶  In Canada, immigrant men (aged 25 to 54) experienced a growing disadvantage in wages relative to 25-to-30-year-old domestic-born university graduates. This was not the case in the U.S. Although labour market conditions for new entrants, whether domestic or foreign-born, may be partly responsible for the different outcomes of new immigrants to Canada and the U.S., this does not appear to be the main story.

The observed pattern in relative wages of university-educated new immigrants could be driven by rising absolute wages of the domestic-born workers and/or falling absolute wages of new immigrants. The nature of these underlying patterns would lead to different interpretations of the observed changes in relative wages of immigrants and the university premium they receive. Charts 5, 6, 7 and 8 address these questions. The falling relative earnings of university-educated new immigrants in Canada are driven by both declines in the absolute wages of successive immigrant cohorts and increases (since the 1990s) in the earnings of Canadian-born workers. In the U.S., earnings have been rising for both U.S.-born workers and entering university-educated immigrants between 1990 and 2000, with the latter group experiencing a more rapid increase than the former; this has resulted in rising relative wages over the 1990s. ¹⁷ 

3.3.2  The university wage premium

The university wage premium, the difference in wages between university graduates and high school graduates, has been rising among the general population in both Canada and the U.S. since the 1980s. In Canada, a recent study shows that the raw university/high school wage differential rose from 1980 to 2005 among men, most in the early 1980s and since 1995. Adjusting for experience, the authors find that the wage premium shows an overall positive trend over the whole period. The adjusted wage premium among women was stable from 1980 to 2000, but increased from 2000 to 2005 (Boudarbat, Lemieux, and Riddell 2010). In the U.S., the university/high school wage premium increased sharply in the 1980s, and continued to increase at a moderate pace in the 1990s for both men and white women (see review by Deere and Vesovic, 2006; and Peracchi 2006). From at least 1990, the university premium has also been higher in the U.S. than Canada. Murphy, Riddell, and Romer (1998) explored this U.S.-Canada difference and concluded that the more rapid increase in the relative supply of the university-educated in Canada was the main determinant of the difference in university wage premium trends between the two countries. In the data, in 1981, some 14% of Canadian-born aged 25 to 30 held a university degree. This share had risen to 27% by 2006. In the U.S., the corresponding numbers were 23% in 1980 and 30% in 2005. The difference between the two countries was even more pronounced among new immigrants aged 25 to 54; the share of university graduates in Canada increased from 25% to 55% between 1981 and 2006 and from 30% to 35% in the U.S.

Even though there were large wage gaps between university-educated new immigrants and the domestic-born in both countries, for most of the last quarter-century, new immigrant and domestic-born men have received a similar university premium (a university premium is defined as the difference between mean log weekly wages of university and high school graduates). ¹⁸  However, new immigrants to Canada and the U.S. experienced opposite trends with respect to the university wage premium (Charts 9 and 10). Basically, the university wage premium has risen much faster in the U.S. than in Canada over the last quarter-century, for both immigrants and the domestic-born. Furthermore, there was a marked decline in the premium in Canada for immigrants following 2000, which was not observed in the U.S.

More specifically, new immigrants and the domestic-born had very similar wage premiums in Canada in 1980 and 1990, but divergent patterns emerged after 1990. The premium for domestic-born men increased considerably, from 0.29 log points in 1990 to 0.40 in 2000, and remained unchanged in 2006. The immigrant premium essentially stagnated between 1990 and 2000 and dropped from 0.31 log points in 2000 to 0.20 log points in 2005. Picot and Hou (2009) conclude that much of this decline since 2000 is associated with the downturn in the technology sector and with the fact that a very high proportion of entering immigrants were high-tech professionals or engineers.

Among women, the university premium is much lower among immigrants than among the domestic-born in both countries (Chart 11-12). This result stems from the much higher wage premium afforded domestic-born women than men in both countries. This same relative gender advantage is not observed among immigrants. The reason for this is unknown.

In the U.S., the unadjusted premium to a university education has been rising from 1980 to 2005 for both immigrant and U.S.-born workers, both men and women, and rising faster than it did in Canada until 2000. For example, among immigrant men the wage premium rose from 0.34 log points (roughly 34%) in 1980 to 0.71 points in 2005.

