Publications

    Culture, Tourism and the Centre for Education Statistics

    Location of Study and the Labour Market Success of Immigrants to Canada

    Chapter 6: Empirical strategies

    Warning View the most recent version.

    Archived Content

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

    Archived

    This page has been archived on the Web.

    Although very useful for analysis, descriptive statistics examined in the previous chapter are not adjusted for differences in key determinants of labour market success, such as labour market experience, field of study, occupation, language skills, province of residence, and so on. By ignoring labour market experience or language skills, for instance, we might mistakenly suggest that labour force status and wage earnings of landed immigrants are more sensitive to location of study than they really are.

    A multivariate analysis allows us to measure the sensitivity of earnings, employment and education-job match rates to location of study while simultaneously taking into account several determinants of labour market success in Canada. This chapter presents the empirical specifications used for labour force status (section 6.1) and employment earnings (section 6.2).

    6.1 Labour force status

    In order to estimate the relative effect that location of postsecondary study has on labour market success of internationally-educated immigrants, we start by assuming that each individual faces seven mutually exclusive labour force statuses, namely: unemployed, not in the labour force, self-employed, undereducated employee, correctly-matched employee, overeducated employee and school attendee. The individual chooses the alternative that maximizes the present value of his/her benefits given observed attributes and labour market dynamics.

    From analytical models (see, McFadden, 1973), we know that the multiple-choice nature of this classification calls for the use of a multinomial Logit analysis. This model, which is an extension of the binary Logit model, is appropriate because it allows having more than two values for labour market behaviour and the exogenous variables to depend on the alternatives.1 Using this technique, we can estimate the effect of a given location of study on the likelihood that a randomly-chosen immigrant (average immigrant) will select any given labour force status compared to his/her randomly chosen Canadian-born (average Canadian-born) counterpart, after having held other characteristics constant.

    The net effect that each location of study has on the labour force status of immigrants relative to Canadian-born is estimated for each of very-recent, recent and established immigrants. In doing so, we are able to quantify the change in the average likelihood that a very-recent, recent or established immigrant who graduated in a given country will be self-employed, for example, compared to their Canadian-born counterparts. A detailed description of the statistical methodology is available upon request to the author.

    6.2 Employment earnings

    The analysis of earnings differences by location of study is limited to paid employees. The self-employed are excluded from this analysis due to issues associated with the accuracy of self-reported labour income.2 We perform two alternative empirical strategies for the earnings function. First, we use ordinary least-squares (OLS) to estimate the relationship between the natural logarithm of annual employment earnings of immigrants (as compared to the Canadian-born) and location of postsecondary study, taking into account the fact that employment earnings are also determined by several other measured attributes.

    However, given that labour earnings are observed only for individuals who are employed in the wage sector, OLS estimates of the earnings equation may be biased because the unemployed and people out of the labour force are not collectively similar to individuals having a paid job.3 Given that labour market behaviour might not be a random process (as pointed out in chapter 3), the effect of location of study on earnings may be underestimated as quantified using a standard OLS coefficient (Heckman, 1979). This in turn may compromise the validity of our key analytical conclusions. Our second empirical strategy consists of relying on a two-stage method proposed by Trost and Lee (1984) to correct for potential sample selection when estimating the wage earnings equation. This regression technique is described in an Appendix available upon request to the author.

    Box 1

    Background explanatory variables

    In order to estimate the sensitivity of labour market success of immigrants to their location of highest postsecondary certificate, degree or diploma completion, we specify that the dependent variables i.e., labour force status and annual employment earnings, are functions of a set of pertinent variables drawn from prior research.

    Human capital: Labour market experience, health status and field of study all have been identified as major determinants of labour market success in a typical market economy (for international literature review, see Psacharopoulos and Patrinos 2002; for recent Canadian evidence, see for example, Boudarbat and Chernoff, 2009; Hansen 2006; Ferrer, Green and Riddell 2006; Sweetman and McBride, 2004). We account for health status via a dichotomous variable that takes the value 1 if activities of daily living are limited by physical problems and 0 if not. In line with previous research,4 age and age squared are used to measure potential labour market experience. Field of study is taken into account with mutually exclusive dichotomous variables. Despite its relevancy for our analysis, education has been omitted from our regression models because it serves to build employment statuses such as undereducated, correctly-matched and overeducated (see Chapter 3). In doing so, we avoid producing spurious correlations between some labour market alternatives and levels of education.

    Language skills: A critical number of empirical studies suggests that in most countries favoured by long-term international migrants, i.e., the United States, Australia and Canada,5 labour market success is a function of proficiency in a country's official language(s) (see, for example, Chiswick and Miller (2002) for the United States; Pendakur and Pendakur (2002a) and Thomas (2009) for Canada; Liebig (2007) for Australia). Mutually-exclusive dichotomous variables are used to account for the knowledge of Canada's official languages, namely: English (reference category), French, English and French, neither English nor French.

    Family circumstances: Our analysis includes a series of dummy variables that contain information on marital status, position in the economic family (i.e., whether an individual is the primary household maintainer or not) and household demographic structure (i.e., the family generation to which each individual of interest belongs and whether this person lives with dependent children or not). These variables have been shown to be associated with the decision to work, labour supply intensity (see for instance, Gunderson, 1998) and may measure the need for job security.

