Article body

Introduction

This study investigates the earnings gap between Black and White workers in the Canadian economy. According to the 2006 Canadian census, the average employment income of full-year, full-time, Black workers was $40,179 in 2005 while the average for all Canadian workers was over $11,000 greater at $51,221.[1] Between 2000 and 2005, average real earnings of Black workers rose at a rate of 2.1 per cent compared to 5.5 per cent for all workers, resulting in a widening earnings gap.[2] Perhaps not surprising in light of these earnings statistics, a disproportionate share of Blacks are found in lower-skilled and lower-paid occupations and conversely are under-represented in many highly-skilled, well-paying occupations. For example, the 2006 Canadian census indicates that only 7.7 per cent of full-year, full-time Black workers were managers, in contrast to 13.3 per cent of all workers, while 7.4 per cent of Blacks were employed in lower-skilled service occupations, roughly 50 per cent more than that of all workers.

Recent research consistently finds that Black workers face one of the largest earnings gaps amongst ethnic groups in Canada. For example, based on 1996 census data restricted to Canadian-born workers, Pendakur and Pendakur (2002) found that Black females faced the largest earnings gap of 26 ethnic groups and Black males faced the second largest earnings gap.[3] Similarly, Hum and Simpson (1999), using 1991 census data, found no statistically significant wage gaps between Canadian-born visible and non-visible workers, with the exception of Black males, and concluded that in light of “the finding of a significant wage differential between blacks and other Canadians, it is time to investigate this phenomenon more carefully.” Pendakur and Pendakur (2007) applied quantile regression methods to 2001 census data and found that, unlike other Canadian-born visible minorities, Black workers face “great disparity across the distribution”, implying that access to good jobs at all skill levels is problematic. Furthermore, Hou and Coulombe (2010) examined the relative earnings of Canadian-born Blacks, Chinese and South Asians using 2006 census data and concluded that the large earnings shortfalls faced by Blacks, particularly Black men, are “striking.” These findings provide the motivation for the investigation conducted in the present study.

Canada is a very diverse country with large visible minority populations in many of its major cities. Nonetheless, instances of workplace and societal discrimination point to potential challenges in the labour market, including racial discrimination being intertwined with incidences of poverty and other socio-economic challenges (Cassin, Krawchenko and Vander Plaat, 2007).[4] Amongst visible minority groups in Canada, survey evidence suggests that Black workers may face the largest potential challenges in the labour market. In focus group analysis, it was revealed that between 10 and 38 per cent of participants had “experienced unequivocal racial discrimination at work or in trying to obtain work” and Blacks were most likely to have experienced discrimination (Canadian Race Relations Foundation, 2000). Similarly, in the 2002 Ethnic Diversity Survey conducted by Statistics Canada, the highest rate of perceived discrimination was reported by Blacks at 49.6 per cent compared to the visible minority average of 35.9 per cent (Reitz and Banerjee, 2007). Consequently, the potential impact that labour market discrimination may exert on the earnings of Black workers warrants further investigation.

The present study aims to provide an understanding of the earnings differential between Black and White workers in the Canadian setting. While there have been a number of recent studies examining visible minority earnings in Canada (e.g., Hou and Coulombe, 2010; Pendakur and Pendakur, 2007, 2011; Yap and Konrad, 2009), this literature lacks a recent investigation of the combined impact of wage discrimination and occupational segregation on the earnings gap faced by Black workers. To our knowledge, Howland and Sakellariou (1993) was the last paper to decompose the earnings gap faced by Canadian Blacks into the portion attributable to differences in productive endowments, and those which may reflect wage discrimination and occupational segregation.[5] We remedy this gap in the literature by utilizing the decomposition method developed by Brown, Moon and Zoloth (1980).

