Immigrant Families in the Canadian Labour Market 378 Christopher Worswick

CHRISTOPHER WORSWICK Department of Economics University of Melbourne Melbourne, Australia

Nous comparons les activités des couples d’immigrés et de non-immigrés sur le marché du travail à l’aide de données de 1981 et 1991 de Recensement Canada. Nous donnons de nouvelles indications sur la performance d’immigrés masculin et féminin en termes de trois composantes annuelle de leurs revenus: les salaires horaires, le nombre d’heures travaillées par semaine et le nombre de semaines travaillées annuellement. Nous ne trouvons aucune indication qui soutiendrait l’argument d’un arbitrage entre la carrière du mari immigré au détriment de la femme. Cependant, la performance de l’épouse sur le marché du travail joue un rôle important dans la création de revenus des familles d’immigrés. Ces résultats vont dans le sens d’une évaluation et d’une mise en place de politiques d’immigration qui favorisent la famille. The labour market activity of immigrant and non-immigrant married couples is compared using data from the 1981 and 1991 Canadian Censuses. New evidence is provided on the performance of immigrant men and women in terms of three components of annual earnings: hourly wage rates, hours worked per week, and weeks worked per year. Evidence of intra-family trade-offs of investments in the immigrant husband’s career at the expense of investments in the wife’s career are not found overall. However, the wife’s labour market performance is found to play a major role in the earnings creation of immigrant families. The findings support a family orientation to both the evaluation and the implementation of immigration policies.

INTRODUCTION

I

n this paper, the labour market performance of immigrant families is analyzed. The success of Canada’s immigration policy is currently under debate in terms of the number of immigrants admitted each year and the composition of the inflow between independent class immigrants, where admission is based on the applicant’s skills, and family class immigrants, where admission is based primarily on the applicant having relatives in Canada (Campbell 1994; Simpson 1994). Given this debate and the recent changes announced to the point system used to select immigrants (Citizenship and Immigration Canada 1995), it is important to gain a full understanding of how the selected immigrants are sucCANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

ceeding in the Canadian labour market. Also, given that the immigration decision is often based upon what is best for the immigrant family rather than a particular family member, it is important to see how different immigrant family members are performing in the labour market. Economic studies of immigrant labour market adjustment have focused on differences in earnings between immigrants and non-immigrants as a function of the immigrant’s years-since-migration (YSM) and arrival cohort. The results for men indicate that immigrant men have lower earnings than similar non-immigrant men immediately after migration. However, the immigrant men experience higher growth in earnings over time with the VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market magnitude of this difference being the focus of much of the debate in the literature (see Chiswick 1978; Borjas 1985; Bloom and Gunderson 1991; LaLonde and Topel 1992; Baker and Benjamin 1994). This paper significantly extends this work by examining the labour force decisions of cohorts of immigrant married women and men in tandem. A specific focus is to determine whether and how the labour force decisions of immigrant wives are related to the human capital investment patterns of immigrant husbands. In what has become known as the Family Investment Hypothesis (FIH), several researchers have argued that financing investments in host-country skills by immigrant husbands is a factor affecting the labour force decisions and human capital investments of immigrant married women (see Long 1980; Beach and Worswick 1993; Duleep and Sanders 1993; Baker and Benjamin 1996). Associated with this general hypothesis are the following specific expectations: in relation to their husbands’ investments in host-country specific skills, immigrant women would be more likely to work and to work longer hours and more weeks than would otherwise be the case. Conditional on their own levels of host-country-specific skills, immigrant women would also be less likely to undertake human capital investments in order to take jobs that paid more during the period in which the husband’s investment in host-country skills was most intense. Subsequently, the wages of immigrant women may be higher than would otherwise be the case during the period in which the husband is investing. By foregoing investment, the wage-age profile of immigrant women would be expected to be flatter than otherwise. Using data from the 1970 US Census, Long (1980) finds that immigrant women with few years of residence in the United States work more hours and have higher earnings than both similar nonimmigrant women and immigrant women with more years of residence. Long argues that these earnings patterns of immigrant women could be explained under the FIH. Using the Canadian Job Mobility

379

Survey, Beach and Worswick (1993) find, in sharp contrast to what had been previously found for men, that the immigrant/non-immigrant earnings differential for women does not depend on the immigrant’s years of residence. It is argued that this may be explained by the FIH. Two data limitations hamper each of these studies: (i) the availability of only a single cross-section survey, and (ii) limited information on the labour supply of the immigrant women. The availability of only one cross-section means that each cohort of immigrant women is observed at only one value of YSM. The poor controls available for labour supply means that it is impossible to distinguish between the dynamic returns to an immigrant’s work time with more YSM and the family’s labour supply decision for the immigrant across time. Duleep and Sanders (1993) use the 1980 US Census and find that the labour force participation rate is higher for immigrant women whose husbands have difficulty adapting to the new labour market and may need to make investments in skills specific to the US labour market. Baker and Benjamin (1994) use the 1986 and 1991 Surveys of Consumer Finance for Canada and estimate a reduced-form labour supply model for married couples. They find evidence that immigrant women increase their hours of work in order to finance the labour market adjustment of their husbands. However, the analysis is hampered by the fact that the time frame over which each cohort is observed is only five years; therefore, the long-term dynamics of labour market adjustment in immigrant families are not observed. Implicit in these studies are two basic approaches to the analysis of the FIH. One approach, as exemplified by Long (1980) and Beach and Worswick (1993), has been to compare the patterns of work and earnings behaviour of immigrant married women relative to non-immigrant women with the patterns of work and earnings behaviour of immigrant and non-immigrant men found in previous studies. Another approach has been to identify the family investment model by using variation in the

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

380 Christopher Worswick likely extent to which immigrant husbands invest in host-country human capital, while holding constant the wife’s own level of host-country skills as measured by her own years-since-migration and level of host-country language proficiency. This approach is adopted by Duleep and Sanders (1993) and Baker and Benjamin (1994). This paper adopts the former strategy. The pattern of work and earnings behaviour of the husband relative to the native-born is compared to the patterns of work and earnings behaviour of the immigrant wife relative to the native-born. An important improvement over the Long (1980), and Beach and Worswick (1993) efforts is that data are used from more than one time period. In addition, this study analyzes the labour market behaviour of immigrant married men and women, relative to their nonimmigrant counterparts, in tandem, rather than comparing the labour market behaviour of married immigrant women to the findings of the literature on immigrant men. In addition to analyzing the annual earnings of immigrant husbands and wives, this paper also analyzes the labour market adjustment of immigrant husbands and wives along the following dimensions: the hourly wage rate, the hours of work per week, and the weeks of work per year.

The remainder of the paper is structured as follows. In the second section the basic regression specifications are described. In section three the data set and estimation sample used are presented. In the remaining sections, results are presented from wage, hours, weeks, and annual earnings regressions for immigrant and non-immigrant husbands and wives. The conclusions and implications for public policy are discussed in the final section.

REDUCED-FORM EQUATIONS AND SPECIFICATION ISSUES To measure the labour market adjustment of immigrants, the labour market outcomes of similar nonimmigrants are used as a benchmark and the deviations of immigrant labour market outcomes from the non-immigrant outcomes are analyzed. Figure 1 demonstrates possible wage adjustment profiles of two immigrant arrival cohorts. The vertical axis is the difference between the log wage of an immigrant and the log wage of a non-immigrant, holding other characteristics the same. The horizontal axis measures the number of years-since-migration (YSM) of the immigrant. The solid line and the dashed line are the Immigrant Adjustment Paths

FIGURE 1 Immigrant Adjustment Path (IAP)

SOURCE : Author’s compilation.

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market (IAP) for the wage rates of immigrants from the 1960s arrival cohort and the 1970s cohort, respectively. To estimate these lines, data are used on wage rates for non-immigrants and for immigrants from these two arrival cohorts across time. The data used in this paper come from two crosssectional surveys, the 1981 and 1991 Census of Canada Family Microdata Files. Three immigrant arrival cohorts can be identified in both surveys: arrival before 1961, arrival between 1961 and 1970, and arrival between 1971 and 1980. In the 1991 sample, immigrants from these three cohorts can be identified as well as immigrants from the following six arrival cohorts: arrival between 1981 and 1985, and arrival in 1986, 1987, 1988, 1989 and 1990. Eight regression equations are estimated in the paper: log wage, log hours, log weeks and log earnings equations for both the husband and the wife. All eight of these equations have the same structure; therefore, consider the husband’s illustrative log wage equation:

In W = ß 0 + Xß+ Aα + (A ∗ Y 91 )α 91 + C1γ + (C1 ∗ Y 91 )γ 91 + C2δ + u (1) where lnW is the log of the husband’s hourly wage. The vector X contains: (i) demographic variables of the husband and (ii) interactions of these demographic variables with the immigrant dummy variable, FB, allowing for separate effects of demographic characteristics on the husband’s wage for immigrants and non-immigrants. 1 The vector A contains age dummy variables identifying whether the person was age 25-29, 30-34, 35-39, 40-44 (default), 45-49 or 50-54 in 1981.2 These dummy variables are also included as interactions with the 1991 survey year dummy variable, Y91. This specification allows for different effects of age on the husband’s wage for each of the six age-in-1981 cohorts, in each survey year. Interpreted in terms of a wage-age profile, this specification leads to estimates of two points on the wage-age profile of each of the six age-in-1981 cohorts (one point from the 1981 Census, the other from the 1991 Census).

381

The regression specification employs the identifying assumption of age effects on the husband’s wage being the same for immigrants and nonimmigrants. This enables the identification of yearssince-migration effects on the wages of immigrants. The vector C1 contains three dummy variables identifying the three immigrant arrival cohorts (pre1961, 1961-70, and 1971-80) which appear in both the 1981 Census and the 1991 Census. These dummy variables are also included as interactions with the 1991 survey year dummy variable. The vector C2 contains six dummy variables identifying immigrants from the six immigrant arrival cohorts who arrived in Canada after 1980 (1981-85, 1986, 1987, 1988, 1989, and 1990). This specification allows for different effects of arrival cohort on the husband’s wage for each arrival cohort in each survey year. Interpreted in terms of a wage IAP, this specification leads to estimates of two points on the wage IAP of each of the three pre-1981 cohorts, and one point on the wage IAP profile for each of the six post-1980 cohorts.3 It is common in the immigration literature to refer to the difference between the wage growth of a cohort of immigrants and the wage growth of nonimmigrants, ceteris paribus, as the wage “assimilation” of that cohort of immigrants. In the wage equation, the vector of coefficients γ91 measures the wage assimilation from 1980 to 1990 of each of the three cohorts of immigrants who arrived in Canada before 1981.4 In terms of Figure 1, the three assimilation coefficients in γ91 measure the movement along the IAP (or the slope of the IAP) over the relevant years-since-migration values for each of the three arrival cohorts.5

THE DATA AND ESTIMATION SAMPLE In the absence of suitable panel data, researchers have used synthetic cohort data in the analysis of immigrant labour market adjustment. The synthetic cohorts are two or more cross-sectional surveys where the samples can be defined so as to represent

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

382 Christopher Worswick the same underlying population at different points in time. The Canadian Census data used in this analysis come from the Family Files which contain 94,745 census families from 1981 and 345,351 census families from 1991. Since the 1991 data are a larger random sample of the population than the 1981 data, weights are used in the analysis to account for the relative oversampling in 1991. Due to the large numbers of non-immigrant households in the two data-sets, a 25 percent random sample is taken of non-immigrant households. The estimation sample contains married couples with both spouses having been age 25 to 54 in 1981. Households where one spouse is an immigrant and the other spouse was born in Canada are excluded from the sample. Duleep and Sanders (1993) also focus on households where both spouses are immigrants. Family trade-offs in terms of one spouse working to subsidize the other spouse’s career are more likely to be necessary in households where both spouses are immigrants. In households where only one spouse is an immigrant, the non-immigrant spouse is already established in Canada with Canadian education credentials, knowledge of at least one of the official languages, and in most cases the nonimmigrant spouse has work experience with Canadian employers. The sample is restricted to households where the husband works in both the reference week and the reference year. This is standard in the economic literature on the labour market adjustment of immigrant men.6 Wage rates are constructed by dividing the product of the reference weekly hours variable and the weeks-worked-per-year variable into the individual’s annual earnings the previous year.7 This is an imperfect measure of the true wage rate offered to the individual in that the weekly hours variable refers to the reference week and not the average weekly hours of the reference year. To construct the wage variable, it was necessary to exclude from the sample households where either spouse had positive earnings and weeks in the reference year, and zero hours in the reference week, or vice versa.8 CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

Table 1 contains sample means of selected variables over the immigrant and non-immigrant samples in each survey. The average wages of immigrant women are $1.37 lower than those of the nonimmigrant women in 1981, but this wage gap has dropped to $0.89 by 1991. Immigrant women work 2.2 hours more per week in 1981 than nonimmigrant women and 1.85 more hours per week in 1991. Immigrant women work one-third of a week more on average in 1981 and less than one week more in 1991. Immigrant men have $0.83 lower wages than non-immigrant men in 1981, but by 1991 the difference has decreased to $0.25. The mean hours per week and weeks per year patterns of immigrant men are very similar to those of nonimmigrant men.

THE WAGE RATES OF IMMIGRANT AND NON-IMMIGRANT COUPLES Table 2 contains estimates from log wage regressions for husbands and wives. The first and third columns list results from the log wage equation estimated over the sample of husbands. 9 The first three coefficient estimates give the estimated difference in log wages in 1981 between immigrant men and non-immigrant men, ceteris paribus, for each of the three immigrant cohorts evaluated at high school completion.10 In terms of Figure 1, each of these coefficient estimates is an estimate of the height of the cohort’s IAP over the relevant range of YSM. Immigrant men who arrived in the 1970s have 28 percent lower wages in 1981 than similar nonimmigrant men, while those arriving earlier have 11 to 15 percent lower wages with each of these differences significant at the 5 percent level. The next three coefficient estimates give the wage assimilation of immigrant men for each immigrant cohort. In terms of Figure 1, these coefficients measure the slope of the IAP between the two survey years. The estimated assimilation of the pre-1961 cohort and the 1971-80 cohort is 8 and 9 percent, respectively, and statistically significant, while the assimilation of the 1961-70 cohort is not significant. VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market

383

TABLE 1 Sample Means for Selected Variables

Immigrants 1981 N=3688

Immigrants 1991 N=10472

Non-Immigrants 1981 N=3808

Non-Immigrants 1991 N=9815

8.00 34.67 43.37 .2400 .1883 .1280 .3364 .0848 .0225

10.03 35.47 46.71 .2042 .1621 .1598 .3346 .1038 .0355

9.37 32.47 43.04 .1253 .2661 .1895 .3364 .0734 .0093

10.92 33.64 46.11 .0877 .2171 .2107 .3732 .0935 .0178

11.36 42.87 47.56 .1956 .1261 .0576 .4431 .1099 .0677

13.29 42.07 48.25 .1748 .1309 .0885 .3965 .1277 .0816

12.09 42.69 48.26 .1322 .1988 .1238 .3994 .1099 .0359

13.54 42.31 48.38 .1110 .1960 .1372 .3933 .1185 .0440

Wife’s Variables WAGE* HOURS* WEEKS* PRIMARY SECOND HSGRAD POSTSEC BACH GRAD

Husband’s Variables WAGE HOURS WEEKS PRIMARY SECOND HSGRAD POSTSEC BACH GRAD

NOTE: *The mean is calculated over the sample of women who worked. SOURCE : Author’s compilation.

The next six coefficient estimates give the estimated differences in log wages in 1991 between immigrant men who arrived after 1980, and nonimmigrant men, ceteris paribus. In the 1991 survey, the immigrant men from these cohorts have 36 to 61 percent lower wages than similar non-immigrant men and each of these differences is significant. A test is performed of the hypothesis that the average value of the wage IAP over the zero to ten YSM range is the same for the 1980s cohort and the 1970s cohort.11 The coefficient estimate for the variable C7180 is the estimate of the wage IAP for the 1970s

cohort over the YSM range of zero to ten years. A weighted average of the coefficients on the controls for immigrant arrival year in the 1980s is derived (because of the detailed arrival year breakdown in the 1991 Census) and the coefficient on C7180 is subtracted off of the weighted average to give the difference in the IAP over the zero to ten range of YSM for the two cohorts.12 The immigrant/nonimmigrant wage differential over zero to ten YSM is found to be 14 percentage points lower for the 1980s cohort compared with the 1970s cohort and this difference is statistically significant.13 A second

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

384 Christopher Worswick TABLE 2 Immigrant/Non-Immigrant Wage Differences by Immigrant Cohort for Husbands and Wives

Husband

Wife

Husband

Wife

CPRE61

-.1474** (.034)

-.0880 (.053)

C8185

-.3550** (.041)

-.2773** (.052)

C6170

-.1094** (.033)

-.1531** (.050)

C86

-.3676** (.078)

-.2666** (.095)

C7180

-.2759** (.035)

-.2551** (.049)

C87

-.3778** (.064)

-.3223** (.074)

CPRE61*YR91

.0777** (.029)

.0816* (.041)

C88

-.6148** (.079)

-.3718** (.070)

C6170*YR91

.0303 (.025)

.0816* (.035)

C89

-.4503** (.066)

-.3422** (.077)

C7180*YR91

.0876** (.028)

.1136** (.035)

C90

-.4983** (.080)

-.4264** (.088)

R2 F N

.083 47.6 27,783

.105 39.4 19,558

NOTES: 1. Standard errors are in parentheses. 2. ** denotes significance at the 1 percent level; * denotes significance at the 5 percent level. 3. The first two and last two columns contain estimates from the husband’s log wage equation and the wife’s log wage equation, and have the same structure as (1). The variables listed in the table are defined in the Appendix. Dummy variables are included in the equations identifying: (i) the education categories 0-8 years, 9-13 years without a high school diploma, a high school diploma (default), completed a bachelor’s level university degree, and completed a university degree above the bachelor’s level; (ii) place of residence in the Atlantic provinces, Quebec outside Montreal, Montreal, Ontario outside Toronto, Toronto (default), the Prairie provinces, British Columbia outside Vancouver, and Vancouver; (iii) language fluency in English but not French (default), French but not English, and both English and French. In the wife’s equation, the number of live births is also included. A complete set of estimates is available upon request from the author. SOURCE : Author’s compilation.

test of a decline in the wage IAP of immigrant and non-immigrant men is performed. The coefficient on C6170 is subtracted from the sum of the coefficients on C7180 and C7180*YR91. This difference is statistically significant and implies that the wage IAP of the 1980s cohort of immigrant men is 8 percentage points lower than that of the 1970s cohort over the 11 to 20 YSM range.14 CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

These results are consistent with the findings of Borjas (1985) for the United States and Baker and Benjamin (1994) for Canada. Immigrant men have significantly lower wages than non-immigrant men at time of arrival and they experience low rates of wage assimilation. Also, the immigrant-nonimmigrant wage differential, holding YSM constant, has worsened for more recent arrival cohorts. VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market The results in the second and fourth columns of Table 2 are from the wage regression estimated over the sample of immigrant women who work.15 The general specification is the same as the one used in the husband’s equation with the wife’s variables replacing those of the husband. Controls for the number of live births of the wife are also included as a proxy for time spent out of the labour market due to caring for young children. From the first three coefficient estimates, we see that in the 1981 data, immigrant women who arrived after 1960 have 15 to 26 percent lower earnings than similar non-immigrant women, while the women who arrived before 1960 do not have significantly different wages than non-immigrant women. The wage assimilations of the three pre-1981 arrival cohorts are 8 to 11 percent and each estimate is significant at the 5 percent level. Therefore, the estimates of the wage assimilation of the immigrant women from these arrival cohorts are larger than what were found for the same cohorts of immigrant men. Under the FIH, one would expect the opposite to be the case. The theory predicts that immigrant wives will have lower wage assimilation than their husbands since they are sacrificing career opportunities in order to accept a job which supports their families while their husbands become established in their careers. From the coefficients on the controls for arrival after 1980, we see that the 1991 wage rate for these cohorts is 27 to 43 percent lower than the wage rate for similar non-immigrant women and these differences are significant. The test is performed of the hypothesis that the average value of the wage IAP over the zero to ten YSM range is the same for the 1980s arrival cohort of immigrant women as the 1970s cohort. The coefficient on C7180 is subtracted off of a weighted average of the coefficients on the controls for arrival after 1980 and the difference is not statistically significant.16 The test of the equality of the wage IAP of the 1960s cohort and the 1970s cohort over the 11 to 20 YSM range is performed for women. The coefficient on C6170 is not significantly different from the sum of the coeffi-

385

cients on C7180 and C7180*YR91, indicating that the issue of declining unexplained productivity of immigrant cohorts does not appear to be a concern for immigrant married women. The first and third columns of Table 3 contain the estimates from the coefficients on the education controls from the log wage equation of the husbands. These coefficients appear in the parameter vector β in equation (1) and represent the returns to education for immigrant and non-immigrant husbands. The wages of non-immigrant men are increasing in their level of education with a particularly large and significant return of 42 to 48 percent to a university degree relative to having only a high school degree. Given the problems of having education credentials recognized after migration (Chiswick 1978), one would expect a flatter wage-education profile for immigrants. However, the results do not support this hypothesis. The returns to education for immigrant men are generally the same as those for nonimmigrant men with the exception of a lower return to a bachelor’s degree for immigrant men. This latter relationship is consistent with the immigrant men having difficulty having foreign educational credentials recognized as equivalent to Canadian credentials by Canadian employers. The wage equations were reestimated replacing the FB*BACH variables with interactions of the BACH variables with each of the immigrant arrival cohort variables. This allows separate returns to a bachelor’s degree according to arrival cohort. For immigrant men, the coefficients on these interaction variables were all close in magnitude and the hypothesis that they were equal could not be rejected. In the women’s equation, immigrant women who arrived in the 1980s had a lower return to a bachelor’s degree than did either non-immigrant women or immigrant women from earlier cohorts. This indicates that the problem of immigrant women having their bachelor’s degrees accepted in the Canadian labour market as equivalent to those of native-born women is largest for the most recently arrived immigrant women. From the estimates of the second and fourth columns of Table 3, we see that the wage-education

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

386 Christopher Worswick profile of non-immigrant women is steeper than what was found for non-immigrant men. For example, there is a 61 to 74 percent higher return to a university education compared to a high school diploma. Unlike what was found for men, immigrant women have a significantly flatter wage-education profile than do non-immigrant women. Immigrant women in the lowest education category have significantly higher offered wages than do nonimmigrant women, ceteris paribus. As was the case for immigrant men, immigrant women with a bachelor’s degree have significantly lower wages than similar non-immigrant women. Similar lower rates of return on education for immigrant women compared with non-immigrant women are found by Long (1980), and Beach and Worswick (1993).

The lower rate of return on education of immigrant women coupled with the fact that immigrant men face the same rate of return on their education as non-immigrant men are consistent with the FIH. If the husband is perceived to be the main earner, it may be that the family invests in job search and retraining for the husband so that he can find a job that suits his education, while the wife settles quickly in a job which may not suit her education but allows her to fund family consumption. However, an alternative explanation is that the family’s migration decision may be more closely related to the husband’s career than the wife’s career. In this case, one would expect the immigrant husband’s education to be more suited to the new labour market than the wife’s education, leading to the

TABLE 3 Estimated Returns to Education for Husbands and Wives

Husbands

Wives

Husbands

Wives

PRIMARY

-.1463** (.024)

-.3095** (.039)

FB*PRIMARY

.0476 (.030)

.1531** (.046)

SECOND

-.0314 (.020)

-.1188** (.024)

FB*SECOND

-.0427 (.026)

.0936** (.036)

POSTSEC

.1129** (.015)

.2267** (.022)

FB*POSTSEC

.0057 (.015)

-.0380 (.031)

BACH

.4150** (.022)

.6106** (.030)

FB*BACH

-.0992** (.028)

-.1414** (.042)

GRAD

.4808** (.029)

.7398** (.058)

FB*GRAD

.0121 (.034)

-.0949 (.072)

NOTES : 1. Standard errors are in parentheses. The regression specification is described in the notes of Table 2. 2. ** denotes significance at the 1 percent level; * denotes significance at the 5 percent level. 3. The default education category contains individuals who have completed high school but do not have postsecondary training. SOURCE : Author’s compilation.

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market differential rates of return on their schooling relative to those of their non-immigrant counterparts.17

387

THE HOURS OF WORK OF IMMIGRANT AND NON-IMMIGRANT MARRIED COUPLES

ally insignificant at the 5 percent level indicating that the hours of work per week of immigrant men do not differ from those of non-immigrant men, ceteris paribus. In particular, the hours assimilation is not significant from zero for each of the three pre1981 cohorts.

Table 4 contains the results from the log hours equations for husbands and wives. The results from the husbands’ equation are presented in the first and third columns. With the exception of the coefficient on C89, all of the coefficient estimates are individu-

The test of the hypothesis that the average values of the hours IAPs of the 1980s cohort and the 1970s cohort are equal over the zero to ten YSM range is performed. The coefficient on C7180 is subtracted from a weighted average of the

TABLE 4 Immigrant/Non-Immigrant Hours Differences by Immigrant Cohort for Husbands and Wives

Husband

Wife

Husband

Wife

CPRE61

.0050 (.012)

-.0010 (.029)

C8185

-.0177 (.016)

.0532 (.032)

C6170

-.0078 (.013)

.0544 (.031)

C86

-.0096 (.032)

.0528 (.060)

C7180

-.0156 (.014)

.0406 (.032)

C87

-.0275 (.025)

.0367 (.047)

CPRE61*YR91

.0065 (.012)

-.0013 (.026)

C88

-.0510 (.028)

.0446 (.044)

C6170*YR91

.0012 (.010)

-.0207 (.023)

C89

-.0623** (.026)

.0396 (.048)

C7180*YR91

-.0010 (.011)

.0082 (.023)

C90

-.0480 (.036)

.0552 (.056)

R2 F N

.011 5.85 27,783

.044 14.3 19,558

NOTES : 1. Standard errors are in parentheses. 2. ** denotes significance at the 1 percent level; * denotes significance at the 5 percent level. 3. The first two and last two columns contain estimates from the husband’s log hours equation and the wife’s log hours equation, and have the same structure as (1). The variables listed in the table are defined in the Appendix. The demographic controls in the husband’s equation are identical to those used in the husband’s wage equation of Table 2. In the wife’s equation, the demographic controls differ from those in the wife’s wage equation of Table 2 in that the controls for the number of live births are replaced by controls for the number of children present in the household (see Table 5). A complete set of estimates is available upon request from the author. SOURCE : Author’s compilation.