    Employment conditions: There are some indications that individual earnings vary across professional occupations in Canada (see, for instance, Boudarbat and Chernoff, 2009; Hansen, 2006). Variations in employment earnings could be due to differences in work schedule or occupation: some workers may be observed with higher earnings because they hold two or more jobs, work longer hours or have secured employment in well-paying occupations. To take into account the effect of hours worked for pay, dummy variables are used to distinguish between mainly part-time (reference category) and mainly full-time work. Finally, the analysis controls for occupation via mutually-exclusive dichotomous variables.

    Market forces: Business cycles, and institutional and environmental factors contribute to the variability of labour market conditions. For instance, a structural shift towards more skill-intensive employment opportunities in a given industry or sector will likely reduce the demand for less-skilled, less-educated, low-paid workers while increasing that for highly-educated, highly-skilled, well-paid workers. Similarly, the hiring of new employees is less likely to be a common practice in industries experiencing economic contraction as opposed to those experiencing economic growth. Further, there could be geographical variation in the type of labour supply, industrial composition or production technologies. To control for economic and institutional forces, we follow prior research (Boudarbat and Chernoff, 2009; Hansen, 2006; to name few) in using province or territory of residence and area of residence as explanatory variables. These variables also allow us to take into account other omitted variables that predict employment and earnings in a highly competitive economy such as the Canadian economy.6 

    Nonmarket forces: Previous studies indicate that regardless of gender, visible-minority immigrants, especially those of black descent, tend to earn significantly less than immigrant whites in Canada's labour markets, even after controlling for measured productive attributes (for example, see Baker and Benjamin, 1994; Pendakur and Pendakur, 1998; 2002b). It has been suggested that the significant earnings disadvantage comparatively experienced by immigrant members of visible minority groups is associated at least in part with economic discrimination in the marketplace, i.e., with the fact that visible minorities are treated differently (less favourably) than members of the majority racial group with identical measured productive attributes.7 Also, it has been claimed that in a typical host economy with multiple source countries, limited information about the productive attributes of many prospective labour market participants may give many prospective employers an incentive to rely on observable characteristics such as race to proxy their productivity and reward them accordingly (see, for instance, Arai and Vilhelmsson, 2001, Junankar, Paul and Yasmeen, 2004). According to Frenette and Morissette (2003) and Aydemir and Skuterud (2004), observed economic underutilization of immigrants' professional skills could result from an amalgam of factors including visible-minority status. Our analysis therefore includes a dummy variable which takes the value 1 for visible minorities and 0 otherwise.8

    Gender: There is some indication of differences in females' and males' occupational distributions in Canada. For instance, the 2006 Employment Equity Data Report reveals that women disproportionately worked in white-collar, clerical, sales and services occupations, whereas men were disproportionately employed in senior, middle and other management, and in occupations unique to manufacturing, transportation or primary sector (Human Resources and Skills Development Canada, 2009). There is also theoretical support for gender differences in labour market participation in prior research. As an example, Nekby (2002) claims that labour market participation may be sensitive to gender because women have childbirth considerations and are more likely than men to allocate a greater share of their disposable time in the production of home-made public goods. To account for gender differences in labour force participation patterns, each of our regression models contains a dichotomous variable equal to 1 for males and 0 for females.

    End of box

    Notes

    1. Ibid.
    2. As underlined in many household surveys, the annual taxable earnings of the self-employed even derived from tax registries might underestimate their true annual earnings. Further, the self-employed are associated with issues that are beyond the scope of our study. These include self-selection in professional occupations: educational credentials of the self-employed are traditionally assessed by consumers whereas prospective employers value those for other wage sector workers. Due to their employment status thus, the self-employed have a lower propensity than other paid workers to automatically rely on educational signal when participating in the labour market, especially in occupations where educational credentials are not crucial (Heywood and Wei, 2004).
    3. This problem is well-known in economics and related areas as sample selection bias (see, for instance, Heckman, 1979).
    4. Ibid.
    5. Two thirds of all people who migrated to another country between 1975 and 1980, for example, went to Australia, Canada or the United States (Friedberg and Hunt, 1995). As of 2005 moreover, those three countries had the highest proportions of foreign-born people: 24.6%, 19.2% and 11.7% respectively (Hawthorne, 2006).
    6. As measured by the Global Competitiveness Index (GCI) of the World Economic Forum, Canada has one of the world's most competitive economies. As of 2006-2007 for instance, Canada observed a higher GCI (often taken as an overall competitiveness measure), i.e., 5.37 in comparison with the other classical immigration country with strong selection policy, namely, Australia (5.29) and with other country members of the Organization of Economic Cooperation and Development such as France (5.31), New Zealand (5.16), South Korea Republic (5.13), Belgium (5.29), Spain (4.77) and Italy (4.46).
    7. Ibid.
    8. The 2006 Census dictionary defines visible minorities as "persons other than Aboriginal peoples, who are non-Caucasian in race or non-white in colour".
    Date modified: