Corps de l’article

Since the 1980s, awareness has increased concerning employment discrimination against gays and lesbians. In fact, all three actors of the Industrial Relations (IR) system (Labour, Government and Private Agencies, and Management) have addressed issues concerning sexual orientation. For example, in Canada, most governments have passed human rights legislation that specifically prohibits discrimination based on sexual orientation (Wintemute 1996). In addition, courts in both Alberta and in the federal jurisdictions have “read in” prohibitions against discrimination based on sexual orientation when the legislation in question did not specifically reference it (Zinn and Brethour 1999). Moreover, several unionized and non-unionized employers have recently added same-sex benefits to their traditional benefits plans (Human Rights Campaign 2002), and some unions have included sexual orientation in the non-discrimination clauses of their collective agreements (Brown 1998; Hunt 1997).

Much of the information concerning gay and lesbian workplace issues consists of web site data consolidated by gay and lesbian activists (e.g., The Center for Lesbian and Gay Civil Rights 2000; Human Rights Campaign 2002; Stratton 2000). To date, there have only been a limited number of studies which have examined the role of Canadian unions in addressing the workplace issues of gays and lesbians.

Perhaps the earliest work in this area was that of Genge (1983). Genge’s work is seminal in that it is the first to openly address the role of Canadian labour in the workplace issues of lesbians and gays. In this work, Genge presents her own experience and overviews the resolutions passed by unions, in particular by the Canadian Union of Public Employees (CUPE), the Ontario Federation of Labour (OFL) as well as by the Canadian Labour Congress (CLC) to include sexual orientation in discrimination clauses of collective agreements, constitutions and laws. Similarly, White (1993), in her book Sisters in Solidarity, provides a brief overview of sexual orientation issues focusing on the legal issues of the early 1990s and labour’s role in promoting these issues. Petersen (1999) builds on this foundation by overviewing recent human rights tribunal rulings and court case results concerning same-sex partner benefits and the role of union activists in these rulings and cases.

Recently, two more in-depth studies have also been conducted. The first, Hunt (1997), represents an in-depth three-phase study. Phase one consisted of telephone surveys of 9 Canadian labour federations and 29 Canadian unions. Phase two consisted of three in-depth case studies of the unions deemed by Hunt to be the most progressive, namely, CUPE, the National Automobile, Aerospace and Agricultural Implement Workers of Canada (CAW), and the Public Service Alliance of Canada (PSAC). Phase three was a descriptive statistical analysis (e.g., frequencies, percentages) of the non-discrimination provisions of over 3400 Ontario collective agreements covering more than 200 employees. Based on this work, Hunt concluded that public sector unions were more active in addressing the issues of gays and lesbians than were private sector unions, and that issues concerning sexual orientation often surfaced in female dominated unions.

Hunt’s (1997) argument that public sector unions were more progressive in this area should not be surprising for at least two reasons. First, the employer of these public sector groups is the government. Given the visibility of government employment decisions and policies, there may be increased pressure for these governments to act as ‘model’ employers relative to their counterparts in the private sector (Gunderson 1995; Ponak and Thompson 2001). Second, there are significant demographic differences between private and public sector unions that may play a role here. For example, Ponak and Thompson (2001: 474) argue that public sector unions, with their approximately 58% female membership relative to the one-third female membership of private sector unions “have been among the strongest advocates of workplace equity” in terms of employment equity, pay equity and anti-discrimination regulations.