The traditional earnings decompositions separate the earnings differential into two distinct portions, namely: (i) the share due to productivity-related differences (i.e., endowments) and (ii) the share not due to endowment differences. In this study, we also conduct the traditional earnings decomposition consisting of the Mincer (1958) approach as well as the Blinder (1973) and the Oaxaca (1973) approaches. These traditional earnings decompositions have been used by several researchers investigating the earnings gap in the Canadian setting (Pendakur and Pendakur, 2002, 2007; Hou and Coulombe, 2010). However, we extend this analysis by also using the Brown, Moon and Zoloth method. This method extends the traditional decomposition by also identifying the role specifically played by occupational differences, namely, differences in the return to productive endowments within an occupation (referred to as wage discrimination) and differential access to occupations despite possessing similar productive endowments (referred to as occupational segregation). By incorporating an occupational attainment model into the more traditional earnings decomposition, the Brown, Moon and Zoloth technique allows for an explicit quantitative estimate of the role of occupational segregation in earnings differences. This has the attractive feature of neither implying that occupational differences are by themselves “legitimate” determinants of the earnings gap nor ignoring occupational attainment in the formal analysis.[6]

The literature suggests that the Brown, Moon and Zoloth technique is particularly well-suited for an examination of Black-White earnings differentials for numerous reasons. First, prior analyses of Canadian data found substantial occupational differences unrelated to individual characteristics (Howland and Sakellariou, 1993; Darden, 2005). Second, previous applications of this technique to Black-White earnings differentials concluded that occupational segregation is indeed a major contributing factor to the earnings gap (Howland and Sakellariou, 1993). For example, using 1986 Canadian census data on male workers, Howland and Sakellariou (1993) attributed about 19 per cent of the Black-White earnings gap to occupational segregation while Gabriel and Schmitz (1989) utilized 1980 U.S. Census of Population data and concluded that the occupational segregation of Black men accounts for nearly half of the observed Black-White earnings differential. Third, analysis of 2006 Canadian census data is highly suggestive that occupational segregation is an important contributing factor to the earnings gap. For instance, Hou and Coulombe (2010) concluded that for Black males working in the public sector, the earnings disadvantage stems more from working part-time and in lower-paying occupations than from lower pay in the same job, while for Black males working in the private sector, job sorting was concluded to play a similar role as lower pay in the same job. Similarly, Torczyner (2010) found that Black workers in Montreal were significantly under-represented in high paying occupations and over-represented in low paying occupations, even when possessing relatively similar educational qualifications as their non-Black co-workers.

The Brown, Moon and Zoloth method attempts to provide further insights into the portion of the earnings gap unexplained by differences in productivity characteristics. Consistent with any empirical analysis, the findings from this method of earnings decompositions are subject to interpretation. For instance, Dougherty (2005) suggests that “discrimination, tastes and circumstances” may explain occupational differences and, in turn, the portion of earnings gap unexplained by differences in productivity characteristics yet attributed to wage discrimination and occupational segregation. On the other hand, Fouad and Byers-Winston (2005) suggest that there are few meaningful differences in occupational preferences across groups. Similarly, evolutionary economic theory suggests that individuals have an incentive to select strategies (occupations) that maximize their expected return from the utilization of their productive endowments. These varying interpretations in the literature suggest that the Brown, Moon and Zoloth method can be conservatively interpreted as providing upper-bound estimates of wage discrimination and occupational segregation (Gabriel and Schmitz, 1989).

The key findings of the study are as follows: The total earning gap identified through the study after standardizing for productive endowments was about $12,400 in 2005 using the 2006 Canadian census. We find that roughly one-fifth of the Black-White earnings gap (equaling $2,600) can be attributed to productivity-related endowment differences. The remaining four-fifths of the earnings gap (equaling $9,800) can be attributed to occupational segregation and wage discrimination. The upper-bound estimate of wage discrimination is calculated at slightly over half the overall earnings differential and occupational segregation is estimated to account for up to the remaining one-fifth of the gap. Black workers were found to be under-represented in high-income occupations while over-represented in low-income occupations. For instance, given the productive endowments possessed, 13.3 per cent of Black workers were predicted to be in the managerial occupations while the actual number was found to be 7.3 per cent. It was predicted that 5.3 per cent of Black workers would be employed in elemental service occupations yet the actual number was 7.7 per cent. In aggregate, these estimates translate into annual earnings losses due to wage discrimination and occupational segregation of approximately $1.5 billion for full-time full-year Black workers in the Canadian workforce.