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

388 Christopher Worswick coefficients on the controls for immigrant arrival in the 1980s.18 The difference equals negative two percentage points but it is not statistically significant. Also, the hypothesis that the coefficient on C6170 equals the sum of the coefficients on C7180 and C7180*YR91 could not be rejected. Therefore, the immigrant/non-immigrant hours differential, holding YSM constant, has not changed significantly for recent arrival cohorts. The second and fourth columns give the estimates from the log hours equation for women.19 The results are very similar to those found for men. All of the immigrant cohort controls are insignificant indicating that immigrant women’s hours are not significantly different from the hours of non-immigrant women, ceteris paribus. The two tests of the hypothesis that the hours IAPs of recent cohorts of immigrant women are equal to the hours IAPs of earlier cohorts are performed and the hypothesis is not rejected in either case.20

In Table 5, estimates are presented of the coefficients on the controls for presence of children in the household from the hours equation for wives. Non-immigrant wives are found to reduce their hours of work when young children are present in the household. This effect is smaller for immigrant women. This is a common finding in studies of the labour market adjustment of immigrant women (see Long 1980; Beach and Worswick 1993; Baker and Benjamin 1996). It may be that adult relatives are more likely to be present in immigrant families than in non-immigrant families. Unfortunately, this information is not available in the data.

THE WEEKS OF WORK OF IMMIGRANT AND NON-IMMIGRANT MARRIED COUPLES Table 6 contains the results from the log weeks regressions. A dummy variable is included in each regression identifying immigrants who arrived in the

TABLE 5 Estimated Response of the Hours of the Wife to the Presence of Children

Interactions with FB variable One child present under age 6

-.0828** (.022)

.0570* (.023)

More than one child present under age 6

-.1525** (.043)

.1266** (.043)

Number of children present age 6-14 (maximum of two)

-.0577** (.008)

.0325** (.010)

More than two children present age 6-14

-.0924** (.030)

-.0100 (.038)

NOTES : 1. Standard errors are in parentheses. 2. ** denotes significance at the 1 percent level; * denotes significance at the 5 percent level. 3. The regressions are described in the notes to Table 4. SOURCE : Author’s compilation.

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market

389

TABLE 6 Immigrant/Non-Immigrant Weeks Differences by Immigrant Cohort for Husbands and Wives

Husband

Wife

Husband

Wife

CPRE61

-.0407** (.012)

-.0460 (.027)

C8185

-.0438** (.015)

-.0363 (.033)

C6170

-.0439** (.011)

-.0291 (.027)

C86

-.0861 (.045)

-.1251 (.067)

C7180

-.0734** (.014)

-.0869** (.028)

C87

-.0523* (.024)

-.0906 (.051)

C80

-.5616** (.092)

-.6422** (.054)

C88

-.0848** (.025)

-.1478** (.047)

CPRE6*YR91

.0330** (.012)

.0106 (.023)

C89

-.2888** (.049)

-.2416** (.053)

C6170*YR91

.0155 (.010)

-.0013 (.019)

C90

-.6060** (.065)

-.5905** (.062)

C7180*YR91

.0369** (.013)

.0625** (.019)

R2 F N

.060 32.6 27,783

.066 21.8 19,558

NOTES : 1. Standard errors are in parentheses. 2. ** denotes significance at the 1 percent level; * denotes significance at the 5 percent level. 3. The first two and last two columns contain estimates from the husband’s log weeks equation and the wife’s log weeks equation, and have the same structure as (1). The variables listed in the table are defined in the Appendix. The demographic controls in the husband’s equation are identical to those used in the husband’s wage equation of Table 2. In the wife’s equation, the demographic controls differ from those in the wife’s wage equation of Table 2 in that the controls for number of live births are replaced by controls for the number of children present in the household. A complete set of estimates is available upon request from the author. SOURCE : Author’s compilation.

1980 reference year. This is intended to pick up the effect on the individual’s weeks of work of the fact that he/she arrived after the start of the year. The coefficients on C80 and C90 are found to be negative and large in magnitude. This is due to the immigrants in these cohorts having a truncated weeks of work for the year.

The first three coefficient estimates indicate that the weeks of work of immigrant men in 1981 were 4 to 7 percent lower than those of similar nonimmigrant men. The weeks’ assimilation is found to be 3 to 4 percent and significant for the pre-1961 cohort and the 1971-80 cohort, respectively. The immigrant men who arrived between 1981 and 1988

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

390 Christopher Worswick have similar weeks of work to the non-immigrants, ceteris paribus. The immigrants who arrived in 1989 have much lower weeks of work (29 percent) than non-immigrants. This may be picking up an increased likelihood of being unemployed in the first years after arriving. The hypothesis that the weeks IAP over the zero to ten YSM range is the same for the 1980s and 1970s cohorts is not rejected. The weighted average of the coefficients on the controls for immigrant arrival year in the 1980s implies that the 1991 hours differential over the zero to ten YSM range is four percentage points smaller than the 1981 differential.21 However, this difference is not statistically significant.22 The second test of differences in the hours IAP across arrival cohorts is performed. The coefficient on C6170 is not found to be significantly different from the sum of the coefficients on C7180 and C7180*YR91. The results of the two cohort tests indicate that differences in unobserved characteristics of immigrant men across arrival cohorts do not lead to significant differences in weeks of work between the cohorts for given levels of YSM. The second and fourth columns of the table contain results from the log weeks regression over the sample of women who work.23 The weeks of work in 1981 of immigrant women who arrived in the 1970s are 9 percent lower than those of nonimmigrant women, ceteris paribus, and this difference is significant. The immigrant wives who arrived in the 1970s experience 6 percent higher growth in weeks of work than non-immigrant wives. The weeks assimilation of the pre-1970 cohorts is not significant indicating that their weeks IAPs are flat over the YSM ranges observed in the data. Immigrant wives who arrived in Canada between 1981 and 1988 work 4 to 15 percent fewer weeks than similar non-immigrant women. The immigrant women who arrived in 1989 have much lower weeks of work (24 percent) than non-immigrants. This may be picking up an increased likelihood of being unemployed in the first years after arriving. CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

Tests are carried out on the hypothesis that the weeks IAP of recent cohorts of immigrant women are equal to the weeks IAP of earlier cohorts. The coefficient on C7180 is subtracted from the weighted average of the coefficients on the 1980s arrival cohort dummy variables. 24 The weeks IAP for the 1980s cohort over the zero to ten YSM range is not found to be significantly different from the weeks IAP of the 1970s cohort over the same YSM range. Also, the coefficient on C6170 is not found to be significantly different from the sum of the coefficient on C7180 and the coefficient on C7180*YR91 indicating that the weeks IAP of the 1970s cohort does not differ significantly from the weeks IAP of the 1960s cohort over the 11 to 20 YSM range.

THE ANNUAL EARNINGS OF IMMIGRANT AND NON-IMMIGRANT MARRIED COUPLES In order to summarize the differences in wages, hours, and weeks of work between immigrants and non-immigrants, conventional annual earnings regressions are estimated and selected coefficients are presented in Table 7. The dependent variable is the natural logarithm of annual earnings which equals the sum of the dependent variables used in the previous sections: the natural logarithms of the wage, hours, and weeks. In the first and third columns, the results are presented from the estimation of the earnings equation over the sample of husbands. Immigrant husbands who arrived prior to 1981 are found to have 16 to 34 percent lower earnings than similar non-immigrant men in the 1981 data. These differences are larger than what was found in the wage equation of Table 2 and this is due to the lower weeks of the immigrants in these cohorts compared with the non-immigrants found in Table 6. The earnings assimilation is positive and significant for the pre1961 cohort and the 1971-80 cohorts of men as was found in the wage equation estimates of Table 2 and the weeks equation estimates of Table 6. The tests of whether or not the annual earnings IAP of immigrant men has shifted down for more VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market

391

TABLE 7 Immigrant/Non-Immigrant Annual Earnings Differences by Immigrant Cohort for Husbands and Wives

Husband

Wife

Husband

Wife

CPRE61

-.1821** (.034)

-.1364* (.063)

C8185

-.4167** (.041)

-.2490** (.062)

C6170

-.1601** (.032)

-.1141 (.060)

C86

-.4631** (.073)

-.3333** (.110)

C7180

-.3447** (.035)

-.2467** (.061)

C87

-.4573** (.063)

-.3782** (.087)

C80

-.8972** (.135)

-1.228** (.146)

C88

-.7504** (.075)

-.4807** (.082)

CPRE61*YR91

.1167** (.029)

.0879 (.049)

C89

-.8014** (.071)

-.5786** (.091)

C6170*YR91

.0465 (.024)

.0516 (.042)

C90

-1.152** (.105)

-1.015** (.104)

C7180*YR91

.1037** (.028)

.1402** (.043)

R2 F N

.133 78.6 27,783

.142 48.1 19,558

NOTES : 1. Standard errors are in parentheses. 2. ** denotes significance at the 1 percent level; * denotes significance at the 5 percent level. 3. The first two and last two columns contain estimates from the husband’s log earnings equation and the wife’s log earnings equation, and have the same structure as (1). The variables listed in the table are defined in the Appendix. The demographic controls in the husband’s equation are identical to those used in the husband’s wage equation of Table 2. In the wife’s equation, the demographic controls differ from those in the wife’s wage equation of Table 2 in that the controls for the number of children present in the household are also included. A complete set of estimates is available upon request from the author. SOURCE : Author’s compilation.

recent cohorts are performed. The hypothesis that the earnings IAP over the zero to ten YSM range is the same for the 1980s and 1970s cohorts is rejected. The weighted average of the coefficients on the controls for immigrant arrival year in the 1980s implies that the 1991 earnings differential over the zero to ten YSM range is 19 percentage points smaller than the 1981 differential. 25 Also, the coefficient on C6170 is found to be 8 percentage points larger than the sum of the coefficient on C7180 and the coeffi-

cient on C7180*YR91 and this difference is statistically significant indicating that the earnings IAP of the 1960s cohort lies above the earnings IAP of the 1970s cohort over the 11 to 20 YSM range. Therefore, the results indicate that successive cohorts of immigrant men have fared worse in terms of annual earnings for given years of residence. The second and fourth columns of Table 7 contain estimates from the annual earnings equation

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

392 Christopher Worswick estimated over the sample of wives who work.26 The pre-1961 cohorts of immigrant women have 11 to 25 percent lower earnings than similar nonimmigrant women which mirrors the results of the wage equation of Table 2. The assimilation of the 1971-80 cohort of women is significant over the period at 14 percent. The tests for cohort differences discussed above are performed. The weighted average of the coefficients on the variables identifying arrival between 1981 and 1989 is found to be 10 percentage points lower than the coefficient on C7180 and this difference is statistically significant. The difference between the coefficient on C6170 and the sum of the coefficients on C7180 and C7180*YR91 is near zero and insignificant. Therefore, there is some evidence of a decline in the earnings of immigrant women from more recent cohorts relative to those from earlier cohorts, ceteris paribus; however, the magnitude of these differences is at most half of the decline found for men.