Brown’s (1998) study built on the work of Hunt by adding a United States comparison. His study consisted of interviews (phone and e-mail) and secondary source data (web sites, collective agreements, union publications) concerning two federations of labour: the American Federation of Labor-Congress of Industrial Organizations (AFL-CIO) and the CLC, as well as four unions, two from Canada (CUPE, CAW) and two similar unions from the United States (American Federation of State, County and Municipal Employees [AFSCME], the United Autoworkers of America [UAW]). Brown suggested that public sector unions were more active in the area of sexual orientation, that a union’s first efforts in the area of sexual orientation focused on freedom from discrimination before moving on to same-sex benefits, and that larger bargaining units were more likely to include sexual orientation clauses in their collective agreements. A potential explanation as to why a union’s initial efforts focus on freedom from discrimination may be the fact that unions view non-discrimination language in collective agreements as a method of educating members concerning discrimination issues (Giles and Starkman 1995). Hence, unions may choose to focus on an education mandate before moving to the benefits arena in an effort to make their members more aware of, and comfortable with, workplace issues concerning sexual orientation. A potential explanation for why larger bargaining units are more active in this area may simply be due to the availability of resources. Larger units may have more resources to dedicate to equality issues (e.g., special committees, educational/awareness programs, etc.) relative to smaller units where union resources may be dedicated to the more traditional areas of negotiations, grievances and arbitrations. Brown’s (1998) research also highlighted the importance of both women’s and employment equity committees. As these committees often examined issues concerning fair treatment, non-discrimination, harassment, and education, much of the initial forays into sexual orientation in the unions in Brown’s study were initiated and championed by women’s and employment equity committees.

While the studies of both Hunt (1997) and Brown (1998) break new ground in this area, these studies leave at least three significant gaps. First, they fail to use a theoretical framework or model to ground their research, develop hypotheses/propositions, or frame their findings. Second, they do not provide a systematic, statistical testing of important hypotheses. Rather, these studies used qualitative case study analysis and non-inferential statistical analysis that precluded the testing of key relationships or hypotheses. Third, while it is argued that IR is a multi-disciplinary field (Hébert, Jain, and Meltz 1988), these past works failed to include relevant research from other disciplines, in particular sociology and psychology, to help frame their research. As such, the over-arching purpose of the present study was to systematically develop and test several of the relationships proposed by Brown and Hunt using quantitative analyses and a more substantive theoretical base.

As argued by Kervin (1992: 54), there has been “a long-simmering” debate concerning the relative merits of qualitative versus quantitative research. My decision to use a quantitative approach in the present study was largely guided by the view that social science “research requires both qualitative and quantitative approaches: the former in an exploratory phase to generate understanding and ideas, the latter to test these ideas” (Kervin 1992: 54). More specifically, my decision to now turn to quantitative research was based on the following. First, the previous work in this area has been largely qualitative. As is the case with qualitative research (Kervin 1992; Saks 2000), these studies have provided excellent, in-depth discussions of the key areas of interest (in this case issues concerning sexual orientation) as these issues play out in a small sample (in this case unions). As such, these past studies have given us the ability to “understand a specific, particular event by getting ‘inside it’” (Kervin 1992: 54). However, limitations of qualitative approaches include that they do not allow for statistical testing of key relationships as “methods of analysis are not well formulated” (Saks 2000: 154) and such approaches cannot assess the extent to which the qualitative findings from one setting hold true in other settings (Kervin 1992; Saks 2000). Second, these limitations of qualitative research can be overcome by well-proven quantitative methods designed to empirically test key relationships. Thus quantitative methods allow us to test the extent to which the findings of the qualitative research hold true in other settings and samples.

In terms of the theoretical framework selected for this study, the systems approach was chosen as it is the basis of the pluralist approach to IR and as this approach is “the starting point for scientific analysis” (Meltz 1993: 164). In particular, Craig’s model of industrial relations was used as it has frequently been used as the basis for industrial relations teaching and interpretation in Canada (Gottlieb Taras, Ponak, and Gunderson 2001; Meltz 1993). While a detailed presentation of Craig’s model is available elsewhere (Craig 1988, 1990; Craig and Solomon 1996), a summary of Craig’s model follows. Craig views the three actors of the IR system being impacted by internal inputs (e.g., power, values, and goals of the actors) that flow from the external inputs of the external environment. These external inputs (or environmental subsystems) include the legal subsystem, economic subsystem, political subsystem, and social subsystem. Through a series of conversion mechanisms (e.g., day-to-day relationships, negotiation processes, third party interventions, strikes and lockouts), the actors produce organizational and worker-oriented outputs (e.g., management rights, union recognition/security, terms and conditions of work, wage and effort bargain, etc.). These outputs, through the feedback loop, impact the external environment and the IR system (in particular the power, values and goals of the actors) such that the system is dynamic and ever changing.