The policy implications of the findings relate to the need for public policy and private sector efforts to address the vast majority of the earnings gap, which is primarily due to wage discrimination and occupational segregation, while the remainder of the earnings gap can be shrunk through individual and community efforts as well as complementary public and private initiatives. By implication, the contemporarily observed socio-economic outcomes (e.g., incidence of poverty and educational attainment) can be reasonably anticipated to be directly related to the disincentives and financial deficits caused by the earnings gap. Additionally, we conclude that the Brown, Moon and Zoloth decomposition technique is an important methodological device for investigating the earnings gap of Black workers in Canada in light of the role of occupational attainment differences in earnings. Furthermore, this technique may hold promise in investigating the earnings gaps pertaining to other groups in the Canadian workforce.

The paper is organized as follows: next section presents the data we utilized; it is followed by a section discussing the econometric methods and a section presenting empirical results. The last section concludes the paper with a discussion of public policy implications and suggestions for further research into this area.

Data

The data set used in the study is derived from the 2006 Canadian Census Public Use Microdata File (PUMF) on individuals. The PUMF contains 844,476 records, representing 2.7 per cent of the Canadian population. In this study, weighted data are generally reported and analyzed. We restrict the sample to individuals between the ages of 18 and 64, who worked mostly full-time (30 hours or more per week) on a full-year basis (i.e., between 49 and 52 weeks) with positive employment earnings in 2005. Self-employed workers are excluded from the analysis by removing those with non-zero self-employment earnings in 2005.[7] These restrictions result in a sample of 209,346.[8]

For much of the analysis, we also restrict the sample to workers who are Black or White. The “Visible Minority Indicator” variable is utilized to categorize an individual’s race.[9] As these data are based on a question posed only to non-Aboriginal Canadians, we exclude Aboriginal individuals from the “Not Visible Minority” category to derive the subset of workers whom we define as White. This latter restriction reduces the sample size to 180,003 of which 4,219 are Black and 175,784 are White. The sample size represents a population proportional to 6,669,135 individuals which is comprised of 156,080 Blacks and 6,503,054 Whites across Canada.

Table 1 contrasts weighted mean values for the Black and White workers in our data sample. In terms of labour market earnings in 2005, White males earned nearly $18,000 more per year or about $340 more per week than Black males, while White females earned about $5,200 more per year or $100 more per week than Black females. The reported difference in the total weeks worked is very minor (0.1 week for either gender) as can be anticipated, since the sample is restricted to those who worked on a full-year basis in 2005. The most marked difference between the White and Black worker samples is the much higher percentage of Black workers who are immigrants (76.4 per cent for males and 75.9 per cent for females) compared to White workers (10.7 per cent of males and 10.1 per cent of females). Similarly, a lower percentage of Black workers in 2005 have English or French as a mother tongue (67.1 per cent for males and 72.3 per cent for females) compared to White workers (about 85 per cent for either gender). Other reported human capital characteristics are fairly similar amongst the Black and White workers in the sample. For instance, years of schooling are nearly the same at 13.7 years for White males and 13.8 for Black males (14.1 for White females and 14.2 for Black females), while years of potential work experience was slightly higher for White workers (by about one percentage point).

Table 1

Weighted Means by Ethnicity and Gender

Weighted Means by Ethnicity and Gender

1. French mother tongue in Quebec, English elsewhere.

2. As information on years of schooling was not collected in the 2006 census, we used imputed years of schooling based on estimated years of schooling by the highest level of certificate or degree for individuals aged 25 to 64 from the 2001 census as presented in Hou and Coulombe (2010) under their footnote 7, page 41.

3. Potential years of work experience calculated as age – years of schooling – 5.

-> See the list of tables

The geographic distribution of White and Black workers is quite different. Black workers are much more likely to reside in Ontario (60.3 per cent of males and 64.9 per cent of females) whereas White workers are more evenly distributed across Canada. Black workers are also more likely to reside in urban areas as 94.8 per cent of Black males live in Census Metropolitan Areas (CMAs) compared to 37.4 per cent of White males (numbers for females are 64.9 per cent and 37.7 per cent, respectively). Since the cost of living is generally higher in large urban centres, these geographical differences imply that Black workers require higher wages than their White counterparts in order to achieve a comparable purchasing power. The earnings gap therefore suggests that an even greater purchasing power gap faces Black workers.