CONCLUSIONS AND IMPLICATIONS FOR IMMIGRATION POLICY In this paper, immigrant and non-immigrant married couples have been compared in terms of the wage rates, hours of work, and weeks of work of both spouses. Immigrant husbands and wives were found to have much lower wage rates than their nonimmigrant counterparts. However, immigrant wage growth over the 1980s exceeded that of nonimmigrants leading to wage rate differences between immigrants from early cohorts and similar nonimmigrants of the range 7 to 20 percent for men and 0 to 14 percent for women. Immigrants were found to work similar hours of work per week to nonimmigrants and these differences do not vary with duration of residence in Canada. The weeks of work of recent immigrants were found to be lower than the weeks of non-immigrants; however, immigrants with more years of residence have weeks of work

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

which are only 4 to 6 percent lower than nonimmigrants. The findings in this paper are consistent with certain predictions of the Family Investment Hypothesis and inconsistent with others. Consistent with the hypothesis, immigrant women are found to have higher wages relative to non-immigrant women than is true for immigrant men relative to nonimmigrant men, although the differences are small in magnitude. Initially, higher wages for immigrant women could occur if immigrant women enter jobs that are initially higher paying but offer less opportunity for career advancement. The return to education relative to non-immigrants is higher for immigrant men than women. Furthermore, the effect of young children on immigrant women’s hours of work is less than it is for non-immigrant women. These outcomes could be viewed as work-related responses by immigrant wives to a family investment strategy in which the wife (controlling for other factors) works more and foregoes investment in human capital in order to help finance her husband’s investment in host-country specific skills. Other outcomes are not consistent with the FIH. In particular, the wage growth of immigrant women relative to non-immigrant women exceeds the wage growth of immigrant men relative to non-immigrant men. Also, immigrant women do not appear to supply more hours than non-immigrant women in the first years after migration as would be predicted by the FIH. Although clear evidence of intra-family career trade-offs was not found, the results indicate that the immigrant wife plays a major role in the earnings creation of immigrant families. The lower immigrant/non-immigrant earnings differentials for wives than husbands imply that the share of family earnings contributed by the wife is larger in immigrant families than in non-immigrant families. Also, the larger decline in the immigrant/nonimmigrant earnings differentials for recent cohorts of immigrant men than immigrant women indicates

VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market that the trend is towards a still greater role for the immigrant wife in the earnings creation of the family. The findings support a family orientation to both the evaluation and the implementation of immigration policies. The relatively stronger labour market performance of recent cohorts of immigrant women than immigrant men means that evaluating the success of past immigration policies at selecting immigrants using data on men will lead to too negative a conclusion. In terms of the implementation of immigration policies, given the important role of immigrant women in earnings creation, policies should focus not only on facilitating the labour market adjustment of the principle applicant (often a man), but also assist the labour market adjustment of other family members. It may be possible to design government programs to evaluate the training needs of families rather than individuals. This would allow for the identification of the family members who are having the most difficulty finding jobs suited to their skills. Such a program could begin at the time of entry into the country when the principle applicant is required to provide both personal and family information. From the information gathered at this stage it may be possible to identify which family members are likely to have difficulty adjusting to the Canadian labour market and training support could be provided to the family. Providing assistance at this early stage could avoid delays in the family members acquiring appropriate training due to their lack of information concerning the Canadian labour market and possible problems of recognition of foreign credentials. There could be significant earnings gains for immigrant families from this kind of program given that recent immigrant women are found to have a very low return to university education relative to either native-born women or immigrant women from earlier arrival cohorts. Future research should pursue an explanation for the lower returns to education of immigrant women than non-immigrant women found in this study and in previous research. Identifying the cause of the

393

lower return on the immigrant wife’s education may lead to new policies which will facilitate the adjustment of immigrant families to Canada. A second issue which needs to be addressed is the cause of the lower weeks of work of immigrant women in the first five to ten years after migration.

NOTES I am grateful to David Green for extensive comments and suggestions. I would also like to thank Craig Riddell, Terry Wales, Charles Beach, Jeff Borland, Ted McDonald, Alice Nakamura and Jim Nason for helpful discussions. The comments of three anonymous referees were also very helpful. Research support was provided through a SSHRC doctoral fellowship and a grant from the Centre for Research in Economic and Social Policy at the University of British Columbia. 1The demographic variables are dummy variables iden-

tifying education level, place of residence and language fluency. The variables are described in detail in the Appendix. 2As is discussed below, the sample is restricted to married couples in the 1981 and 1991 Census files where both the husband and the wife were between the ages of 25 and 54 in 1981. 3The

remaining terms in the equation are: ß0 which is the intercept; ß, α, α91 , γ, γ91 and δ which are parameter vectors; and u the error term. 4To

see this, note that the expected change over the 1980s in the log wage of an immigrant husband who arrived before 1981 is:

™ W FB= Aα 91 + C1γ 91 where the vector A has appropriate values to identify the individual’s age in 1981 and the vector C1 has appropriate values to identify the immigrant’s arrival cohort. Similarly, the expected change over the 1980s in the log wage of an otherwise identical non-immigrant husband is:

™ W NB = Aα 91 Therefore, the wage assimilation for an immigrant husband is:

™ W FB− ™ W NB = C1γ 91

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

394 Christopher Worswick 5The

combination of separate age effects by birth cohort and separate years-of-residence effects by immigrant arrival cohort means that it is not possible to control for time effects (or year effects). Differences in the labour market conditions between 1981 and 1991 will be absorbed into the coefficients on the age and immigrant cohort variables. In order to identify labour market assimilation, it is assumed that possible differences in macroeconomic conditions across the survey years affect the labour market outcomes of immigrants and nonimmigrants to the same extent.

weights are .4446, .0669, .1325, .1410, .1239, and .0911, for the variables C8185, C86, C87, C88, C89 and C90, respectively. 13 The

Wald test statistic is distributed asymptotically according to the Chi-square distribution with one degree of freedom. 14 The wage IAP of the pre-1961 cohort is not compared with that of the 1961-70 cohort since the exact YSM range of the earlier cohort is not known. 15 The

6The

sample used in most studies of the labour market adjustment of immigrant men contains only men who worked full-time hours and more than 40 weeks in the reference year. Therefore, the sample restrictions used in this paper are not as selective as in previous studies and are more representative of the underlying population of men. 7The annual earnings variable from the 1991 data is deflated into 1981 dollars using the GDP deflator for Canada. 8Two

final sample restrictions are used. Households are excluded from the sample if either spouse is predominantly self-employed or is in the military. 9The

coefficient estimates are distributed asymptotically according to the Normal distribution. White’s (1980) heteroskedasticity consistent covariance matrix is used. 10 To

be precise, this is the wage differential for immigrants in the education default category of completed high school. The differential returns to education for immigrants and non-immigrants are discussed below. The wage, hours, and weeks differences for immigrants were also compared at both the immigrant and non-immigrant mean values of education as suggested by Borjas (1985). Since the results do not depend in general on the point of comparison, the immigrant/non-immigrant differentials holding education at the default level are the focus of discussion in this paper. 11Given

the broad arrival year categories, this is as close to a test of the wage IAP’s intercept being the same for the two cohorts as is possible. 12 Each weight is the ratio of the number of immigrant men in the sample who have a value of one for the cohort dummy variable, divided by the total number of immigrant men in the sample who arrived in the 1980s. The

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

wife’s equation is estimated using a two-step selection correction method (see Heckman 1979). The first stage Probit index is defined using both the wife’s and the husband’s variables. The results of the Probit estimation are available on request from the author. The coefficient on the inverse Mill’s ratio variable is positive and significant at the one percent level indicating that women with higher wage offers are more likely to be in the labour market. The equation was reestimated by OLS ignoring the selection into the labour market with only minor changes in the coefficients of the wage regression. 16 Each

weight is the ratio of the number of immigrant women in the sample who have a value of one for the cohort dummy variable, divided by the total number of immigrant women in the sample who arrived in the 1980s. The weights are .4814, .0629, .1216, .1472, .1100, and .0769, for the variables C8185, C86, C87, C88, C89 and C90, respectively. 17A

family level wage ratio equation was also estimated, where the dependent variable is the natural logarithm of the ratio of the husband’s wage to the wife’s wage. The exogenous variables included controls for both the husband’s and wife’s demographic characteristics as well as controls for both the husband’s and the wife’s arrival cohort. In principle, this specification allows for the measurement of the effects of one spouse’s arrival cohort and YSM holding the arrival cohort and YSM of the other spouse fixed. However, due to the high collinearity of the husband’s arrival year and the wife’s arrival year and the fact that broad arrival year blocks are used in the raw data, very few of the coefficients on the arrival cohort variables were significant from zero. Equivalent family equations were also estimated where the dependent variable was the natural logarithm of the ratio of the husband’s to wife’s hours, weeks or annual earnings. In each case, the standard errors on the cohort variables were large

VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market and the coefficients on the immigrant controls were insignificant in virtually all cases. Therefore, the estimates of these equations are not included in the paper. 18 The

weights are the same as those used in the analogous test of section four. 19As in the previous section, a two-step estimation procedure is used to address the selection into employment by the wives. The log hours equation was reestimated by OLS (ignoring the selection into the labour market) with only minor changes in the estimates. 20 The

tests are analogous to the ones described above for the husbands’ equation. The weights are the same as the ones used in the equivalent test for wives in section four. 21 The

coefficient on C80 is not used in the analysis to eliminate the truncation of the weeks variable for that cohort. For the same reason, the coefficient on C90 is not included in the weighted average. Therefore, the weights are .4892, .0736, .1457, .1551, and .1364 for the variables C8185, C86, C87, C88, and C89, respectively. 22 The

Wald test statistic is distributed asymptotically according to the Chi-square distribution with one degree of freedom. 23 Again,

a two-step procedure is used to address the endogeneity of the wife’s participation decision. The Probit specification is the same as the one used for the wage and hours regressions. 24 The coefficient on C90 was not included in the average to eliminate the truncation of weeks of that cohort. Therefore, the weights are .5215, .0681, .1317, .1595, and .1192 for the variables C8185, C86, C87, C88, and C89, respectively. 25 The

coefficient on C80 is not used in the analysis to eliminate the truncation of the earnings variable for that cohort. For the same reason, the coefficient on C90 is not included in the weighted average. The weights are the same as those used in the analogous test for the weeks equation. The Wald test statistic is distributed asymptotically according to the Chi-square distribution with one degree of freedom. 26 Again,

a two-step procedure is used to address the endogeneity of the wife’s participation decision. The Probit specification is the same as the one used for the wage and hours regressions.

395

REFERENCES Abbott, M.G. and C.M. Beach (1993), “Immigrant Earnings Differentials and Birth-year Effects for Men in Canada: Postwar-1972,” Canadian Journal of Economics 26:505-24. Baker, M. and D. Benjamin (1994), “The Performance of Immigrants in the Canadian Labour Market,” Journal of Labor Economics 12: 369-405. ____ (1996), “The Role of Family in Immigrants’ Labour Market Activity: An Evaluation of Alternative Explanations,” American Economic Review 86. Beach, C.M. and C. Worswick (1993), “Is There a Double-Negative Effect on the Earnings of Immigrant Women?” Canadian Public Policy 19:36-53. Bloom, D. and M. Gunderson (1991), “An Analysis of the Earnings of Canadian Immigrants,” in Immigration, Trade and the Labor Market, ed. J. Abowd and R. Freeman (Chicago: University of Chicago Press for the NBER). Borjas, G.J. (1985), “Assimilation, Change in Cohort Quality, and the Earnings of Immigrants,” Journal of Labor Economics 3:463-89. Borjas, G.J. and S. Bronars (1991), “Immigration and the Family,” Journal of Labor Economics 9:123-49. Campbell, M. (1994), “Too Many Immigrants, Many Say,” The Globe and Mail, 10 March. Chiswick, B.R. (1978), “The Effect of Americanization on the Earnings of Foreign-Born Men,” Journal of Political Economy 86: 897-921. Citizenship and Immigration Canada (1995), “Canada Changes Immigration Criteria for Skilled Workers,” Press Release, 17 November. Duleep, H. and S. Sanders (1993), “The Decision to Work by Married Immigrant Women,” Industrial and Labor Relations Review 46:677-90. Heckman, J.J. (1979), “Sample Selection Bias as a Specification Error,” Econometrica 47:153-61. LaLonde, R. and R. Topel (1992), “Immigrants in the American Labor Market: Quality, Assimilation, and Distributional Effects,” AEA Papers and Proceedings, pp. 297-302. Long, J. (1980), “The Effect of Americanization on Earnings: Some Evidence for Women,” Journal of Political Economy 88:620-29. Simpson, J. (1994), “Immigration Policy Needs to Balance Families, Refugees and Independents,” The Globe and Mail, 26 October. White, H. (1980), “A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test of Heteroskedasticity,” Econometrica 48:817-38.

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

396 Christopher Worswick APPENDIX DEFINITIONS OF VARIABLES LISTED IN TABLES

OTHER CONTROLS

1.

WAGE: hourly wage rate (derived by dividing the product of hours worked in the reference week and weeks worked in the reference year into annual earnings in the reference year). Deflated into 1981 dollars using the GDP deflator for Canada.

2.

HOURS: hours of work in the reference week.

3.

WEEKS: weeks of work in the reference year.

4.

PRIMARY: has zero to eight years of education.

5.

SECOND: has nine to thirteen years of educa-

The following variables are included in all equations. In the husband’s equations the variables relate to the husband and in the wife’s equation they relate to the wife: dummy variables identifying: (i) place of residence in the Atlantic provinces (New Brunswick, Newfoundland, Nova Scotia, and Prince Edward Island), Quebec outside Montreal, Montreal, Ontario outside Toronto, Toronto (default), the Prairie provinces (Manitoba, Saskatchewan, and Alberta), British Columbia outside Vancouver, and Vancouver; and (ii) language fluency in English but not French (default), French but not English, and both English and French.

tion but has not graduated from high school. 6.

HSGRAD: has graduated from high school.

7.

POSTSEC: holds a postsecondary certificate or

diploma. 8.

BACH: holds a bachelor’s degree.

9.

GRAD: holds a professional or graduate university degree.

10. FB : Immigrant. 11. CPRE61, C6170, C7180, C80, C8185, C86, C87, C88, C89, C90: Indicator variables for immigrants whose year of arrival was before 1961, between 1961-70, 1971-80, in 1980, between 1981-85, in 1986, 1987, 1988, 1989 and 1990, respectively.

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

Also, in the wife’s wage equation the number of live births is also included as a regressor. In the wife’s hours equation and weeks equation controls for the number of children present in the household are also included as regressors (see Table 5 for their definitions). In the wife’s earnings equation, both the number of live births variable and the controls for the presence of children are included as regressors.