As previously stated, the outputs of the IR system are influenced by the inputs from the environment. In terms of the current external inputs, we see a legal subsystem that is increasing its recognition of gay and lesbian workplace rights (Petersen 1999; Wintemute 1996), a political subsystem that has taken legislative action to entrench these rights (Peterson 1999; Wintemute 1996), and a social subsystem where public opinion concerning gays and lesbians is changing. For example, in Canada we have recently seen several provinces (e.g., British Columbia, Quebec and Ontario) permit same-sex unions as a recognized form of marriage (Makin 2003) or as a recognized form of civil union (“N84, Loi instituant l’union civile” 2003). In fact, in the first four months after Ontario permitted same-sex marriages, over 780 same-sex licenses were issued by the city of Toronto alone (Harding 2003)!

In terms of the social subsystem, and in particular the public opinion element, research has shown that public attitudes towards gays and lesbians are not uniform and that a number of interesting differences exist. First, past research from the United States and Canada has highlighted the role of gender differences in terms of reactions to gays and lesbians. For example, Herek (2002: 41–42), when summarizing a series of research papers (Herek and Capitanio 1995, 1996, 1998) resulting from national phone surveys of over 1,000 Americans, stated that research has shown that: “heterosexual men consistently displayed more negative affective reactions to homosexuality than did heterosexual women.” Moreover, in Herek’s (2002) phone survey of over 650 Americans, women were significantly more supportive of passing laws that protected gays and lesbians from discrimination in the workplace, as well as laws that recognized gay male marriages and same-sex benefits for gay males. Herek also found that while men and women were supportive of gays and lesbians having equal rights in terms of job opportunities, they were less supportive of passing a law that specified that gays and lesbians should have equal employment rights. From a Canadian perspective, a study of approximately 200 university students also showed that male students had less positive attitudes towards gay men than did female students (Schellenberg, Hirt, and Sears 1999).

Second, while Hofstede (1998) noted that there are many cultural similarities between Canada and the United States, Canadians, when compared to Americans, are generally more likely to support measures designed to reduce discrimination based on sexual orientation and have been found to be more supportive of the need for equal rights for gays and lesbians in the military and teaching positions relative to Americans (Mazzuca 2002). Moreover, over 60% of Canadians responding to an Angus Reid poll were concerned with discrimination against gays and lesbians, suggesting that “Canadians view gay and lesbian discrimination in the workplace . . . as a fundamental issue of human rights and equality” (“Stop bias against gays” 1995: A19). Additional evidence suggests that 56% of 1500 Canadians surveyed in a Southam News-Angus Reid public opinion poll supported amending the Canadian Human Rights Act to provide protection from discrimination on the basis of sexual orientation (Bryden 1996) and 75% of 1500 people surveyed by Angus Reid supported legislation barring employment and housing discrimination against gays and lesbians (Ovenden 1998; Wright 1998). A limitation of these polls is that they strictly reported descriptive statistics and largely reported the percentages of each province that supported sexual orientation issues. None of these reports statistically examined whether attitudes differed by gender. However, a recent Canadian study examined the relationship between gender, education—both in terms of discipline (e.g., Arts, Science, Social Sciences, etc.) and level (e.g., number or years of education)—and people’s attitudes towards homosexuality. Specifically, Schellenberg et al. (1999) found that Arts and Social Sciences students (regardless of gender) had more positive attitudes toward gay men as compared to Science and Business majors and that the attitudes of male students towards gay men improved with education levels.