The distribution of workers within fairly broad occupational categories varies between the two groups of workers. White workers have greater representation in managerial, professional, supervisory, and trade occupations whereas Black workers have a relatively higher representation within the clerical, sales, semi-skilled and low level service occupations.[10] As seen in Table 2, the occupations in which Whites are over-represented are the four occupations with the highest annual earnings (managerial, professional, trade, and supervisory). In contrast, those occupational categories where Black workers are over-represented are the four lowest paying occupations (elemental services, clerical, semi-skilled and sales).

Table 2

Annual Earnings by Ethnicity, Gender and Occupation ($)

Annual Earnings by Ethnicity, Gender and Occupation ($)

-> See the list of tables

The average earnings for the sample group consisting of Black and White workers between 18 and 64 is found to be $52,880. This average income, however, only partially reflects the dispersion of worker income on the basis of occupation. For instance, managers are at the top-end of the annual earnings distribution with an average income of $81,909. In comparison, individuals in the elemental services occupation are earning an annual income well below the average at $27, 294.

Table 3 provides further insight into the earnings gap between Black and White workers by examining the gap along occupational and gender lines. It shows the striking differences in earnings differentials by gender; the male Black-White earnings gap is more than three times that of females. This table shows that while Black females nearly achieve earnings parity with White females across a number of occupations (e.g., professionals, supervisory, clerical and sales), Black males significantly lag White males irrespective of occupation. Table 3 also indicates that the earnings gap varies significantly across occupations, both in absolute and relative terms.[11] At the low end of the spectrum, Black managers earned 69.4 per cent of White managers’ earnings, whereas Blacks who worked in elemental services, earned 92.7 per cent of their White counterparts. These earnings differentials are further expanded when gender and ethnicity are jointly considered (i.e., earnings of Black females as compared with White males in Table 3).

Table 3

Annual Earnings Differences by Ethnicity, Gender and Occupation

Annual Earnings Differences by Ethnicity, Gender and Occupation

-> See the list of tables

Given that 76.2 per cent of the black workers in the data sample are immigrants, compared to a corresponding figure of 10.4 per cent for White workers, it might be thought that the observed earnings differentials are largely spurious. Specifically, recent immigrants have been shown to fare poorly in the Canadian labour market due to a combination of factors, including lack of Canadian work experience, non-recognition of foreign credentials, lower levels of English or French language proficiency and lack of social and job networks (Nadeau and Seckin, 2010; Statistics Canada, 2003). However, the general patterns in the raw earnings gap persist, even when the analysis is restricted to Canadian-born workers.[12]

Econometric Model

The Black-White earnings gap is investigated as follows: first, we apply the standard econometric and decomposition approaches and, second, we utilize the Brown, Moon and Zoloth technique. For the standard econometric approach, Equation 1 represents the traditional Mincer (1958) earnings function:

where the index i refers to the individual, ln ei denotes the natural logarithm of earnings, xi are the observed productivity-related characteristics that determine earnings (including a dichotomous variable capturing ethnicity), β is a vector of coefficients (including the constant term) and ui is the error term.

The second econometric approach to derive the magnitude of the earnings gap is the decomposition technique advanced by Blinder (1973) and Oaxaca (1973). In this model, Equation 1 is estimated separately for both groups under examination (i.e., Black and White workers). While the dummy variable approach essentially equates discrimination with the difference between the intercepts of the two earnings regression lines, the decomposition approach equates discrimination with the part of the gap resulting from differences in coefficient values. Specifically, it can be shown that the difference in average earnings can be decomposed as follows (Gunderson, 1979)[13]:

where: the w and b indexes refer to White and Black workers respectively. In this equation, the first term on the right hand side represents that part of the earnings differential that is unexplained by attribute differences and the second portion represents the part of the earnings gap explained by differences in the mean values of characteristics. The first term on the right hand side of Equation 2 is often interpreted as an upper-bound measure of labour market discrimination (Dougherty, 2005; Nadeau and Seckin, 2010).

