Do immigrant and Canadian-born cognitive skills differ?

Charts 4.1.A and 4.2.A plot the kernel density functions of the individual averages of the four cognitive skill scores for males and females, respectively.¹ For both men and women the cumulative distribution function (CDF) for Canadian-born scores lies to the right of the immigrant CDF throughout the sample range and stochastically dominates the immigrant CDF at conventional significance levels.² The differences between immigrants and native-born Canadians are especially large at the low end of the skill distribution. For example, among women the immigrant-Canadian born gap is over 50 points at the 10^th percentile, 38 points at the median, and 25 points at the 90^th percentile. The skill gaps for men are smaller but follow the same pattern – a differential of 42 points at the 10^th percentile, 30 points at the 50^th percentile, and 14 points at the 90^th percentile. Charts 4.1.B and 4.2.B plot the cognitive skill distributions with the respondents who did not complete the main tests removed. The immigrant literacy distributions now appear more similar, though still inferior, to the Canadian born distributions.

Charts 4.1.C and D, and 4.2.C and D, show the skills distributions for immigrants with and without Canadian education relative to that of the Canadian born. The skill distributions of both immigrant groups are inferior to those of the Canadian born, and the difference between the respective distributions is largest for immigrants educated outside of Canada.³ There is also less dispersion in the cognitive skills of immigrants who completed their education in Canada than is the case for those educated outside of Canada. In addition, for immigrant men educated in Canada the upper tail of the distribution is similar to that for Canadian born men, whereas this is not the case for male immigrants who were educated prior to arrival in Canada, nor is it the case for either group of female immigrants. For both men and women a noteworthy difference between immigrants who obtained some or all of their education in Canada and those who did not is the relative absence of individuals with high skill levels in the latter group.

The final two charts, 4.1.E and F, and 4.2.E and F, compare the skills distributions of the two immigrant groups to those of the Canadian born after removing individuals who did not complete the main cognitive skill tasks. Doing so makes little difference to the distributions for immigrants with Canadian education, but makes a substantial difference to the distributions for immigrants without Canadian education. A significant number of those with low skill levels are evidently in this group.

Chart 4.1
Distribution of average skill score – males

Chart 4.2
Distribution of average skill score – females

Charts 4.3 to 4.10 provide further evidence of differences in the distribution of skills among native-born Canadians and the two immigrant groups. These charts recreate charts 4.1 and 4.2 for each of the four individual skills. For both men and women the immigrant distributions are clearly inferior to those of the Canadian born. The immigrant – Canadian-born skill gaps are most evident for prose literacy and least evident for numeracy. This latter pattern may reflect the tendency for numeracy to be less language dependent. As was the case with the average scores, separating immigrants into two sub-samples delineated by where they obtained their education reveals substantial differences between the two groups. For each of the four cognitive skills and both genders the distributions for immigrants educated in Canada have lower dispersion than those for foreign-educated immigrants and proportions of individuals with high skill levels that are closer to those observed among the Canadian born. This similarity at the top of the skill distribution is especially evident for males. Both immigrant groups have larger proportions of their respective distributions with low skill levels (scores below 200) than is the case for the Canadian born. This concentration in the lower tail of the distribution is especially pronounced for immigrants who completed their education prior to arriving in Canada.