As previously discussed, Craig’s model suggests that external environmental factors (such as the public opinion results previously discussed) influence the goals and values of actors as well as the outputs of the industrial relations system. As such, we would expect that based on this attitudinal research, unions with memberships that are: (1) largely female, (2) highly educated, and (3) comprised of people from the social sciences and arts disciplines would have more favourable attitudes towards issues associated with sexual orientation. An industry known for being female dominated, highly educated, and largely comprised of people who studied in the social sciences and arts disciplines is the public sector (Ponak and Thompson 2001). As such, it is not surprising that Hunt (1997) and Brown (1998) found that this sector was more active than the private sector in issues concerning sexual orientation.

Based on the past research of Brown (1998) and Hunt (1997), and the preceding literature concerning public attitudes, I set out to test the following hypotheses:

  • Hypothesis 1: Relative to collective agreements in the private sector, public sector collective agreements are more likely to have anti-discrimination clauses that include sexual orientation and same-sex benefits provisions.

  • Hypothesis 2: Larger bargaining units are more likely to have anti-discrimination clauses which include sexual orientation and same-sex benefits relative to smaller bargaining units.

  • Hypothesis 3: As compared to collective agreements not containing equality clauses, collective agreements containing equality clauses (e.g., employment equity, sexual harassment, equal pay and disability) are more likely to have anti-discrimination clauses that include sexual orientation and same-sex benefits.

  • Hypothesis 4: Collective agreements that have anti-discrimination language that provides protection based on sexual orientation are more likely to have same-sex benefits language than those that do not.



The sample consisted of 248 collective agreements from the Human Resources Development Canada’s (HRDC) collective agreement library. In September of 1999, when this sample was drawn, there were approximately 1076 collective agreements in the HRDC’s collective agreement library. These 1076 agreements only included provincial agreements with over 200 employees and federal agreements of over 500 employees. The total number of employees in the sample of 248 collective agreements was 493,812. The agreements in the sample had expiration dates ranging from 1993 to 2004 with the mode expiration date being 1998. The library codes all collective agreements in a database. The initial sample was drawn by an electronic search conducted by the HRDC library staff. They selected all agreements in the sample that HRDC had coded as having some language concerning prohibited grounds of discrimination. This decision was made because I wished to examine the specific language concerning discrimination, as my primary intent was to examine the extent to which sexual orientation was included as a prohibited grounds of discrimination in these agreements.

The decision to focus on collective agreement language was based on the view that these agreements remain “at the center of the Canadian industrial relations system” (Giles and Starkman 1995: 368) as they provide a framework that guides the day-to-day relationships of the parties and, therefore, the rights of employees. In terms of my decision to look at discrimination provisions in collective agreements, this decisions was guided by the view that unions seek to include these clauses as they have the advantages of educating members and allowing these issues to be enforceable via the grievance procedure (Giles and Starkman 1995).


The procedure included three key steps. First, once the sample had been drawn, a research assistant, who had been previously employed in the HRDC library, took the list of agreements identified by the HRDC search and manually searched each collective agreement. She photocopied all language in the agreement that related to discrimination and prohibited forms of discrimination.

Second, a second research assistant and the researcher reviewed the language provided by the first research assistant. They independently assessed the extent to which the collective agreement language prohibited discrimination on the basis of sexual orientation. Specifically, they independently coded the discrimination variable as follows: 0 = no provision concerning sexual orientation, 1 = clause refers to human rights code which includes sexual orientation protection, and 2 = sexual orientation explicitly stated. Collective agreements that used the terms sexual preference, gay and/or lesbian, were also coded in this final category.

Third, as the secondary purpose of this study was to examine the extent to which the collective agreements in the sample contained same-sex benefits language, the HRDC staff also conducted an electronic search for same-sex benefits provisions. Only five agreements were found in the sample that contained such benefits. The exact contract language was then provided to the researcher by the HRDC staff.


The key independent variables in this study included public sector, bargaining unit size, and presence of equality issues clauses in collective agreements.

Public Sector. This was a dichotomous variable coded as private sector = 0 and public sector = 1. Consistent with the definition of Ponak and Thompson (2001), public sector was deemed to include three levels of government (local, provincial, federal), health and welfare, as well as education. In the present sample, 152 (61.3%) of the collective agreements were in the public sector.

