Home About Us Reports Research Paper 2002 Age Discrimination and the Employment Rights of Elderly Canadian Immigrants Page 11

About Us

Reports

Research Paper

Age Discrimination and the Employment Rights of Elderly Canadian Immigrants

The earnings gap between immigrants and the native-born is generally estimated using cross-sectional data and an equation which fits employment income to personal and employment characteristics. Personal characteristics included as explanatory variables differ depending upon the hypothesis to be tested: some studies include marital status to control for expected differences in the choice of hours of work, while others include country of source or ethnicity. Employment income is used as the dependent variable in order to capture differences in wages, hours of work, and unemployment rates. Consider a conventional income function:

log(Y) = a₀ + a_i(Z_i) + a_j(P_j)

where log(Y) = the logarithm of annual employment income;

Z_i = a vector of employment characteristics; and

P_j = a vector of personal characteristics.

Employment characteristics include years of schooling (SCH), work experience (EXP) and language skills (LANG) as measures of human capital endowment. Since the earnings-experience profile is expected to increase at a decreasing rate, experience also enters the equation in quadratic form (EXP, EXPSQ). To test for the difference in earnings between immigrant and native-born workers, a dummy variable is included for migration status (IMMIG) and the years since migration (YSM):

log(Y) = a₀ + a₁(SCH) + a₂(EXP_j) + a₃(EXPSQ) + a₄(LANG) + a₅(IMMIG) + a₆(YSM)

where SCH = years of education;

EXP = years of work experience, estimated using Mincer=s rule of thumb that EXP = (Age - SCH - 6);

EXPSQ = years of work experience squared;

LANG = a dummy variable assuming a value of 0 for no knowledge of English or French, and 1 otherwise;

IMMIG = a dummy variable assuming the value of 0 for native-born and 1 for immigrants;

YSM = years since migration.

If years of schooling, years of experience and language ability capture all observable skill differences, then years since migration measures the “assimilation effect,” or the growth in earnings independent of observable measures of labour quality.

To capture possible cohort effects, first elaborated by Borjas, [28] it is necessary to track immigrant cohorts over time. Owing to the lack of large-scale longitudinal data sets, “synthetic cohorts” are created by “pooling” data from two or more censuses. The regression model would then be expanded by including dummy variables for different “vintages” of immigrants:

log(Y) = a₀ + a₁(SCH) + a₂(EXP_j) + a₃(EXPSQ) + a₄(LANG) + a₅(IMMIG) + a₆(YSM)

+ c₁(COHORT₁) + c_n(COHORT_n)

Specifying the income equation in this manner makes it possible to determine whether each immigrant cohort faces a different entry penalty (or initial wage gap), while the assimilation rate is assumed to be the same across cohorts.

In order to consider retirement income, we assume that the chief variation is due to accumulated savings out of previous employment income. Past employment income is expected to vary according to the years of education (SCH) and knowledge of either official language (LANG, a dummy variable assuming a value of 1 for knowledge of French or English and 0 otherwise). No estimator for years of employment is included. We also include a variable for current age (AGE) since we expect dissaving to occur after the age of retirement, and a dummy variable for marital status (NEVERMARRIED). The expectation is that men who remain single throughout their life have a lower income (since they were not supporting a spouse or family) and higher for women (since they had no spousal income). Finally, a dummy variable for gender in included (GENDER = 1 for men and 0 for women).

Our estimating equation, therefore, is:

log(Y) = α₀ + α₁SCH + α₂LANG + α₃NEVERMARRIED + α₄AGE + α₅IMMIG + α₆GENDER + α₇COHORT_i + ε

The same equation can be applied to separate samples for men and women with the exclusion of the variable for gender.

The estimated results are summarized in Appendix Table 2. All of the predictors have the expected sign with the exception of AGE for women, and all are significant at the 95 percent level. The results suggest that an individual’s human capital endowment (SCH and LANG) plays an important part in increasing income; that as elderly men age, their income falls (due to dissaving), but for elderly women, income rises with age; that being or having been married increases a man’s income but lowers a woman’s income; and that despite these observations, a significant unexplained gap remains in the income of men and women. The effect of immigration status and the immigrant’s cohort group is particularly revealing. The empirical results suggest that, with the exception of very early arrivals, immigrants receive less money than native-born Canadians, and that the difference becomes more pronounced with the later immigration cohorts. This is consistent with the view that more recent immigrants have less current income because of a shorter history of work in the Canadian economy.

Appendix Table 3 summarizes the estimated marginal effects of changes in the value of each variable upon expected income. Given the mean values for individuals in the sample (for men, 72.7 years old, with 9.7 years of schooling, native born, speaks an official language and is married; for women, 73.7 years old, with 9.5 years of schooling, native born, speaks an official language and is married), the estimated annual income for the representative man and woman is $21,344 and $13,337 respectively. The marginal effects displays how a change in the value of each variable is predicted to affect average income.

These results, while admittedly exploratory, suggest that further research is needed on the relationship between the incomes of elderly Canadians and immigration status.