The university premium could alternatively be defined as the difference in wages between those with a high school diploma and those with a bachelor's degree. This would abstract from any differences in the share of workers with graduate degrees among those with at least one university degree and/or in returns to a graduate degree versus a bachelor's degree, both between immigrants and the domestic-born and between the two countries. The unadjusted premium, therefore, was recalculated restricting the university graduates' sample to those with a bachelor's degree as the highest educational attainment. For men, the premium was lower for both the domestic-born and new immigrants; however, the trend remained essentially unchanged. ¹⁹ 

The large drop in the university wage premium of new immigrant men who arrived in Canada between 2001 and 2005 was due to a larger drop in wages of university-educated immigrants than in those of high school graduates (see Charts 5, 6, 7 and 8). ²⁰  Earnings of immigrant women in both education groups fell during this period as well, with a larger drop among high school graduates than university graduates. In the U.S., the rising skill premium was driven by rising wages of university graduates. This pattern was true of both U.S.-born and immigrant workers alike until 2000. In 2005, wages of new immigrants in both education groups fell. The patterns were similar for men and women.

The raw data in Charts 5, 6, 7 and 8 are also consistent with the notion that, at least during the first few years in Canada, entering immigrants are competing to a considerable extent in the same labour market as lower-skilled Canadians. University-educated immigrants in Canada, both men and women, start out with average wages similar to those of domestic-born high school graduates. In the U.S., by contrast, especially since the 1990s, university-educated immigrants have earned wages that have considerably exceeded those of domestic-born high school graduates.

4   Do changes in source country composition account for the changing labour market performance of university-educated immigrants?

Existing research on declining entry earnings of immigrants in Canada identifies several determinants. The shift in source country composition away from European countries is an important explanation (Aydemir and Skuterud 2005; Bloom, Grenier, and Gunderson 1995; Picot and Hou 2003). Immigrants from non-European, non-English speaking countries may have difficulties in the Canadian labour market as a result of insufficient host country language skills, the quality of their foreign-acquired education, cultural differences, limited local networks, and potentially discrimination (Bloom, Grenier, and Gunderson 1995; Sweetman 2004). Changing source country composition of successive immigrant flows is also cited as a major contributor to declining immigrant earnings in the U.S. (Borjas 1992). This section examines the extent to which compositional changes account for the Canada-U.S. difference in the trends in relative wages of university-educated new immigrants and the university wage premium.

4.1  Approach

The following model of log weekly wages is estimated on a pooled sample of domestic-born and new immigrant paid workers, separately by gender and country (where the individual-specific index is suppressed for simplicity):

where the D superscripts refer to parameters pertaining to domestic-born workers, the I superscripts refers to immigrant workers, the subscript t refers to the census year, IM is a dummy variable equal to one for immigrants and zero otherwise, and S_j represents dummy variables for four education groups (less than high school; high school graduate; non-university post-secondary; and university graduate). The slope of the age profiles is allowed to vary by year and immigrant status, but the curvature of the age profile (Age²) is restricted to be time invariant and common to both groups. The vector X includes an indicator for a non-official language being spoken at home, an indicator for whether the respondent speaks English (English or French in Canada), a set of indicators for region of origin, and a set of indicators for region of residence (for Canada, the regions are the provinces and territories, with the Atlantic provinces grouped together and the territories grouped together, excluding Vancouver, Toronto, and Montreal, which are controlled for separately; for the U.S., the regions are the individual states). The estimated coefficients, reported in Appendix Text table 4, are used to obtain adjusted wage gaps between university-educated new immigrants and university-educated domestic-born as well as the university premium for domestic-born and new immigrant workers.

4.2  Adjusted relative wages among the university-educated

The U.S.-Canada divergence in the relative wages of university-educated new immigrants persists after controlling for age, non-official language spoken at home, ability to speak English (English or French in Canada), region of origin, and region of residence (see Table 4). The principal difference between Canada and the U.S. in entry earnings trends occurred during the 1990s - the entry earnings gap (relative to the earnings of the domestic-born) among the university-educated declined in the U.S. over the decade, while rising in Canada. This cross-country difference is observed both in the raw data and after controlling for changes in characteristics (i.e., the adjusted results).