VOL . XXII, NO . 4 1996

CHRISTOPHER WORSWICK Department of Economics University of Melbourne Melbourne, Australia

Nous comparons les activités des couples d’immigrés et de non-immigrés sur le marché du travail à l’aide de données de 1981 et 1991 de Recensement Canada. Nous donnons de nouvelles indications sur la performance d’immigrés masculin et féminin en termes de trois composantes annuelle de leurs revenus: les salaires horaires, le nombre d’heures travaillées par semaine et le nombre de semaines travaillées annuellement. Nous ne trouvons aucune indication qui soutiendrait l’argument d’un arbitrage entre la carrière du mari immigré au détriment de la femme. Cependant, la performance de l’épouse sur le marché du travail joue un rôle important dans la création de revenus des familles d’immigrés. Ces résultats vont dans le sens d’une évaluation et d’une mise en place de politiques d’immigration qui favorisent la famille. The labour market activity of immigrant and non-immigrant married couples is compared using data from the 1981 and 1991 Canadian Censuses. New evidence is provided on the performance of immigrant men and women in terms of three components of annual earnings: hourly wage rates, hours worked per week, and weeks worked per year. Evidence of intra-family trade-offs of investments in the immigrant husband’s career at the expense of investments in the wife’s career are not found overall. However, the wife’s labour market performance is found to play a major role in the earnings creation of immigrant families. The findings support a family orientation to both the evaluation and the implementation of immigration policies.

INTRODUCTION

I

n this paper, the labour market performance of immigrant families is analyzed. The success of Canada’s immigration policy is currently under debate in terms of the number of immigrants admitted each year and the composition of the inflow between independent class immigrants, where admission is based on the applicant’s skills, and family class immigrants, where admission is based primarily on the applicant having relatives in Canada (Campbell 1994; Simpson 1994). Given this debate and the recent changes announced to the point system used to select immigrants (Citizenship and Immigration Canada 1995), it is important to gain a full understanding of how the selected immigrants are sucCANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

ceeding in the Canadian labour market. Also, given that the immigration decision is often based upon what is best for the immigrant family rather than a particular family member, it is important to see how different immigrant family members are performing in the labour market. Economic studies of immigrant labour market adjustment have focused on differences in earnings between immigrants and non-immigrants as a function of the immigrant’s years-since-migration (YSM) and arrival cohort. The results for men indicate that immigrant men have lower earnings than similar non-immigrant men immediately after migration. However, the immigrant men experience higher growth in earnings over time with the VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market magnitude of this difference being the focus of much of the debate in the literature (see Chiswick 1978; Borjas 1985; Bloom and Gunderson 1991; LaLonde and Topel 1992; Baker and Benjamin 1994). This paper significantly extends this work by examining the labour force decisions of cohorts of immigrant married women and men in tandem. A specific focus is to determine whether and how the labour force decisions of immigrant wives are related to the human capital investment patterns of immigrant husbands. In what has become known as the Family Investment Hypothesis (FIH), several researchers have argued that financing investments in host-country skills by immigrant husbands is a factor affecting the labour force decisions and human capital investments of immigrant married women (see Long 1980; Beach and Worswick 1993; Duleep and Sanders 1993; Baker and Benjamin 1996). Associated with this general hypothesis are the following specific expectations: in relation to their husbands’ investments in host-country specific skills, immigrant women would be more likely to work and to work longer hours and more weeks than would otherwise be the case. Conditional on their own levels of host-country-specific skills, immigrant women would also be less likely to undertake human capital investments in order to take jobs that paid more during the period in which the husband’s investment in host-country skills was most intense. Subsequently, the wages of immigrant women may be higher than would otherwise be the case during the period in which the husband is investing. By foregoing investment, the wage-age profile of immigrant women would be expected to be flatter than otherwise. Using data from the 1970 US Census, Long (1980) finds that immigrant women with few years of residence in the United States work more hours and have higher earnings than both similar nonimmigrant women and immigrant women with more years of residence. Long argues that these earnings patterns of immigrant women could be explained under the FIH. Using the Canadian Job Mobility

379

Survey, Beach and Worswick (1993) find, in sharp contrast to what had been previously found for men, that the immigrant/non-immigrant earnings differential for women does not depend on the immigrant’s years of residence. It is argued that this may be explained by the FIH. Two data limitations hamper each of these studies: (i) the availability of only a single cross-section survey, and (ii) limited information on the labour supply of the immigrant women. The availability of only one cross-section means that each cohort of immigrant women is observed at only one value of YSM. The poor controls available for labour supply means that it is impossible to distinguish between the dynamic returns to an immigrant’s work time with more YSM and the family’s labour supply decision for the immigrant across time. Duleep and Sanders (1993) use the 1980 US Census and find that the labour force participation rate is higher for immigrant women whose husbands have difficulty adapting to the new labour market and may need to make investments in skills specific to the US labour market. Baker and Benjamin (1994) use the 1986 and 1991 Surveys of Consumer Finance for Canada and estimate a reduced-form labour supply model for married couples. They find evidence that immigrant women increase their hours of work in order to finance the labour market adjustment of their husbands. However, the analysis is hampered by the fact that the time frame over which each cohort is observed is only five years; therefore, the long-term dynamics of labour market adjustment in immigrant families are not observed. Implicit in these studies are two basic approaches to the analysis of the FIH. One approach, as exemplified by Long (1980) and Beach and Worswick (1993), has been to compare the patterns of work and earnings behaviour of immigrant married women relative to non-immigrant women with the patterns of work and earnings behaviour of immigrant and non-immigrant men found in previous studies. Another approach has been to identify the family investment model by using variation in the

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

380 Christopher Worswick likely extent to which immigrant husbands invest in host-country human capital, while holding constant the wife’s own level of host-country skills as measured by her own years-since-migration and level of host-country language proficiency. This approach is adopted by Duleep and Sanders (1993) and Baker and Benjamin (1994). This paper adopts the former strategy. The pattern of work and earnings behaviour of the husband relative to the native-born is compared to the patterns of work and earnings behaviour of the immigrant wife relative to the native-born. An important improvement over the Long (1980), and Beach and Worswick (1993) efforts is that data are used from more than one time period. In addition, this study analyzes the labour market behaviour of immigrant married men and women, relative to their nonimmigrant counterparts, in tandem, rather than comparing the labour market behaviour of married immigrant women to the findings of the literature on immigrant men. In addition to analyzing the annual earnings of immigrant husbands and wives, this paper also analyzes the labour market adjustment of immigrant husbands and wives along the following dimensions: the hourly wage rate, the hours of work per week, and the weeks of work per year.

The remainder of the paper is structured as follows. In the second section the basic regression specifications are described. In section three the data set and estimation sample used are presented. In the remaining sections, results are presented from wage, hours, weeks, and annual earnings regressions for immigrant and non-immigrant husbands and wives. The conclusions and implications for public policy are discussed in the final section.

REDUCED-FORM EQUATIONS AND SPECIFICATION ISSUES To measure the labour market adjustment of immigrants, the labour market outcomes of similar nonimmigrants are used as a benchmark and the deviations of immigrant labour market outcomes from the non-immigrant outcomes are analyzed. Figure 1 demonstrates possible wage adjustment profiles of two immigrant arrival cohorts. The vertical axis is the difference between the log wage of an immigrant and the log wage of a non-immigrant, holding other characteristics the same. The horizontal axis measures the number of years-since-migration (YSM) of the immigrant. The solid line and the dashed line are the Immigrant Adjustment Paths

FIGURE 1 Immigrant Adjustment Path (IAP)

SOURCE : Author’s compilation.

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market (IAP) for the wage rates of immigrants from the 1960s arrival cohort and the 1970s cohort, respectively. To estimate these lines, data are used on wage rates for non-immigrants and for immigrants from these two arrival cohorts across time. The data used in this paper come from two crosssectional surveys, the 1981 and 1991 Census of Canada Family Microdata Files. Three immigrant arrival cohorts can be identified in both surveys: arrival before 1961, arrival between 1961 and 1970, and arrival between 1971 and 1980. In the 1991 sample, immigrants from these three cohorts can be identified as well as immigrants from the following six arrival cohorts: arrival between 1981 and 1985, and arrival in 1986, 1987, 1988, 1989 and 1990. Eight regression equations are estimated in the paper: log wage, log hours, log weeks and log earnings equations for both the husband and the wife. All eight of these equations have the same structure; therefore, consider the husband’s illustrative log wage equation:

In W = ß 0 + Xß+ Aα + (A ∗ Y 91 )α 91 + C1γ + (C1 ∗ Y 91 )γ 91 + C2δ + u (1) where lnW is the log of the husband’s hourly wage. The vector X contains: (i) demographic variables of the husband and (ii) interactions of these demographic variables with the immigrant dummy variable, FB, allowing for separate effects of demographic characteristics on the husband’s wage for immigrants and non-immigrants. 1 The vector A contains age dummy variables identifying whether the person was age 25-29, 30-34, 35-39, 40-44 (default), 45-49 or 50-54 in 1981.2 These dummy variables are also included as interactions with the 1991 survey year dummy variable, Y91. This specification allows for different effects of age on the husband’s wage for each of the six age-in-1981 cohorts, in each survey year. Interpreted in terms of a wage-age profile, this specification leads to estimates of two points on the wage-age profile of each of the six age-in-1981 cohorts (one point from the 1981 Census, the other from the 1991 Census).

381

The regression specification employs the identifying assumption of age effects on the husband’s wage being the same for immigrants and nonimmigrants. This enables the identification of yearssince-migration effects on the wages of immigrants. The vector C1 contains three dummy variables identifying the three immigrant arrival cohorts (pre1961, 1961-70, and 1971-80) which appear in both the 1981 Census and the 1991 Census. These dummy variables are also included as interactions with the 1991 survey year dummy variable. The vector C2 contains six dummy variables identifying immigrants from the six immigrant arrival cohorts who arrived in Canada after 1980 (1981-85, 1986, 1987, 1988, 1989, and 1990). This specification allows for different effects of arrival cohort on the husband’s wage for each arrival cohort in each survey year. Interpreted in terms of a wage IAP, this specification leads to estimates of two points on the wage IAP of each of the three pre-1981 cohorts, and one point on the wage IAP profile for each of the six post-1980 cohorts.3 It is common in the immigration literature to refer to the difference between the wage growth of a cohort of immigrants and the wage growth of nonimmigrants, ceteris paribus, as the wage “assimilation” of that cohort of immigrants. In the wage equation, the vector of coefficients γ91 measures the wage assimilation from 1980 to 1990 of each of the three cohorts of immigrants who arrived in Canada before 1981.4 In terms of Figure 1, the three assimilation coefficients in γ91 measure the movement along the IAP (or the slope of the IAP) over the relevant years-since-migration values for each of the three arrival cohorts.5

THE DATA AND ESTIMATION SAMPLE In the absence of suitable panel data, researchers have used synthetic cohort data in the analysis of immigrant labour market adjustment. The synthetic cohorts are two or more cross-sectional surveys where the samples can be defined so as to represent

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

382 Christopher Worswick the same underlying population at different points in time. The Canadian Census data used in this analysis come from the Family Files which contain 94,745 census families from 1981 and 345,351 census families from 1991. Since the 1991 data are a larger random sample of the population than the 1981 data, weights are used in the analysis to account for the relative oversampling in 1991. Due to the large numbers of non-immigrant households in the two data-sets, a 25 percent random sample is taken of non-immigrant households. The estimation sample contains married couples with both spouses having been age 25 to 54 in 1981. Households where one spouse is an immigrant and the other spouse was born in Canada are excluded from the sample. Duleep and Sanders (1993) also focus on households where both spouses are immigrants. Family trade-offs in terms of one spouse working to subsidize the other spouse’s career are more likely to be necessary in households where both spouses are immigrants. In households where only one spouse is an immigrant, the non-immigrant spouse is already established in Canada with Canadian education credentials, knowledge of at least one of the official languages, and in most cases the nonimmigrant spouse has work experience with Canadian employers. The sample is restricted to households where the husband works in both the reference week and the reference year. This is standard in the economic literature on the labour market adjustment of immigrant men.6 Wage rates are constructed by dividing the product of the reference weekly hours variable and the weeks-worked-per-year variable into the individual’s annual earnings the previous year.7 This is an imperfect measure of the true wage rate offered to the individual in that the weekly hours variable refers to the reference week and not the average weekly hours of the reference year. To construct the wage variable, it was necessary to exclude from the sample households where either spouse had positive earnings and weeks in the reference year, and zero hours in the reference week, or vice versa.8 CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

Table 1 contains sample means of selected variables over the immigrant and non-immigrant samples in each survey. The average wages of immigrant women are $1.37 lower than those of the nonimmigrant women in 1981, but this wage gap has dropped to $0.89 by 1991. Immigrant women work 2.2 hours more per week in 1981 than nonimmigrant women and 1.85 more hours per week in 1991. Immigrant women work one-third of a week more on average in 1981 and less than one week more in 1991. Immigrant men have $0.83 lower wages than non-immigrant men in 1981, but by 1991 the difference has decreased to $0.25. The mean hours per week and weeks per year patterns of immigrant men are very similar to those of nonimmigrant men.