Size. The size of the bargaining unit ranged from 200 to 40,000. The average bargaining unit was 1992 (sd = 3767). The variable was collapsed into equal thirds of small (below 600 employees), medium (600–1100 employees) and large (over 1101 employees).

Equality Issues. As previously stated, sexual orientation issues are often championed by unions concerned with equality issues (e.g., employment equity and women’s committees). The collective agreement library database captured whether or not collective agreements contained language concerning disabilities (an area covered by human rights and employment equity legislation), employment equity, and two factors that could be seen as women’s issues (namely, sexual harassment and equal pay). In the data set, each of these variables was coded dichotomously where 1 = provision present and 0 = provision not present. Overall, 112 (45.2%) had language concerning disabilities, 45 (18.1%) had language concerning employment equity, 98 (39.5%) had language concerning sexual harassment, and 36 (14.5%) had language concerning equal pay. A multicolinearity check was conducted using Pearson correlations between these variables. None of the correlations exceeded .30; hence, there was no evidence of multicolinearity.

The key dependent variables in this study were anti-discrimination clauses that included sexual orientation and same-sex benefits.

Anti-discrimination Clauses. As previously discussed, a research assistant and the researcher independently read and coded the discrimination clauses. They coded the clauses as follows: 0 = no provision concerning sexual orientation, 1 = clause refers to human rights code which includes sexual orientation protection, and 2 = sexual orientation explicitly stated. There was 100% agreement between the research assistant and the researcher regarding the coding of the discrimination language. Sample anti-discrimination clauses are located in Table 1. The frequencies of these levels were 119 (48.0%), 34 (13.7%), and 95 (38.3%), respectively.

As previously stated, American research (Herek 2002) found that while men and women were supportive of gays and lesbians having equal rights in terms of job opportunities, they were less supportive of passing a law that specified that gays and lesbians should have equal employment rights. As such, the researcher decided to examine two different dependent variables. The first (coverage) was coded as 1 if sexual orientation was included by virtue of making reference to the law or specific contract language and was coded as 0 if there was no protection based on sexual orientation. This variable was similar to Herek’s (2002) ‘general version’ variable that assessed whether respondents felt that gays and lesbians should have equal rights in the workplace. The resulting frequencies were 129 (52.0%) and 119 (48.0%), respectively.

Table 1

Sample Clauses

Sample Clauses

-> Voir la liste des tableaux

The second dichotomous dependent variable (specific coverage) was coded as 1 if sexual orientation was specifically identified in the non-discrimination clause of the collective agreement; otherwise, it was coded as 0. This variable was similar to Herek’s (2002) ‘law version’ variable that assessed whether respondents believed that a law should be passed that specified that gays and lesbians should have employment protection. As previously identified, 95 (38.3%) of the collective agreements in the sample specifically covered sexual orientation.

Same-sex Benefits. In this sample, HRDC’s database reflected whether agreements provided gay and lesbian partners with same-sex benefits. In the present case, all same-sex benefits clauses present did so through an extension of the spousal definition. For example, “the definition of spouse and dependent amended to include same sex partner and dependents of same sex partners” (Memorandum of Agreement, Case: 066805E, italics added for emphasis). As such, this variable was a dichotomous variable where 0 = no provision for same-sex partners and 1 = same-sex partners included in spousal definition. In the present sample, only 5 (2.0%) of the agreements provided some level of same-sex benefits through such clauses. Sample same-sex benefits to which these same-sex partners were entitled included bereavement leave and healthcare benefits. Of these five agreements, 80% were Ontario agreements, 80% specifically reference sexual orientation in their discrimination language, and all had settlements dates between 1997–1998.


Crosstabs. Consistent with Herek (2002), an initial investigation and examination of the hypothesized relationships was carried out using crosstabs. This initial investigation permitted the researcher to see whether there was any evidence of the hypothesized relationships before a full logistic regression was carried out. Table 2 includes a summary of the results. Note that given the very small number of collective agreements with same-sex benefits, crosstabs were not conducted. This is consistent with the recommendations of Weinberg and Abramowitz (2002).