In addition to estimating Equation 2, we utilize the approach originally advanced by Brown, Moon and Zoloth (1980), who point out that when occupational control variables are included in Equation 2, differences in occupational representation appear in the part of the earnings gap that is characterized as “legitimate.” They contend, however, that if occupational attainment is even partially a result of discrimination, then this is an inappropriate claim. Therefore, an alternative formulation of the earnings gap is required and represented as follows:

where lne̅w and lne̅b are the natural logarithm of mean earnings of Whites and Blacks, and pwj and pbj are the proportions of White and Black workers in the jth occupation. The first term on the right hand side of Equation 3 is the portion of the earnings gap due to inter-occupational differences[14] and the second term of this gap represents the earnings gap within occupations (i.e., the intra-occupational differences).[15]

Howland and Sakellaniou (1993) show that Equation 3 can be further decomposed into the following components:

where pb’j is a measure of the predicted share of Blacks in the jth occupation according to White’s predicted occupational distribution. Equation 4 allows the earnings gap to be decomposed into four components. These components are: (1) Black-White differences in individual attributes; (2) occupational segregation due to differences in individual attributes; (3) differences in returns to characteristics; and (4) differences in occupational segregation not due to differences in characteristics.[16] On the one hand, since the first two components of this wage gap are due to differences in productivity-related characteristics, they can be described or characterized as “justifiable.” On the other hand, these latter two components are not justifiable according to endowment differences and represent, to some extent, intra-occupational differences in returns (i.e., wage discrimination) and differential access to occupations (i.e., occupational segregation), respectively.

To obtain the pb’j term appearing in the second and fourth elements of Equation 4, we utilize a multinomial logit model of occupational attainment as in Liu, Zhang and Chong (2004). Specifically, we model occupational choice, for nine discrete occupational categories, as a function of: immigration status, official language spoken, gender, marital status, years of schooling, years of work experience, years of work experience squared, province of residence, and urban status. These factors represent a fairly standard list of variables utilized in similar studies (e.g., Gabriel and Schmitz, 1989; Howland and Sakellariou, 1993; Liu, Zhang and Chong, 2004). Unfortunately, some factors likely important in modeling occupational attainment, such as the occupational attainment of parents and career aspirations, are unavailable in the census data utilized in this study (Harper and Haq, 2001)

The equations from the multinomial can be solved to compute the predicted probabilities.[17] Following the notation appearing in Liu, Zhang and Chong (2004), the predicted probabilities of occupational attainment are calculated as follows:

where N is the sample size, J is the number of occupational categories, Pmj is the probability of individual m working in occupation j, Zm is a vector of determinants of occupational choice, and ϒj is a vector of these coefficients corresponding to the kth occupation.

Results

The results emerge from the application of econometric techniques to the standard approach represented by Equation 1, the Oaxaca-Blinder decomposition approach reflected by Equation 2, and the Brown, Moon and Zoloth approach in Equation 4, which includes a multinomial logit model as specified by Equation 5.

Standard Approach

The initial results are derived from the estimation of Equation 1. Table 4 shows the results of three regressions where the natural logarithm of weekly earnings is regressed separately by gender. The explanatory variables included in this model are fairly conventional characteristics such as immigration status, language proficiency, marital status, years of education, highest level of schooling, work experience (years), work experience squared, province of residence, urban status, occupation, industry sector and gender (in the case of the pooled male and female regression). As can be seen, Black workers faced significant negative earnings differentials of approximately 7 per cent for females, 20 per cent for males and 14 per cent for the combined sample. When these data samples are restricted to Canadian-born workers, the magnitude on the negative pay differentials faced by Black workers fall somewhat, but are still quite pronounced at 7 per cent for females, 18 per cent for males and 12 per cent for the combined sample. Restricting the analysis to immigrants results in larger estimated earnings differentials faced by Black workers: 10 per cent for females, 23 per cent for males and 17 per cent for the combined sample.[18]

Table 4

Earnings Regressions by Gendre, All Workers

Earnings Regressions by Gendre, All Workers

-> See the list of tables

Table 4 (continued)

Notes: Reference categories in square brackets, Absolute value of t statistics in parentheses, Significance is denoted by *at 10%; ** at 5%; and *** at 1%.