THE WAGE RATES OF IMMIGRANT AND NON-IMMIGRANT COUPLES Table 2 contains estimates from log wage regressions for husbands and wives. The first and third columns list results from the log wage equation estimated over the sample of husbands. 9 The first three coefficient estimates give the estimated difference in log wages in 1981 between immigrant men and non-immigrant men, ceteris paribus, for each of the three immigrant cohorts evaluated at high school completion.10 In terms of Figure 1, each of these coefficient estimates is an estimate of the height of the cohort’s IAP over the relevant range of YSM. Immigrant men who arrived in the 1970s have 28 percent lower wages in 1981 than similar nonimmigrant men, while those arriving earlier have 11 to 15 percent lower wages with each of these differences significant at the 5 percent level. The next three coefficient estimates give the wage assimilation of immigrant men for each immigrant cohort. In terms of Figure 1, these coefficients measure the slope of the IAP between the two survey years. The estimated assimilation of the pre-1961 cohort and the 1971-80 cohort is 8 and 9 percent, respectively, and statistically significant, while the assimilation of the 1961-70 cohort is not significant. VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market

383

TABLE 1 Sample Means for Selected Variables

Immigrants 1981 N=3688

Immigrants 1991 N=10472

Non-Immigrants 1981 N=3808

Non-Immigrants 1991 N=9815

8.00 34.67 43.37 .2400 .1883 .1280 .3364 .0848 .0225

10.03 35.47 46.71 .2042 .1621 .1598 .3346 .1038 .0355

9.37 32.47 43.04 .1253 .2661 .1895 .3364 .0734 .0093

10.92 33.64 46.11 .0877 .2171 .2107 .3732 .0935 .0178

11.36 42.87 47.56 .1956 .1261 .0576 .4431 .1099 .0677

13.29 42.07 48.25 .1748 .1309 .0885 .3965 .1277 .0816

12.09 42.69 48.26 .1322 .1988 .1238 .3994 .1099 .0359

13.54 42.31 48.38 .1110 .1960 .1372 .3933 .1185 .0440

Wife’s Variables WAGE* HOURS* WEEKS* PRIMARY SECOND HSGRAD POSTSEC BACH GRAD

Husband’s Variables WAGE HOURS WEEKS PRIMARY SECOND HSGRAD POSTSEC BACH GRAD

NOTE: *The mean is calculated over the sample of women who worked. SOURCE : Author’s compilation.

The next six coefficient estimates give the estimated differences in log wages in 1991 between immigrant men who arrived after 1980, and nonimmigrant men, ceteris paribus. In the 1991 survey, the immigrant men from these cohorts have 36 to 61 percent lower wages than similar non-immigrant men and each of these differences is significant. A test is performed of the hypothesis that the average value of the wage IAP over the zero to ten YSM range is the same for the 1980s cohort and the 1970s cohort.11 The coefficient estimate for the variable C7180 is the estimate of the wage IAP for the 1970s

cohort over the YSM range of zero to ten years. A weighted average of the coefficients on the controls for immigrant arrival year in the 1980s is derived (because of the detailed arrival year breakdown in the 1991 Census) and the coefficient on C7180 is subtracted off of the weighted average to give the difference in the IAP over the zero to ten range of YSM for the two cohorts.12 The immigrant/nonimmigrant wage differential over zero to ten YSM is found to be 14 percentage points lower for the 1980s cohort compared with the 1970s cohort and this difference is statistically significant.13 A second

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

384 Christopher Worswick TABLE 2 Immigrant/Non-Immigrant Wage Differences by Immigrant Cohort for Husbands and Wives

Husband

Wife

Husband

Wife

CPRE61

-.1474** (.034)

-.0880 (.053)

C8185

-.3550** (.041)

-.2773** (.052)

C6170

-.1094** (.033)

-.1531** (.050)

C86

-.3676** (.078)

-.2666** (.095)

C7180

-.2759** (.035)

-.2551** (.049)

C87

-.3778** (.064)

-.3223** (.074)

CPRE61*YR91

.0777** (.029)

.0816* (.041)

C88

-.6148** (.079)

-.3718** (.070)

C6170*YR91

.0303 (.025)

.0816* (.035)

C89

-.4503** (.066)

-.3422** (.077)

C7180*YR91

.0876** (.028)

.1136** (.035)

C90

-.4983** (.080)

-.4264** (.088)

R2 F N

.083 47.6 27,783

.105 39.4 19,558

NOTES: 1. Standard errors are in parentheses. 2. ** denotes significance at the 1 percent level; * denotes significance at the 5 percent level. 3. The first two and last two columns contain estimates from the husband’s log wage equation and the wife’s log wage equation, and have the same structure as (1). The variables listed in the table are defined in the Appendix. Dummy variables are included in the equations identifying: (i) the education categories 0-8 years, 9-13 years without a high school diploma, a high school diploma (default), completed a bachelor’s level university degree, and completed a university degree above the bachelor’s level; (ii) place of residence in the Atlantic provinces, Quebec outside Montreal, Montreal, Ontario outside Toronto, Toronto (default), the Prairie provinces, British Columbia outside Vancouver, and Vancouver; (iii) language fluency in English but not French (default), French but not English, and both English and French. In the wife’s equation, the number of live births is also included. A complete set of estimates is available upon request from the author. SOURCE : Author’s compilation.

test of a decline in the wage IAP of immigrant and non-immigrant men is performed. The coefficient on C6170 is subtracted from the sum of the coefficients on C7180 and C7180*YR91. This difference is statistically significant and implies that the wage IAP of the 1980s cohort of immigrant men is 8 percentage points lower than that of the 1970s cohort over the 11 to 20 YSM range.14 CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

These results are consistent with the findings of Borjas (1985) for the United States and Baker and Benjamin (1994) for Canada. Immigrant men have significantly lower wages than non-immigrant men at time of arrival and they experience low rates of wage assimilation. Also, the immigrant-nonimmigrant wage differential, holding YSM constant, has worsened for more recent arrival cohorts. VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market The results in the second and fourth columns of Table 2 are from the wage regression estimated over the sample of immigrant women who work.15 The general specification is the same as the one used in the husband’s equation with the wife’s variables replacing those of the husband. Controls for the number of live births of the wife are also included as a proxy for time spent out of the labour market due to caring for young children. From the first three coefficient estimates, we see that in the 1981 data, immigrant women who arrived after 1960 have 15 to 26 percent lower earnings than similar non-immigrant women, while the women who arrived before 1960 do not have significantly different wages than non-immigrant women. The wage assimilations of the three pre-1981 arrival cohorts are 8 to 11 percent and each estimate is significant at the 5 percent level. Therefore, the estimates of the wage assimilation of the immigrant women from these arrival cohorts are larger than what were found for the same cohorts of immigrant men. Under the FIH, one would expect the opposite to be the case. The theory predicts that immigrant wives will have lower wage assimilation than their husbands since they are sacrificing career opportunities in order to accept a job which supports their families while their husbands become established in their careers. From the coefficients on the controls for arrival after 1980, we see that the 1991 wage rate for these cohorts is 27 to 43 percent lower than the wage rate for similar non-immigrant women and these differences are significant. The test is performed of the hypothesis that the average value of the wage IAP over the zero to ten YSM range is the same for the 1980s arrival cohort of immigrant women as the 1970s cohort. The coefficient on C7180 is subtracted off of a weighted average of the coefficients on the controls for arrival after 1980 and the difference is not statistically significant.16 The test of the equality of the wage IAP of the 1960s cohort and the 1970s cohort over the 11 to 20 YSM range is performed for women. The coefficient on C6170 is not significantly different from the sum of the coeffi-

385

cients on C7180 and C7180*YR91, indicating that the issue of declining unexplained productivity of immigrant cohorts does not appear to be a concern for immigrant married women. The first and third columns of Table 3 contain the estimates from the coefficients on the education controls from the log wage equation of the husbands. These coefficients appear in the parameter vector β in equation (1) and represent the returns to education for immigrant and non-immigrant husbands. The wages of non-immigrant men are increasing in their level of education with a particularly large and significant return of 42 to 48 percent to a university degree relative to having only a high school degree. Given the problems of having education credentials recognized after migration (Chiswick 1978), one would expect a flatter wage-education profile for immigrants. However, the results do not support this hypothesis. The returns to education for immigrant men are generally the same as those for nonimmigrant men with the exception of a lower return to a bachelor’s degree for immigrant men. This latter relationship is consistent with the immigrant men having difficulty having foreign educational credentials recognized as equivalent to Canadian credentials by Canadian employers. The wage equations were reestimated replacing the FB*BACH variables with interactions of the BACH variables with each of the immigrant arrival cohort variables. This allows separate returns to a bachelor’s degree according to arrival cohort. For immigrant men, the coefficients on these interaction variables were all close in magnitude and the hypothesis that they were equal could not be rejected. In the women’s equation, immigrant women who arrived in the 1980s had a lower return to a bachelor’s degree than did either non-immigrant women or immigrant women from earlier cohorts. This indicates that the problem of immigrant women having their bachelor’s degrees accepted in the Canadian labour market as equivalent to those of native-born women is largest for the most recently arrived immigrant women. From the estimates of the second and fourth columns of Table 3, we see that the wage-education

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

386 Christopher Worswick profile of non-immigrant women is steeper than what was found for non-immigrant men. For example, there is a 61 to 74 percent higher return to a university education compared to a high school diploma. Unlike what was found for men, immigrant women have a significantly flatter wage-education profile than do non-immigrant women. Immigrant women in the lowest education category have significantly higher offered wages than do nonimmigrant women, ceteris paribus. As was the case for immigrant men, immigrant women with a bachelor’s degree have significantly lower wages than similar non-immigrant women. Similar lower rates of return on education for immigrant women compared with non-immigrant women are found by Long (1980), and Beach and Worswick (1993).

The lower rate of return on education of immigrant women coupled with the fact that immigrant men face the same rate of return on their education as non-immigrant men are consistent with the FIH. If the husband is perceived to be the main earner, it may be that the family invests in job search and retraining for the husband so that he can find a job that suits his education, while the wife settles quickly in a job which may not suit her education but allows her to fund family consumption. However, an alternative explanation is that the family’s migration decision may be more closely related to the husband’s career than the wife’s career. In this case, one would expect the immigrant husband’s education to be more suited to the new labour market than the wife’s education, leading to the

TABLE 3 Estimated Returns to Education for Husbands and Wives

Husbands

Wives

Husbands

Wives

PRIMARY

-.1463** (.024)

-.3095** (.039)

FB*PRIMARY

.0476 (.030)

.1531** (.046)

SECOND

-.0314 (.020)

-.1188** (.024)

FB*SECOND

-.0427 (.026)

.0936** (.036)

POSTSEC

.1129** (.015)

.2267** (.022)

FB*POSTSEC

.0057 (.015)

-.0380 (.031)

BACH

.4150** (.022)

.6106** (.030)

FB*BACH

-.0992** (.028)

-.1414** (.042)

GRAD

.4808** (.029)

.7398** (.058)

FB*GRAD

.0121 (.034)

-.0949 (.072)

NOTES : 1. Standard errors are in parentheses. The regression specification is described in the notes of Table 2. 2. ** denotes significance at the 1 percent level; * denotes significance at the 5 percent level. 3. The default education category contains individuals who have completed high school but do not have postsecondary training. SOURCE : Author’s compilation.

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market differential rates of return on their schooling relative to those of their non-immigrant counterparts.17

387

THE HOURS OF WORK OF IMMIGRANT AND NON-IMMIGRANT MARRIED COUPLES

ally insignificant at the 5 percent level indicating that the hours of work per week of immigrant men do not differ from those of non-immigrant men, ceteris paribus. In particular, the hours assimilation is not significant from zero for each of the three pre1981 cohorts.

Table 4 contains the results from the log hours equations for husbands and wives. The results from the husbands’ equation are presented in the first and third columns. With the exception of the coefficient on C89, all of the coefficient estimates are individu-

The test of the hypothesis that the average values of the hours IAPs of the 1980s cohort and the 1970s cohort are equal over the zero to ten YSM range is performed. The coefficient on C7180 is subtracted from a weighted average of the

TABLE 4 Immigrant/Non-Immigrant Hours Differences by Immigrant Cohort for Husbands and Wives

Husband

Wife

Husband

Wife

CPRE61

.0050 (.012)

-.0010 (.029)

C8185

-.0177 (.016)

.0532 (.032)

C6170

-.0078 (.013)

.0544 (.031)

C86

-.0096 (.032)

.0528 (.060)

C7180

-.0156 (.014)

.0406 (.032)

C87

-.0275 (.025)

.0367 (.047)

CPRE61*YR91

.0065 (.012)

-.0013 (.026)

C88

-.0510 (.028)

.0446 (.044)

C6170*YR91

.0012 (.010)

-.0207 (.023)

C89

-.0623** (.026)

.0396 (.048)

C7180*YR91

-.0010 (.011)

.0082 (.023)

C90

-.0480 (.036)

.0552 (.056)

R2 F N

.011 5.85 27,783

.044 14.3 19,558

NOTES : 1. Standard errors are in parentheses. 2. ** denotes significance at the 1 percent level; * denotes significance at the 5 percent level. 3. The first two and last two columns contain estimates from the husband’s log hours equation and the wife’s log hours equation, and have the same structure as (1). The variables listed in the table are defined in the Appendix. The demographic controls in the husband’s equation are identical to those used in the husband’s wage equation of Table 2. In the wife’s equation, the demographic controls differ from those in the wife’s wage equation of Table 2 in that the controls for the number of live births are replaced by controls for the number of children present in the household (see Table 5). A complete set of estimates is available upon request from the author. SOURCE : Author’s compilation.