Table 2

Crosstabs Results (Pearson Chi-Square)

Crosstabs Results (Pearson Chi-Square)


*** = significant at p < .01; ** = significant at p < .05; * = significant at p < .10

-> Voir la liste des tableaux

Hypothesis 1 stated that public sector unions would be more likely than private sector unions to have anti-discrimination language and same-sex benefits in their collective agreements. Crosstabs revealed a relationship between public sector unions and sexual orientation coverage (χ2 = 3.40, p < .07) and specific coverage (χ2 = 4.87, p < .05).

Hypothesis 2 stated that larger bargaining units would be more likely to have anti-discrimination language which included sexual orientation as well as same-sex benefits as compared to smaller bargaining units. Crosstabs did not find a relationship between larger bargaining units and sexual orientation coverage (χ2 = 2.21, p > .05). However, there was a relationship between larger bargaining units and specific coverage (χ2 = 9.11, p < .01).

Hypothesis 3 predicted that collective agreements containing clauses concerning equality issues would be more likely to have sexual orientation language and same-sex benefits as compared to agreements that did not contain these clauses. Crosstabs revealed that coverage of sexual orientation was negatively related to disability language (χ2 = 4.45, p < .05) and positively related to employment equity (χ2 = 17.25, p < .01) and sexual harassment (χ2 = 11.47, p < .01) clauses. Specific inclusion of sexual orientation language was positively related to agreements that included employment equity (χ2 = 21.76, p < .01) and sexual harassment clauses (χ2 = 14.93, p < .01). However, no relationships were found between equal pay and sexual orientation coverage or specific coverage (χ2 = .39, p > .05; χ2 = 01, p > .05).

Hypothesis 4 stated that collective agreements with anti-discrimination language that provides protection based on sexual orientation are more likely to have same-sex benefits language than those that do not. Given the small number of collective agreement with same-sex benefits, this could not be tested.

Binary Logistic Regression. The results of the preceding crosstabs suggest that collective agreements were more likely to have anti-discrimination language concerning sexual orientation if these agreements were in the public sector, were large bargaining units, and had specific clauses concerning equality issues (e.g., employment equity and sexual harassment language). As such, binary regression was conducted to examine the extent to which each variable increased the likelihood of anti-discrimination language pertaining to sexual orientation. Again, consistent with Herek (2002), two dependent variables were used, one for sexual orientation coverage and a second for specific coverage of sexual orientation. The results of these regressions are presented in Tables 3 and 4, respectively.

Table 3

Binary Logistic Regression Results for Sexual Orientation Coverage

Binary Logistic Regression Results for Sexual Orientation Coverage

Model Chi-square = 32.84***; –2 Log Likelihood = 310.55


*** = significant at p < .01; ** = significant at p < .05; * = significant at p < .10

-> Voir la liste des tableaux

Table 4

Binary Logistic Regression Results for Sexual Orientation Specifically Mentioned

Binary Logistic Regression Results for Sexual Orientation Specifically Mentioned

Model Chi-square = 40.55***; –2 Log Likelihood = 289.56


*** = significant at p < .01; ** = significant at p < .05; * = significant at p < .10

-> Voir la liste des tableaux

Consistent with the previous crosstabs analyses concerning disability, employment equity and sexual harassment, the results revealed that collective agreements with employment equity clauses and sexual harassment clauses were 3.91 (p < .01) and 1.99 (p < .05) times more likely to provide sexual orientation coverage while those that included disability were slightly less likely to provide such coverage (Exp(B) = .59; p < .10). Thus, there was partial support for hypothesis 3.