In 1990, among men, the adjusted entry wage gaps were very similar in the two countries. For example, recent immigrant men had adjusted wages 0.44 log points (roughly 44%) lower than the average wage of domestic-born university graduates in Canada and 0.40 log points lower than the average wage of domestic-born university graduates in the U.S. This disadvantage rose to 0.51 log points in 2000 in Canada, but declined to 0.28 log points in the U.S. The entry wage gap rose in both countries between 2000 and 2005, but the difference between countries was very large by 2005, 0.35 log points in the U.S. and 0.63 log points in Canada. Over the entire quarter-century, the adjusted wage gap between university-educated new immigrant and domestic-born men in the U.S. barely changed, moving from -0.32 log points in 1980 to -0.35 in 2005. In contrast, the corresponding gap expanded from -0.37 to -0.63 log points in Canada.

Among university-educated female workers in Canada, the adjusted wage gap between new immigrants and the domestic-born was consistently higher than among men and increased between 1990 and 2005. ²¹  In the U.S., the adjusted gap among women fluctuated over time in a way similar to that for men, but the gap has been larger among women than among men since 1990, as in Canada.

While changes in characteristics account for much of the change in relative wages of university-educated immigrants in Canada between 1980 and 1990, they explain far less of the change in later periods in Canada and over the entire period in the U.S. The findings are qualitatively similar for men and women. For the sake of brevity, the discussion that follows focuses on men, though the trends also apply to immigrant women.

Between 1980 and 2005, the change in the composition of entering highly skilled male immigrants resulted in a 0.16 log point (roughly 16 percentage points) decline in relative wages, whereas, in the U.S., it resulted in only a 0.09 log point decline. This is not surprising, since the composition of immigrants by source region and language changed much more in Canada than in the U.S. This larger effect in Canada was concentrated in the 1980s. Among men, nearly two-thirds of the decline in relative wages of university-educated new immigrants in Canada between 1980 and 1990 can be attributed to changes in characteristics (0.11 of the 0.18 log points decline in relative earnings; see Table 4). The effect was much smaller in the 1990s and early 2000s. Nonetheless, even after accounting for the greater negative effect of compositional changes on relative earnings in Canada than in the U.S., the poorer outcomes in Canada persist. After controlling for compositional shifts, relative earnings fell 0.26 log points (roughly 26 percentage points) in Canada between 1980 and 2005, and only 0.03 log points in the U.S. Clearly factors other than compositional shifts played a role in the increased gap. Section VI returns to this point.

As noted earlier, the rise in the entry wage gap has been much more prominent among the university-educated than the less educated in both Canada and the U.S., especially in Canada. This is particularly true after controlling for changes in observable characteristics (Table 5). In fact, after conditioning on standard socio-demographic characteristics, the relative wages of high-school-educated immigrant men do not exhibit much long-term decline, although there is some decline in relative wages of immigrant women.

Two additional issues were considered. First, the sample was restricted to residents of metropolitan areas, where the majority of immigrants are located. While the gap in wages between immigrants and domestic-born workers is somewhat larger in cities, the trends are basically the same, in both Canada and the U.S. ²²  Second, immigrants were separated into those originating from traditional and non-traditional source countries; the difference in trends in relative immigrant wages between Canada and the U.S. was observed for both types of immigrants. ²³ 

4.3  Adjusted university wage premium

There are two major observations regarding the university wage premium. First, even though university-educated immigrants earn significantly less than their domestic-born counterparts for the first many years in the host country, for most of the last quarter-century, new immigrant and domestic-born men have had a similar university premium, i.e., the difference in earnings between university graduates and high school graduates (Table 6). Over the entire period, there was an economic advantage to having a university degree (on average), and this relative advantage was similar for immigrants and domestic-born men, at least until post-2000 in Canada, when it fell significantly among immigrant men. Among women, the striking fact is that the domestic-born have a much higher premium than their immigrant counterparts.

Second, the premium has risen much faster in the U.S. than in Canada, for both immigrants and the domestic-born. The university wage premium was not that different between the two countries in 1980 (although marginally higher in the U.S. for males). By 2005, however, the university wage premium was 0.40 log points (roughly 40 percent, see footnote 14) in Canada compared to 0.65 log points in the U.S. among domestic-born men, and 0.20 log points in Canada compared to 0.71 log points in the U.S. among new immigrants. The differences were in the same direction for women, but smaller.

Adjusting for standard socio-demographic characteristics did not affect the patterns or the magnitude of the university premium for immigrants in Canada. It did, however, account for some of the rise in the immigrant university premium in the U.S. in more recent years.

The university premium for new immigrant men in Canada rose slowly until 2000, following which it fell significantly. Immigrant men who arrived in the early 2000s, however, faced the lowest university premium over the study period, at 0.19 log points. In the U.S., in contrast, the premium for immigrants continued to rise, reaching around 0.49 log points in 2005.

5   Outcomes after 11 to 15 years in host country

Thus far, this study has focused on immigrant entry earnings for reasons described at the start. The larger relative (to domestic-born) economic disadvantage faced by immigrants in Canada shortly after arrival warrants less concern if these same immigrants catch up to their domestic-born counterparts relatively quickly. This section presents the wage gap for synthetic cohorts ²⁴  of university-educated immigrants at entry and again ten years later (i.e., 1 to 5 years after arrival and 11 to 15 years after arrival). ²⁵  These results are meant to provide a general idea of differences in earnings trajectories between Canada and the U.S.

The decline in immigrant entry earnings is accompanied by a decline in relative (to domestic-born) earnings for successive cohorts after 11 to 15 years (Table 7). This is true for both countries. However, just as the entry earnings gap among the university-educated (both adjusted and unadjusted) is higher in Canada than in the U.S., so too is the earnings gap after 11 to 15 years in the country, at least for the unadjusted gap (results for the adjusted gap were not produced). For example, in Canada, after 11 to 15 years in the host country, the late 1960s cohort of immigrants earned nearly 90% of what their domestic-born counterparts made; this figure fell to 75% for the late 1980s cohort and 70% for the early 1990s cohort. In the U.S., the comparable percentages were 101%, 84%, and 88%. A similar pattern is observed among immigrant women in both countries. The larger relative economic disadvantage (relative to the domestic-born) faced by university-educated immigrants to Canada than the U.S. does not disappear after 11 to 15 years in the host country.

6   Why did the wage gap between immigrants and the domestic-born increase in Canada and fall in the U.S. over the 1990s?

During the 1990s, the wage gap between immigrants and the domestic-born fell among highly educated entering immigrants in the U.S., but rose in Canada. Little of this divergence was associated with differential trends in the composition of entering immigrants, as demonstrated earlier. This is a significant puzzle.

First, one must ask what brought about the improvement in earnings among highly educated entering immigrants in the United States. Borjas and Friedberg (2007) suggest that the improvement may have been related to shifts in the immigration category used and employment in the IT sector. They conclude that the bulk of the improvement in relative entry wages was concentrated among those employed in computer and engineering occupations, i.e. in jobs associated with the high-tech boom. They speculate that the improvement may have been linked to increased use of the H-1B temporary work visa, which brings immigrants into pre-arranged jobs in "specialty occupations." ²⁶ 

However, similar trends were occurring in Canada. The number of entering immigrants in IT and engineering occupations increased dramatically during the 1990s (Picot and Hou 2009), and, among this particular group, the relative (to domestic-born) wage gap was also reduced over the decade (Hou 2010), as in the U.S. Hence, differences in outcomes among IT and engineering workers between the two countries is likely not the major explanation, even though within each country it is an important phenomenon. To test this, IT workers and engineers were excluded from the samples, and the Canada-U.S. results changed little (results not reported).

The U.S. does use pre-arranged employment among immigrants (as opposed to temporary workers) more so than does Canada. About 12% of all legal immigrants (including children and spouses) entered the U.S. under a class with pre-arranged employment between 1995 and 2000 (U.S. Department of Homeland Security 2006). This group is much skewed towards the highly educated, so that a significant proportion of university-educated immigrants may enter under this category. This approach was used much less in Canada during the 1990s. However, the main concern lies in the change in outcomes over the 1990s, and hence there would have to have been a significant change in the relative use of employer-sponsored entry (which provides better initial labour market outcomes) between the two countries, and there is little evidence of that (aside from the H-1B temporary visa mentioned above).

7   Conclusion

Many immigrant receiving countries, including Canada, the U.S., Australia, New Zealand, and the U.K., are seeking and receiving highly educated immigrants. Canada has done very well in attracting such immigrants, and continues to do so under its "human capital" approach to immigrant selection. By the late 1990s, new immigrants to Canada were much more highly educated and came from very different source regions than their counterparts in the U.S. The pool of university-educated "new" immigrants has undergone considerable change over the last quarter-century in Canada. The share of total immigration that university-educated "new" immigrants comprise rose, and the supply increased significantly, particularly during the 1990s. The source region distribution was altered, away from Europe towards Asia. The share with a home language other than English or French rose dramatically.

The pool of university-educated "new" immigrants to the U.S., on the other hand, was marked by much more stability. Their share of total immigration changed little over time, the supply rose less sharply than it did in Canada, the source region distribution changed little, and the share with a home language other than English was also little changed (although it remained at very high levels).

Relative entry wages of the highly educated "new" immigrants have taken very different paths in the two countries over the last quarter-century and have generally been superior in the U.S. This was particularly evident during the 1990s, when the earnings gap at entry among the university-educated rose sharply in Canada, while falling in the U.S. The adjusted (controlling for standard background observable characteristics) entry earnings gap was very similar in the two countries in 1990, but, by 2005, was 0.35 log points (roughly 35%) in the U.S., compared to a much larger 0.63 log points in Canada. These general trends were evident in both the raw data and the adjusted results.

Changing background characteristics (age, source region, language spoken at home) had a smaller negative effect on relative earnings in the U.S. than in Canada. This is not surprising, given the relative stability of the characteristics distributions in that country. In Canada, changing characteristics accounted for a successively smaller share of the rise in the entry earnings gap as one moves from 1980 to 2005, accounting for little among the entering cohort of the early 2000s. This too is not surprising, since the changes in the source region and language distributions among the highly educated were less significant in the later periods.

Although wages during the first five years in the host country are much lower among university-educated "new" immigrants than among their domestic-born counterparts, the relative economic advantage of having a degree (compared to the high school educated) has remained very similar for immigrants and domestic-born (among males), in both countries. This held true until the early 2000s in Canada, when the university wage premium fell sharply among immigrant males. But, again, the trends in the university wage premium were very different in the two countries. Generally speaking, the university wage premium has increased much more in the U.S. than in Canada, for both domestic-born and "new" immigrants. For the population as a whole, these trend differences were noted in earlier studies and have been attributed to the more rapidly increasing supply of university graduates in Canada than the U.S. (see Murphy, Riddell, and Romer 1998).

Our focus, however, is on new immigrants. Among this group, the adjusted wage premium was not that dissimilar between the two countries in 1980 (identical for women, 30% higher for men in the U.S.). However, by 2000, the premium was dramatically higher in the U.S. than in Canada, 160% higher among males and 70% higher among females.

The reasons for this divergent path between the two countries in the economic outcomes of the highly educated new immigrants are not known, and are beyond the scope of this paper. This study does show that controlling for the standard observable background characteristics accounted for only part of the difference; the increasing gap in outcomes between the two countries persisted (although at a reduced level) after accounting for compositional shifts over the 25 years. Different levels of economic success of immigrants trained in information technology and engineering fields does not appear to contribute much to explaining the Canada-U.S. difference in trends in relative earnings of new immigrants, and neither does the different degree to which the two countries rely on employment-based immigration categories. There are a host of other possibilities that are potential topics for future research, including:

1. The more rapid increase in the supply of the highly educated new immigrants in Canada than in the U.S. Over the 1990s, the share of new, adult immigrants who held a university degree jumped from around 25% to 47% in Canada, but only from 30% to 34% in the U.S.
2. The more pronounced change in language ability (which is not well controlled for in this and most other studies) in Canada as compared to the U.S., associated with the greater shift towards non-traditional source regions in Canada;
3. Changes in unobserved characteristics among entering immigrants in the two countries, such as the possibility that the more able university-educated new immigrants increasingly chose the U.S. over Canada. This seems possible, given the relative (to the U.S.) decline in economic outcomes in Canada, and the more rapid increase in the supply of highly educated immigrants;
4. Possible declines in the quality of the degrees held by entering immigrants, which may be more pronounced in Canada than in the U.S., again in light of the rapid increase in the supply and the more pronounced shift towards "non-traditional" source regions in Canada;
5. Differences in the occupational mix among the highly educated in the two countries, and its relationship to the occupational demand for labour.

The sorting out of the causes of the different economic paths for new immigrants in Canada and the U.S. will be left to future research.