-> See the list of tables

These results are consistent with those of other studies. Pendakur and Pendakur (2002), using 1996 census data, found that Canadian-born black females faced an earnings disadvantage of 22 per cent and black males faced an earnings disadvantage of 36 per cent. These estimated negative earnings differentials exceed our findings for two main reasons. First, they did not include occupational categories as explanatory variables. Second, while our reference category for Blacks was all White workers, their reference group was more narrowly defined as British-origin Whites who have relatively high earnings compared to other ethnic groups (e.g., Greek, Spanish, etc.).

Hum and Simpson (1999), using 1993 data from the Survey of Labour and Income Dynamics (SLID), estimated a version of Equation 1 with a correction for sample-selection bias. Specifically, they utilized an inverse Mills ratio term constructed from a probit regression to enable them to analyze differences in wage offers rather than observed wages. Hum and Simpson’s (1999) estimated earnings disadvantage for Canadian-born Black males of 24 per cent and Canadian-born Back females of 9 per cent are close to our findings (18 per cent for males and 7 per cent for females) despite differences in the data sets, time periods covered, explanatory variables, and treatment for sample selection.

Oaxaca-Blinder Decomposition Approach

Table 5 presents the Oaxaca-Blinder decompositions (i.e., Equation 2) which are based on earnings regressions very similar to those reported in Table 4.[19]

Table 5

Oaxaca-Blinder Decomposition of Earnings Differentials

Oaxaca-Blinder Decomposition of Earnings Differentials

Note: For ease of interpretation, when differences are expressed in dollar terms, the percentages figures are based on applying the percentage of gap to the actual mean dollar figures which appear in Table 1. An alternative method is to derive the dollar figures as differences in the antilogs of the mean logarithms. Differences between the two methods are trivial in terms of the figures generated.

-> See the list of tables

According to the Oaxaca-Blinder decomposition method, the $240.66 weekly earnings differential between Black and White workers is comprised of $99.15 (41.2 per cent) due to differences in endowments of productive characteristics and $141.15 (58.8 per cent) that is not attributable to endowment differences. The latter is attributed to wage discrimination. Nonetheless, it is important to note that correlated and confounding factors (e.g., self-selection and preferences) may have emerged from historical events and influenced occupational choices even though current circumstances may or may not still precipitate these occupational selections. For females, the percentage attributable to wage discrimination is smaller than for males (57.5 per cent compared to 64.9 per cent) and combined with the fact that the raw gap is much larger for Black males than for females ($344.37 compared to $97.95), the corresponding discrimination-related earnings losses experienced by males is approximately four times that of females ($223.50 compared to $56.32).

Brown, Moon, and Zoloth Decomposition Approach

The multinomial logit estimation models the determinants of access into nine occupational categories and is based on the White workers in the data sample.[20] After these coefficients are estimated, the predicted share of Black workers in a given occupation is calculated at the mean value of each characteristic parameter (i.e.,pb’j in Equation 4). These predicted results along with the actual distribution for Black and White workers are presented in Table 6.

Table 6

Actual and Predicted Occupational Distribution

Actual and Predicted Occupational Distribution

-> See the list of tables

As can be seen, based on their productive attributes, Black workers are predicted to be more highly represented in managerial positions (actual probability of 7.3 per cent compared to predicted probability of 13.3 per cent), professional occupations (actual probability of 17.1 per cent compared to predicted probability of 17.6 per cent), semi-professional occupations (actual probability of 8.0 percent compared to predicted probability of 8.2 per cent) and supervisory positions (from 3.0 percent actual to a 3.8 percent predicted probability). Consistent with these findings, Black workers are also over-represented in the lowest level occupations. For example, 7.7 per cent of Black workers are concentrated in elemental services while based on their attributes, only 5.3 percent of Black workers are predicted to hold these occupations. Although these results suggest that the observed earnings gap between Black and White workers is largely due to occupational differences, a decomposition of the earnings differential as specified in Equation 4 is necessary to confirm these impressions.