Chart 4.3
Distribution of prose literacy score – males

Chart 4.4
Distribution of prose literacy score – females

Chart 4.5
Distribution of document literacy score – males

Chart 4.6
Distribution of document literacy score – females

Chart 4.7
Distribution of numeracy score – males

Chart 4.8
Distribution of numeracy score – females

Chart 4.9
Distribution of problem solving score – males

Chart 4.10
Distribution of problem solving score – females

Removing individuals who had their skill scores imputed has a substantial impact on the skill distribution, particularly in the case of immigrants. Thus, it is informative to look at the characteristics of those who did and did not complete the main skill tasks – a comparison that is provided in Table 4.1. Given that most of the analysis in this paper focuses on the earners sample, we report the results of this comparison for our sample of immigrant workers. Immigrant workers who did not complete the main tasks were, on average, older at the time of migration than those who completed these tasks, have lower educational attainment and are roughly half as likely to have English or French as a first language. Nonetheless, around 75% of immigrants who did not complete the main skill tasks report that they speak English on a regular basis at work. About the same proportion reports believing they have sufficient reading skills in English to do their job well, although about 10% fewer report being satisfied with their English writing skills. Literacy use at work is also much lower among immigrants who did not complete the main skill tasks. For example, about half report rarely using writing skills in English at work, compared to only about 20% among immigrants who did complete the main skill tasks. Finally, immigrant workers who did not complete the main skill tasks receive around 70%-80% of the average annual earnings received by immigrants who did complete these tasks, although those earnings are still fairly high ($31,435 for men and $25,134 for women).

The principal objective of this paper is to examine the relationship between cognitive skills and the labour market outcomes experienced by immigrants and native-born Canadians. However, because of its relevance to our analytical framework we also examine the determinants of individual skills, with particular focus on the relationship between cognitive skills and human capital variables like education and age or experience. Table 4.2 reports the results from regressions of the average skill score on age and education plus parents’ education, language proficiency, source country, and province of residence (not reported). The dependent variable is the log of the average skill score, so the coefficients can be interpreted as showing the percentage effect of a unit change in the variable of interest on the average score. We also include a dummy variable to control for the individuals who did not complete the main tasks, and whose test scores were therefore imputed. We do not view this regression as representing a causal story of how cognitive skills are generated. Instead, we interpret the coefficients as revealing partial correlations that are useful for summarizing skill patterns in the population.

Table 4.1
Summary statistics by imputed and non-imputed skill scores among immigrant workers

Table 4.2
Estimated coefficients from regressions with log of average cognitive skill score as the dependent variable

Regression results are reported separately for males and females as some of the determinants of skills vary by gender. We also allow the effects of age and education to differ across the three groups – Canadian-born, immigrants with Canadian education, and immigrants who completed their education prior to arrival. Columns 1 and 3 allow the age profiles (as well as education effects) to differ across the three groups, while columns 2 and 4 also include a “years since migration” variable that allows the immigrant age profile after arrival to differ from that prior to arrival. For ease of interpretation, the age and education variables used in the Table 4.2 estimation are defined in such a way that the immigrant coefficients stand on their own; that is, they are not defined relative to the Canadian-born coefficients.

In the Table 4.2 regressions the omitted category consists of native-born Canadians with less than high school education, whose mothers and fathers also had not completed secondary school, whose first language is English or French and who completed the main skill tasks. Relative to these Canadians, the skills of immigrants with Canadian education are 8% to 11% lower, although the estimate associated with this immigrant dummy is imprecise and not significantly different from zero. Foreign-educated immigrants have much lower skills than the reference group – in the order of 20% lower for females and more than 30% lower for males. These estimates are statistically significant. A further reduction of 11-12% is associated with inability to complete the main skill tasks.

The age profiles reported in columns 1 and 3 are similar for men and women. For all three groups there is a positive partial correlation between age and skills, but one that is small in magnitude and diminishes with increased age. The estimated impact is about the same for Canadian-born and Canadian-educated immigrants – about half of 1% per year at early ages, diminishing to zero by age 25 to 30. For foreign-educated immigrants the magnitudes of the estimated coefficients are similar to those for the other two groups but the estimates are imprecise and not significantly different from zero. These results suggest that among adults (i.e. after about age 20) there is essentially no relationship between age (or experience) and the individual’s skills. This conclusion is consistent with Green and Riddell (2003) who find that years of experience are essentially uncorrelated with the individual’s skill level across various specifications in the IALS data that is predominantly made up of adult workers. Adding controls for years since arrival provides some evidence suggesting that years in Canada matter more than years abroad, especially for immigrants with Canadian education. Nonetheless the main message is the small effect of age on an individual’s skills.

On the other hand, there is a strong relationship between education and cognitive skills for all three groups. Canadian-born high school graduates have skills that are approximately 13% higher than those of high school dropouts, while university graduates have average skill scores that are 22-24% higher than individuals who have not completed high school. University graduates also have dramatically higher skills than graduates from other post-secondary institutions.

Among Canadian-educated immigrants the impact of educational attainment on skills is similar to that of native-born Canadians – gains of 10-11% associated with secondary school, somewhat larger gains associated with non-university post-secondary, and skill gains in the order of 26% associated with university education. In contrast, the educational gradients are much larger for immigrants who obtained their education abroad – estimated skill gains of 18% for high school and approximately 35% for university education.

In summary, for all three groups there are substantial gains in skills associated with higher educational attainment.⁴ The skill gains associated with higher education are greatest for immigrants without Canadian education, and similar in size for the Canadian born and immigrants with Canadian education. Combining the negative immigrant dummies with the positive effects of higher education, foreign-educated male immigrants with a university degree have an average cognitive skill score that is 20-25% lower than that for a university-educated Canadian-born male. In contrast, among male high school dropouts the skills gap between these two groups is approximately 30-40%. The comparable estimates for females are approximately 5-10% and 15-20%, respectively. For both genders a university education narrows the skills gap relative to high school dropouts by approximately 10 percentage points.

Parents’ education is also positively associated with the child’s cognitive skills, but the estimated impacts are relatively modest. Among men the key factor is having a father or mother with at least a high school education; beyond this level there is little additional impact of the father’s education. Among women there is a clear gradient, with the estimated effect of either parent being a university graduate being approximately double that of either parent being a high school graduate. The estimates imply that if both parents are university graduates the average skills of men are 8% higher than if both parents are high school dropouts, while those of women are 10% higher.

Language also appears to exert only a modest effect on cognitive skills. Having one’s first language being other than English or French is associated with average skills that are 2-3% lower compared to individuals whose first language is one of the two official languages. Although the estimated impacts of language are very similar for males and females, the language coefficients for males are less precisely estimated and not statistically significantly different from zero.

These results indicate that, for foreign-educated immigrants, there is a substantial skills deficiency relative to the Canadian born, and that deficiency declines somewhat with education. In contrast, for immigrants who obtained some of their education in Canada the skills disadvantage relative to native-born Canadians is much smaller, but that disadvantage is not reduced at higher education levels. Notice that in obtaining these results we control for region of origin, and that immigrants from the United States or United Kingdom do not face as large a skills disadvantage. These differences in measured skills could arise because of differences in the quality of education across source countries or because immigrants not from the United States or United Kingdom have some difficulties in English or French.⁵

Notes

1. We estimate the kernel density functions with the density function in Stata, using the Epanechnikov kernel and Stata's default bandwidth formula.
2. More precisely, we cannot reject the null hypothesis that the Canadian born CDF first order stochastically dominates the immigrant CDF at conventional significance levels, using the test for first order stochastic dominance described in Barrett and Donald (2003).
3. For example, the median skill score of female immigrants without Canadian education is 54 points below that of natives (Canadian born), while that of female immigrants with Canadian education is 16 points below that of natives. The story for males is similar: a differential in the median skill score of 45 points between foreign-educated immigrants and native Canadians versus a differential of 17 points between Canadian-educated immigrants and natives.
4. As noted previously these estimates should not be given a causal interpretation. The positive association between education and skills may (or may in part) arise because education exerts a causal influence on skill formation. However, the association could also arise because both educational attainment and cognitive skills are correlated with unobserved factors such as innate ability.
5. Note, however, that the regressions control for the respondent's first language being other than English or French, so any additional effect of language on skills would need to arise for reasons that are not captured in our language variable.