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

388 Christopher Worswick coefficients on the controls for immigrant arrival in the 1980s.18 The difference equals negative two percentage points but it is not statistically significant. Also, the hypothesis that the coefficient on C6170 equals the sum of the coefficients on C7180 and C7180*YR91 could not be rejected. Therefore, the immigrant/non-immigrant hours differential, holding YSM constant, has not changed significantly for recent arrival cohorts. The second and fourth columns give the estimates from the log hours equation for women.19 The results are very similar to those found for men. All of the immigrant cohort controls are insignificant indicating that immigrant women’s hours are not significantly different from the hours of non-immigrant women, ceteris paribus. The two tests of the hypothesis that the hours IAPs of recent cohorts of immigrant women are equal to the hours IAPs of earlier cohorts are performed and the hypothesis is not rejected in either case.20

In Table 5, estimates are presented of the coefficients on the controls for presence of children in the household from the hours equation for wives. Non-immigrant wives are found to reduce their hours of work when young children are present in the household. This effect is smaller for immigrant women. This is a common finding in studies of the labour market adjustment of immigrant women (see Long 1980; Beach and Worswick 1993; Baker and Benjamin 1996). It may be that adult relatives are more likely to be present in immigrant families than in non-immigrant families. Unfortunately, this information is not available in the data.

THE WEEKS OF WORK OF IMMIGRANT AND NON-IMMIGRANT MARRIED COUPLES Table 6 contains the results from the log weeks regressions. A dummy variable is included in each regression identifying immigrants who arrived in the

TABLE 5 Estimated Response of the Hours of the Wife to the Presence of Children

Interactions with FB variable One child present under age 6

-.0828** (.022)

.0570* (.023)

More than one child present under age 6

-.1525** (.043)

.1266** (.043)

Number of children present age 6-14 (maximum of two)

-.0577** (.008)

.0325** (.010)

More than two children present age 6-14

-.0924** (.030)

-.0100 (.038)

NOTES : 1. Standard errors are in parentheses. 2. ** denotes significance at the 1 percent level; * denotes significance at the 5 percent level. 3. The regressions are described in the notes to Table 4. SOURCE : Author’s compilation.

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market

389

TABLE 6 Immigrant/Non-Immigrant Weeks Differences by Immigrant Cohort for Husbands and Wives

Husband

Wife

Husband

Wife

CPRE61

-.0407** (.012)

-.0460 (.027)

C8185

-.0438** (.015)

-.0363 (.033)

C6170

-.0439** (.011)

-.0291 (.027)

C86

-.0861 (.045)

-.1251 (.067)

C7180

-.0734** (.014)

-.0869** (.028)

C87

-.0523* (.024)

-.0906 (.051)

C80

-.5616** (.092)

-.6422** (.054)

C88

-.0848** (.025)

-.1478** (.047)

CPRE6*YR91

.0330** (.012)

.0106 (.023)

C89

-.2888** (.049)

-.2416** (.053)

C6170*YR91

.0155 (.010)

-.0013 (.019)

C90

-.6060** (.065)

-.5905** (.062)

C7180*YR91

.0369** (.013)

.0625** (.019)

R2 F N

.060 32.6 27,783

.066 21.8 19,558

NOTES : 1. Standard errors are in parentheses. 2. ** denotes significance at the 1 percent level; * denotes significance at the 5 percent level. 3. The first two and last two columns contain estimates from the husband’s log weeks equation and the wife’s log weeks equation, and have the same structure as (1). The variables listed in the table are defined in the Appendix. The demographic controls in the husband’s equation are identical to those used in the husband’s wage equation of Table 2. In the wife’s equation, the demographic controls differ from those in the wife’s wage equation of Table 2 in that the controls for number of live births are replaced by controls for the number of children present in the household. A complete set of estimates is available upon request from the author. SOURCE : Author’s compilation.

1980 reference year. This is intended to pick up the effect on the individual’s weeks of work of the fact that he/she arrived after the start of the year. The coefficients on C80 and C90 are found to be negative and large in magnitude. This is due to the immigrants in these cohorts having a truncated weeks of work for the year.

The first three coefficient estimates indicate that the weeks of work of immigrant men in 1981 were 4 to 7 percent lower than those of similar nonimmigrant men. The weeks’ assimilation is found to be 3 to 4 percent and significant for the pre-1961 cohort and the 1971-80 cohort, respectively. The immigrant men who arrived between 1981 and 1988

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

390 Christopher Worswick have similar weeks of work to the non-immigrants, ceteris paribus. The immigrants who arrived in 1989 have much lower weeks of work (29 percent) than non-immigrants. This may be picking up an increased likelihood of being unemployed in the first years after arriving. The hypothesis that the weeks IAP over the zero to ten YSM range is the same for the 1980s and 1970s cohorts is not rejected. The weighted average of the coefficients on the controls for immigrant arrival year in the 1980s implies that the 1991 hours differential over the zero to ten YSM range is four percentage points smaller than the 1981 differential.21 However, this difference is not statistically significant.22 The second test of differences in the hours IAP across arrival cohorts is performed. The coefficient on C6170 is not found to be significantly different from the sum of the coefficients on C7180 and C7180*YR91. The results of the two cohort tests indicate that differences in unobserved characteristics of immigrant men across arrival cohorts do not lead to significant differences in weeks of work between the cohorts for given levels of YSM. The second and fourth columns of the table contain results from the log weeks regression over the sample of women who work.23 The weeks of work in 1981 of immigrant women who arrived in the 1970s are 9 percent lower than those of nonimmigrant women, ceteris paribus, and this difference is significant. The immigrant wives who arrived in the 1970s experience 6 percent higher growth in weeks of work than non-immigrant wives. The weeks assimilation of the pre-1970 cohorts is not significant indicating that their weeks IAPs are flat over the YSM ranges observed in the data. Immigrant wives who arrived in Canada between 1981 and 1988 work 4 to 15 percent fewer weeks than similar non-immigrant women. The immigrant women who arrived in 1989 have much lower weeks of work (24 percent) than non-immigrants. This may be picking up an increased likelihood of being unemployed in the first years after arriving. CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

Tests are carried out on the hypothesis that the weeks IAP of recent cohorts of immigrant women are equal to the weeks IAP of earlier cohorts. The coefficient on C7180 is subtracted from the weighted average of the coefficients on the 1980s arrival cohort dummy variables. 24 The weeks IAP for the 1980s cohort over the zero to ten YSM range is not found to be significantly different from the weeks IAP of the 1970s cohort over the same YSM range. Also, the coefficient on C6170 is not found to be significantly different from the sum of the coefficient on C7180 and the coefficient on C7180*YR91 indicating that the weeks IAP of the 1970s cohort does not differ significantly from the weeks IAP of the 1960s cohort over the 11 to 20 YSM range.

THE ANNUAL EARNINGS OF IMMIGRANT AND NON-IMMIGRANT MARRIED COUPLES In order to summarize the differences in wages, hours, and weeks of work between immigrants and non-immigrants, conventional annual earnings regressions are estimated and selected coefficients are presented in Table 7. The dependent variable is the natural logarithm of annual earnings which equals the sum of the dependent variables used in the previous sections: the natural logarithms of the wage, hours, and weeks. In the first and third columns, the results are presented from the estimation of the earnings equation over the sample of husbands. Immigrant husbands who arrived prior to 1981 are found to have 16 to 34 percent lower earnings than similar non-immigrant men in the 1981 data. These differences are larger than what was found in the wage equation of Table 2 and this is due to the lower weeks of the immigrants in these cohorts compared with the non-immigrants found in Table 6. The earnings assimilation is positive and significant for the pre1961 cohort and the 1971-80 cohorts of men as was found in the wage equation estimates of Table 2 and the weeks equation estimates of Table 6. The tests of whether or not the annual earnings IAP of immigrant men has shifted down for more VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market

391

TABLE 7 Immigrant/Non-Immigrant Annual Earnings Differences by Immigrant Cohort for Husbands and Wives

Husband

Wife

Husband

Wife

CPRE61

-.1821** (.034)

-.1364* (.063)

C8185

-.4167** (.041)

-.2490** (.062)

C6170

-.1601** (.032)

-.1141 (.060)

C86

-.4631** (.073)

-.3333** (.110)

C7180

-.3447** (.035)

-.2467** (.061)

C87

-.4573** (.063)

-.3782** (.087)

C80

-.8972** (.135)

-1.228** (.146)

C88

-.7504** (.075)

-.4807** (.082)

CPRE61*YR91

.1167** (.029)

.0879 (.049)

C89

-.8014** (.071)

-.5786** (.091)

C6170*YR91

.0465 (.024)

.0516 (.042)

C90

-1.152** (.105)

-1.015** (.104)

C7180*YR91

.1037** (.028)

.1402** (.043)

R2 F N

.133 78.6 27,783

.142 48.1 19,558

NOTES : 1. Standard errors are in parentheses. 2. ** denotes significance at the 1 percent level; * denotes significance at the 5 percent level. 3. The first two and last two columns contain estimates from the husband’s log earnings equation and the wife’s log earnings equation, and have the same structure as (1). The variables listed in the table are defined in the Appendix. The demographic controls in the husband’s equation are identical to those used in the husband’s wage equation of Table 2. In the wife’s equation, the demographic controls differ from those in the wife’s wage equation of Table 2 in that the controls for the number of children present in the household are also included. A complete set of estimates is available upon request from the author. SOURCE : Author’s compilation.

recent cohorts are performed. The hypothesis that the earnings IAP over the zero to ten YSM range is the same for the 1980s and 1970s cohorts is rejected. The weighted average of the coefficients on the controls for immigrant arrival year in the 1980s implies that the 1991 earnings differential over the zero to ten YSM range is 19 percentage points smaller than the 1981 differential. 25 Also, the coefficient on C6170 is found to be 8 percentage points larger than the sum of the coefficient on C7180 and the coeffi-

cient on C7180*YR91 and this difference is statistically significant indicating that the earnings IAP of the 1960s cohort lies above the earnings IAP of the 1970s cohort over the 11 to 20 YSM range. Therefore, the results indicate that successive cohorts of immigrant men have fared worse in terms of annual earnings for given years of residence. The second and fourth columns of Table 7 contain estimates from the annual earnings equation

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

392 Christopher Worswick estimated over the sample of wives who work.26 The pre-1961 cohorts of immigrant women have 11 to 25 percent lower earnings than similar nonimmigrant women which mirrors the results of the wage equation of Table 2. The assimilation of the 1971-80 cohort of women is significant over the period at 14 percent. The tests for cohort differences discussed above are performed. The weighted average of the coefficients on the variables identifying arrival between 1981 and 1989 is found to be 10 percentage points lower than the coefficient on C7180 and this difference is statistically significant. The difference between the coefficient on C6170 and the sum of the coefficients on C7180 and C7180*YR91 is near zero and insignificant. Therefore, there is some evidence of a decline in the earnings of immigrant women from more recent cohorts relative to those from earlier cohorts, ceteris paribus; however, the magnitude of these differences is at most half of the decline found for men.