Table 7

Decomposition of White-Black Earnings Differential

Decomposition of White-Black Earnings Differential

Notes: The numbers in parenthesis refer to the constituent parts on the right hand side of Equation 4. As in Table 5, when differentials are expressed in dollar terms, percentages derived from the econometric estimations are applied to the actual mean earnings.

-> See the list of tables

Table 7 below presents the decomposition of the earnings differential based on Equation 4. On the one hand, it shows that endowment differences explain a relatively small 21.1 per cent of the earnings gap. On the other hand, differential returns within occupations explain about 56.4 per cent of the earnings gap and differential access to occupations explains another 22.5 per cent of the gap.[21] These two components represent upper-bound estimates of wage discrimination and occupational segregation.

The results in Table 7, therefore, conclude that nearly 80 per cent of the total weekly earnings differential of $240.66 is not explained by differences in endowments of productivity-related attributes. These findings clearly highlight the value of explicitly modeling occupational attainment. In Table 5, where occupation is exogenously determined, the component of the earnings differential attributed to endowment differences was about twice as large (i.e., 41.2 per cent compared to 21.1). Since Equation 2 and Table 5 do not account for occupational segregation, the earnings differential attributable to endowment differences can be anticipated to be overstated. Correspondingly, the scale of discrimination estimated is likely understated. Nevertheless, even as estimated at $141.51 on a weekly basis, lost earnings of this magnitude impose a substantial economic burden on individuals.[22]

Across Canada, the total cost of occupational segregation and wage discrimination affecting full-year, full-time, Black workers is estimated at $1.5 billion.[23] Since immigration status is controlled for in estimations of occupational attainment and earnings, these estimated earnings losses cannot be dismissed as arising from assumptions of the relatively poor labour market performance of new Canadians. Indeed, when the analysis as presented in Table 5 is repeated with only Canadian-born workers, the portion of the earnings gap remaining unexplained by productivity differences falls from the estimated 58.8 per cent, yet remains relatively high at 44.3 per cent.[24]

The estimated aggregate costs of occupational and wage discrimination of $1.5 billion may understate the true economic impact of occupational segregation and wage discrimination in two respects. First, the conventional multiplier effect associated with the absence of earnings equivalent to $1.5 billion. Specifically, the loss of income would result in lower spending and employment activities attributable to Black workers, including contributing to lower economic activities within these communities. Second, there may be intergenerational aspects reflective of current income losses that constrain the human capital investments and bequests of current and future generations as well as community prosperity. These impacts are likely to be particularly acute for recent immigrants who depend on the social capital and prosperity of their family and community for support in their integration into their new society. Hence, historical wage and occupational discrimination can be anticipated to exhibit socio-economic persistence and intergenerational transference will be observed in current and future earnings, education, and occupational outcomes, if not redressed through countervailing policy efforts.

Conclusion

The study utilized 2006 census data to estimate the earnings differential between Black and White workers in the Canadian labour market. Specifically, when the earnings gap was equated with the difference in intercepts between regressions, the earnings differential between Black and White workers in the Canadian labour market was estimated at 6.9 per cent for females, 20.3 per cent for males and 13.7 per cent for the combined sample. When the sample was restricted to Canadian-born individuals, the earnings differential fell to 11.9 per cent for the combined male and female sample and, when it was restricted to immigrants, the differential rose to 17.2 per cent. These findings suggest that social capital (e.g., social networks, knowledge of institutions, language proficiency and Canadian workplace culture and communication skills or “soft-skills”) may play a role in determining wage outcomes so favouring Canadian over foreign born workers and, in turn, influencing the wage gap in the Canadian labour market. Nonetheless, the productive value of social capital to individual employer cannot be directly ascertained from the data. All else being equal, the larger earnings gap faced by immigrants as opposed to Canadian-born individuals suggests the combination of ethnicity and immigration status warrants further analysis.

When the conventional Oaxaca-Blinder decomposition technique was utilized, the upper-bound estimate of the discrimination component was calculated at 58.8 per cent, while this increased to 78.9 per cent when the Brown, Moon and Zoloth technique was used. This illustrates the sensitivity of estimates of labour market discrimination to the econometric technique employed and the efficacy of broadening the notion of discrimination to include occupational segregation. Since previous research has concluded that, of any visible minority, Black men are “most profoundly affected by labour market discrimination” (Swidinsky and Swidinsky, 2002), the Brown, Moon and Zoloth method appears to provide useful complementary evidence to that derived from the Oaxaca-Blinder decomposition method (e.g., Hou and Coulombe, 2010).

Results from the Brown, Moon and Zoloth method applied in this study reveal that full-year, full-time Black workers lose up to an average of $9,800 annually due to wage discrimination and occupational segregation. Applying this upper-bound estimate on an aggregate basis implies that the total costs of occupational segregation and wage discrimination are $1.5 billion annually in the Canadian economy. Pendakur and Pendakur (2011) suggest that the earnings gap faced by visible minorities in Canada has not diminished over the past two decades. Hence, the economic losses to Black workers and others from occupational segregation and wage discrimination represent significant social losses across multiple generations of workers.

The policy implications of this study pertain to the actions on the part of governments and corporations in general, as well as individuals and organizations in the Black community to address the earnings gap issues identified. Specifically, the results suggest that government and organizations can investigate workplaces and implement ameliorating policies to address: (i) wage discrimination, by ensuring that workers are earning equal pay (or eliminating pay inequity) within the same occupations; (ii) occupational segregation, by ensuring that workers’ occupational attainment is consistent with their educational attainment, work experience, and other productivity-enhancing attributes; and (iii) productive endowment gaps, by promoting educational attainment and emphasizing strategic occupational selection.

Canada has a wide range of policies aimed at achieving equity in the labour market, including the Canadian Human Rights Act and Equal Wages Guidelines, 1986, at the federal level and the Pay Equity Act at the provincial level (i.e., in Manitoba, New Brunswick, Nova Scotia, Ontario, Prince Edward Island, and Quebec).[25] Nonetheless, the amount of wage discrimination and occupational segregation identified in this study suggest that Canadian labour markets are failing to compensate all workers on the basis of their productive endowments. To the extent that lower rates of return to education and training are responsible for the earnings gaps between Black and White workers, distortionary incentives exist for the acquisition of human capital by Black workers, thereby causing the economy to perform below its potential output. It can therefore be anticipated that the Canadian economy would experience socio-economic gains through effective action addressing wage discrimination and occupational segregation in the labour market.

Labour market failure can be addressed through mechanisms that result in private and public sector employers taking corrective action to internalizing the distortion. Affirmative action and employment equity initiatives have often been utilized to attain these results. Organizations may also utilize training and professional development initiatives to support efforts to overcome labour market failure, including ensuring that there are fair and equitable opportunities and access to these initiative for all employees. Organizations have also established principles of workplace diversity and equity within their mission statements and strategic plans. However, the record of accomplishments of organizations in relation to these principles are generally not found in their public documents, such as annual reports. To support accountability and transparency, publicly-traded companies and government agencies could be required to report their performance in addressing wage discrimination and occupational segregation within their organizations. Furthermore, the consequences of labour market failure can be anticipated to be reflected in the distribution of income, wealth and social outcome translating into a growing sense of marginalization and social disconnectedness between those individuals facing wage discrimination and occupational segregation and the rest of society. The earning gaps identified in this study can also be anticipated to lead to under-investment in education and, correspondingly, costly programs and actions aimed at addressing socio-economic outcomes that are, in fact, rooted in issues of wage discrimination and occupation segregation.

The present study points to scope for further research. First, the estimate of economic magnitude of earnings differentials faced by Black and other visible minority workers in the Canadian labour market could be extended to investigate the effectiveness of different programs and policies in private and public sectors aimed at addressing wage discrimination and occupational segregation. Second, the trend over a number of decades in the earnings gap due to wage discrimination, occupational segregation and productive endowment differences would be helpful in determining how Canada is progressing as a society in promoting inclusion and equity within the labour market. Third, the analysis conducted in this paper could be extended to the investigation of earning gaps within other ethno-cultural groups.