CONCLUSIONS AND IMPLICATIONS FOR IMMIGRATION POLICY In this paper, immigrant and non-immigrant married couples have been compared in terms of the wage rates, hours of work, and weeks of work of both spouses. Immigrant husbands and wives were found to have much lower wage rates than their nonimmigrant counterparts. However, immigrant wage growth over the 1980s exceeded that of nonimmigrants leading to wage rate differences between immigrants from early cohorts and similar nonimmigrants of the range 7 to 20 percent for men and 0 to 14 percent for women. Immigrants were found to work similar hours of work per week to nonimmigrants and these differences do not vary with duration of residence in Canada. The weeks of work of recent immigrants were found to be lower than the weeks of non-immigrants; however, immigrants with more years of residence have weeks of work

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

which are only 4 to 6 percent lower than nonimmigrants. The findings in this paper are consistent with certain predictions of the Family Investment Hypothesis and inconsistent with others. Consistent with the hypothesis, immigrant women are found to have higher wages relative to non-immigrant women than is true for immigrant men relative to nonimmigrant men, although the differences are small in magnitude. Initially, higher wages for immigrant women could occur if immigrant women enter jobs that are initially higher paying but offer less opportunity for career advancement. The return to education relative to non-immigrants is higher for immigrant men than women. Furthermore, the effect of young children on immigrant women’s hours of work is less than it is for non-immigrant women. These outcomes could be viewed as work-related responses by immigrant wives to a family investment strategy in which the wife (controlling for other factors) works more and foregoes investment in human capital in order to help finance her husband’s investment in host-country specific skills. Other outcomes are not consistent with the FIH. In particular, the wage growth of immigrant women relative to non-immigrant women exceeds the wage growth of immigrant men relative to non-immigrant men. Also, immigrant women do not appear to supply more hours than non-immigrant women in the first years after migration as would be predicted by the FIH. Although clear evidence of intra-family career trade-offs was not found, the results indicate that the immigrant wife plays a major role in the earnings creation of immigrant families. The lower immigrant/non-immigrant earnings differentials for wives than husbands imply that the share of family earnings contributed by the wife is larger in immigrant families than in non-immigrant families. Also, the larger decline in the immigrant/nonimmigrant earnings differentials for recent cohorts of immigrant men than immigrant women indicates

VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market that the trend is towards a still greater role for the immigrant wife in the earnings creation of the family. The findings support a family orientation to both the evaluation and the implementation of immigration policies. The relatively stronger labour market performance of recent cohorts of immigrant women than immigrant men means that evaluating the success of past immigration policies at selecting immigrants using data on men will lead to too negative a conclusion. In terms of the implementation of immigration policies, given the important role of immigrant women in earnings creation, policies should focus not only on facilitating the labour market adjustment of the principle applicant (often a man), but also assist the labour market adjustment of other family members. It may be possible to design government programs to evaluate the training needs of families rather than individuals. This would allow for the identification of the family members who are having the most difficulty finding jobs suited to their skills. Such a program could begin at the time of entry into the country when the principle applicant is required to provide both personal and family information. From the information gathered at this stage it may be possible to identify which family members are likely to have difficulty adjusting to the Canadian labour market and training support could be provided to the family. Providing assistance at this early stage could avoid delays in the family members acquiring appropriate training due to their lack of information concerning the Canadian labour market and possible problems of recognition of foreign credentials. There could be significant earnings gains for immigrant families from this kind of program given that recent immigrant women are found to have a very low return to university education relative to either native-born women or immigrant women from earlier arrival cohorts. Future research should pursue an explanation for the lower returns to education of immigrant women than non-immigrant women found in this study and in previous research. Identifying the cause of the

393

lower return on the immigrant wife’s education may lead to new policies which will facilitate the adjustment of immigrant families to Canada. A second issue which needs to be addressed is the cause of the lower weeks of work of immigrant women in the first five to ten years after migration.

NOTES I am grateful to David Green for extensive comments and suggestions. I would also like to thank Craig Riddell, Terry Wales, Charles Beach, Jeff Borland, Ted McDonald, Alice Nakamura and Jim Nason for helpful discussions. The comments of three anonymous referees were also very helpful. Research support was provided through a SSHRC doctoral fellowship and a grant from the Centre for Research in Economic and Social Policy at the University of British Columbia. 1The demographic variables are dummy variables iden-

tifying education level, place of residence and language fluency. The variables are described in detail in the Appendix. 2As is discussed below, the sample is restricted to married couples in the 1981 and 1991 Census files where both the husband and the wife were between the ages of 25 and 54 in 1981. 3The

remaining terms in the equation are: ß0 which is the intercept; ß, α, α91 , γ, γ91 and δ which are parameter vectors; and u the error term. 4To

see this, note that the expected change over the 1980s in the log wage of an immigrant husband who arrived before 1981 is:

™ W FB= Aα 91 + C1γ 91 where the vector A has appropriate values to identify the individual’s age in 1981 and the vector C1 has appropriate values to identify the immigrant’s arrival cohort. Similarly, the expected change over the 1980s in the log wage of an otherwise identical non-immigrant husband is:

™ W NB = Aα 91 Therefore, the wage assimilation for an immigrant husband is:

™ W FB− ™ W NB = C1γ 91

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

394 Christopher Worswick 5The

combination of separate age effects by birth cohort and separate years-of-residence effects by immigrant arrival cohort means that it is not possible to control for time effects (or year effects). Differences in the labour market conditions between 1981 and 1991 will be absorbed into the coefficients on the age and immigrant cohort variables. In order to identify labour market assimilation, it is assumed that possible differences in macroeconomic conditions across the survey years affect the labour market outcomes of immigrants and nonimmigrants to the same extent.

weights are .4446, .0669, .1325, .1410, .1239, and .0911, for the variables C8185, C86, C87, C88, C89 and C90, respectively. 13 The

Wald test statistic is distributed asymptotically according to the Chi-square distribution with one degree of freedom. 14 The wage IAP of the pre-1961 cohort is not compared with that of the 1961-70 cohort since the exact YSM range of the earlier cohort is not known. 15 The

6The

sample used in most studies of the labour market adjustment of immigrant men contains only men who worked full-time hours and more than 40 weeks in the reference year. Therefore, the sample restrictions used in this paper are not as selective as in previous studies and are more representative of the underlying population of men. 7The annual earnings variable from the 1991 data is deflated into 1981 dollars using the GDP deflator for Canada. 8Two

final sample restrictions are used. Households are excluded from the sample if either spouse is predominantly self-employed or is in the military. 9The

coefficient estimates are distributed asymptotically according to the Normal distribution. White’s (1980) heteroskedasticity consistent covariance matrix is used. 10 To

be precise, this is the wage differential for immigrants in the education default category of completed high school. The differential returns to education for immigrants and non-immigrants are discussed below. The wage, hours, and weeks differences for immigrants were also compared at both the immigrant and non-immigrant mean values of education as suggested by Borjas (1985). Since the results do not depend in general on the point of comparison, the immigrant/non-immigrant differentials holding education at the default level are the focus of discussion in this paper. 11Given

the broad arrival year categories, this is as close to a test of the wage IAP’s intercept being the same for the two cohorts as is possible. 12 Each weight is the ratio of the number of immigrant men in the sample who have a value of one for the cohort dummy variable, divided by the total number of immigrant men in the sample who arrived in the 1980s. The

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

wife’s equation is estimated using a two-step selection correction method (see Heckman 1979). The first stage Probit index is defined using both the wife’s and the husband’s variables. The results of the Probit estimation are available on request from the author. The coefficient on the inverse Mill’s ratio variable is positive and significant at the one percent level indicating that women with higher wage offers are more likely to be in the labour market. The equation was reestimated by OLS ignoring the selection into the labour market with only minor changes in the coefficients of the wage regression. 16 Each

weight is the ratio of the number of immigrant women in the sample who have a value of one for the cohort dummy variable, divided by the total number of immigrant women in the sample who arrived in the 1980s. The weights are .4814, .0629, .1216, .1472, .1100, and .0769, for the variables C8185, C86, C87, C88, C89 and C90, respectively. 17A

family level wage ratio equation was also estimated, where the dependent variable is the natural logarithm of the ratio of the husband’s wage to the wife’s wage. The exogenous variables included controls for both the husband’s and wife’s demographic characteristics as well as controls for both the husband’s and the wife’s arrival cohort. In principle, this specification allows for the measurement of the effects of one spouse’s arrival cohort and YSM holding the arrival cohort and YSM of the other spouse fixed. However, due to the high collinearity of the husband’s arrival year and the wife’s arrival year and the fact that broad arrival year blocks are used in the raw data, very few of the coefficients on the arrival cohort variables were significant from zero. Equivalent family equations were also estimated where the dependent variable was the natural logarithm of the ratio of the husband’s to wife’s hours, weeks or annual earnings. In each case, the standard errors on the cohort variables were large

VOL . XXII, NO . 4 1996

Immigrant Families in the Canadian Labour Market and the coefficients on the immigrant controls were insignificant in virtually all cases. Therefore, the estimates of these equations are not included in the paper. 18 The

weights are the same as those used in the analogous test of section four. 19As in the previous section, a two-step estimation procedure is used to address the selection into employment by the wives. The log hours equation was reestimated by OLS (ignoring the selection into the labour market) with only minor changes in the estimates. 20 The

tests are analogous to the ones described above for the husbands’ equation. The weights are the same as the ones used in the equivalent test for wives in section four. 21 The

coefficient on C80 is not used in the analysis to eliminate the truncation of the weeks variable for that cohort. For the same reason, the coefficient on C90 is not included in the weighted average. Therefore, the weights are .4892, .0736, .1457, .1551, and .1364 for the variables C8185, C86, C87, C88, and C89, respectively. 22 The

Wald test statistic is distributed asymptotically according to the Chi-square distribution with one degree of freedom. 23 Again,

a two-step procedure is used to address the endogeneity of the wife’s participation decision. The Probit specification is the same as the one used for the wage and hours regressions. 24 The coefficient on C90 was not included in the average to eliminate the truncation of weeks of that cohort. Therefore, the weights are .5215, .0681, .1317, .1595, and .1192 for the variables C8185, C86, C87, C88, and C89, respectively. 25 The

coefficient on C80 is not used in the analysis to eliminate the truncation of the earnings variable for that cohort. For the same reason, the coefficient on C90 is not included in the weighted average. The weights are the same as those used in the analogous test for the weeks equation. The Wald test statistic is distributed asymptotically according to the Chi-square distribution with one degree of freedom. 26 Again,

a two-step procedure is used to address the endogeneity of the wife’s participation decision. The Probit specification is the same as the one used for the wage and hours regressions.

395

REFERENCES Abbott, M.G. and C.M. Beach (1993), “Immigrant Earnings Differentials and Birth-year Effects for Men in Canada: Postwar-1972,” Canadian Journal of Economics 26:505-24. Baker, M. and D. Benjamin (1994), “The Performance of Immigrants in the Canadian Labour Market,” Journal of Labor Economics 12: 369-405. ____ (1996), “The Role of Family in Immigrants’ Labour Market Activity: An Evaluation of Alternative Explanations,” American Economic Review 86. Beach, C.M. and C. Worswick (1993), “Is There a Double-Negative Effect on the Earnings of Immigrant Women?” Canadian Public Policy 19:36-53. Bloom, D. and M. Gunderson (1991), “An Analysis of the Earnings of Canadian Immigrants,” in Immigration, Trade and the Labor Market, ed. J. Abowd and R. Freeman (Chicago: University of Chicago Press for the NBER). Borjas, G.J. (1985), “Assimilation, Change in Cohort Quality, and the Earnings of Immigrants,” Journal of Labor Economics 3:463-89. Borjas, G.J. and S. Bronars (1991), “Immigration and the Family,” Journal of Labor Economics 9:123-49. Campbell, M. (1994), “Too Many Immigrants, Many Say,” The Globe and Mail, 10 March. Chiswick, B.R. (1978), “The Effect of Americanization on the Earnings of Foreign-Born Men,” Journal of Political Economy 86: 897-921. Citizenship and Immigration Canada (1995), “Canada Changes Immigration Criteria for Skilled Workers,” Press Release, 17 November. Duleep, H. and S. Sanders (1993), “The Decision to Work by Married Immigrant Women,” Industrial and Labor Relations Review 46:677-90. Heckman, J.J. (1979), “Sample Selection Bias as a Specification Error,” Econometrica 47:153-61. LaLonde, R. and R. Topel (1992), “Immigrants in the American Labor Market: Quality, Assimilation, and Distributional Effects,” AEA Papers and Proceedings, pp. 297-302. Long, J. (1980), “The Effect of Americanization on Earnings: Some Evidence for Women,” Journal of Political Economy 88:620-29. Simpson, J. (1994), “Immigration Policy Needs to Balance Families, Refugees and Independents,” The Globe and Mail, 26 October. White, H. (1980), “A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test of Heteroskedasticity,” Econometrica 48:817-38.

CANADIAN PUBLIC POLICY – ANALYSE DE P OLITIQUES,

VOL. XXII, NO. 4 1996

396 Christopher Worswick APPENDIX DEFINITIONS OF VARIABLES LISTED IN TABLES

OTHER CONTROLS

1.

WAGE: hourly wage rate (derived by dividing the product of hours worked in the reference week and weeks worked in the reference year into annual earnings in the reference year). Deflated into 1981 dollars using the GDP deflator for Canada.

2.

HOURS: hours of work in the reference week.

3.

WEEKS: weeks of work in the reference year.

4.

PRIMARY: has zero to eight years of education.

5.

SECOND: has nine to thirteen years of educa-

The following variables are included in all equations. In the husband’s equations the variables relate to the husband and in the wife’s equation they relate to the wife: dummy variables identifying: (i) place of residence in the Atlantic provinces (New Brunswick, Newfoundland, Nova Scotia, and Prince Edward Island), Quebec outside Montreal, Montreal, Ontario outside Toronto, Toronto (default), the Prairie provinces (Manitoba, Saskatchewan, and Alberta), British Columbia outside Vancouver, and Vancouver; and (ii) language fluency in English but not French (default), French but not English, and both English and French.

tion but has not graduated from high school. 6.

HSGRAD: has graduated from high school.

7.

POSTSEC: holds a postsecondary certificate or

diploma. 8.

BACH: holds a bachelor’s degree.

9.

GRAD: holds a professional or graduate university degree.

10. FB : Immigrant. 11. CPRE61, C6170, C7180, C80, C8185, C86, C87, C88, C89, C90: Indicator variables for immigrants whose year of arrival was before 1961, between 1961-70, 1971-80, in 1980, between 1981-85, in 1986, 1987, 1988, 1989 and 1990, respectively.

CANADIAN PUBLIC POLICY – ANALYSE DE POLITIQUES,

Also, in the wife’s wage equation the number of live births is also included as a regressor. In the wife’s hours equation and weeks equation controls for the number of children present in the household are also included as regressors (see Table 5 for their definitions). In the wife’s earnings equation, both the number of live births variable and the controls for the presence of children are included as regressors.

VOL . XXII, NO . 4 1996