Two potential explanations for these results are as follows. First, a potential explanation for the negative relationship between coverage and disability is industry. Many of the disability clauses (92.4%) were contained in private sector companies, which, as demonstrated in the preceding crosstabs analyses, were less likely than public sector companies to have sexual orientation language in their collective agreement. Moreover, 66.7% of collective agreements containing this language were found in the manufacturing, mining and construction industries where the language focused on the option to transfer to a new job if the employee was injured. This finding may simply reflect the fact that workplace accidents are more common in manufacturing, mining and construction operations as opposed to the office environments typically found in the public sector. For example, in 1998, the loss time injury rate was 5.82 per 100 employees in manufacturing versus 2.68 in government services (HRDC 2000). As such, these private sector, manufacturing, mining and construction unions may have negotiated specific language in their agreements given the higher rates of workplace accidents that would require accommodation (e.g., transfer to another role, adjusted hours of work, etc.).

Second, potential explanations for the positive relationships between “coverage” and employment equity and sexual harassment are public sector and gender. That is, the previously discussed crosstabs revealed that public sector agreements were more likely to include language concerning employment equity and sexual harassment relative to private sector agreements. Moreover, the public sector has historically employed more females relative to the private sector (Ponak and Thompson 2001) and females hold more favourable attitudes towards gay and lesbian issues relative to males (Herek 2002). In addition, the works of Brown (1998) and Hunt (1997) indicated a relationship between unions with a ‘strong’ female presence (such as women’s committees) and the provisions of sexual orientation language.

When the dependent variable was “specific” coverage, the variables of public sector, employment equity, and sexual harassment were significant. Consistent with hypothesis 1, the results demonstrated that public sector bargaining units were 2.08 (p < .05) times more likely to include specific coverage. Moreover, those agreements that had employment equity and sexual harassment clauses were 3.56 (p < .01) and 2.16 (p < .01) times, respectively, more likely to provide such specific coverage. However, size and disability were not significant variables in this analysis.

Again, given the small number of agreements providing same-sex benefits, binary logistic regressions were not conducted. This decision was made as there was limited variation in the dependent variable; thus, the assumptions of the technique would have been violated (SPSS 2001).


The significance of this study is five-fold. First, this study fills a void in the current literature. As previously stated, past studies in this area have failed to include a substantive theoretical grounding, have not incorporated literature from other disciplines and have not empirically tested key relationships. As such, this is the first study, to the author’s knowledge, that empirically examines the activities of unions in the area of gay and lesbian rights. Specifically, the study examined whether collective agreements included sexual orientation in anti-discrimination language and/or provided same-sex benefits. Moreover, the present study integrates research from other disciplines (e.g., psychology, sociology) and uses a substantive theoretical basis to do so (e.g., Craig’s IR system).

Second, given the increased attention concerning gay and lesbian workplace issues as seen on sitcoms such as Will & Grace and the current debate in Canada concerning same-sex marriages (Abbate and Lunman 2002), we can expect that the workplace issues of gays and lesbians will become more important in the near term. As such, this research provides an initial foray into an area that will become increasingly important as organizations become more diverse and inclusive.

Third, the present study builds on the past case studies of Hunt (1997) and Brown (1998). In particular, the analyses supported many of the propositions set forth by these authors. Specifically, larger bargaining units that are based in the public sector, and have negotiated equality provisions, are more likely to have sexual orientation included, either by reference to law or by specific mention, in their anti-discrimination language than are private sector units without equality provisions. Moreover, in support of Brown’s (1998) proposition that many unions first provide language prohibiting discrimination based on sexual orientation before moving to same-sex benefits, the data revealed that 52% of collective agreements provided such protection as compared to 2% which provided same-sex benefits.

Fourth, the results of these analyses highlight the importance of a union negotiating equality provisions. The results, particularly the binary regression analyses, demonstrated the importance of these provisions in predicting the likelihood of anti-discrimination language in collective agreements. This finding is important for a number of reasons. First, both Hunt (1997) and Brown (1998) discussed the importance of women’s caucuses and/or equity committees in putting forward and championing issues concerning gays and lesbians (e.g., non-discrimination clauses, same-sex benefits, awareness training, etc). The results of the present analyses support this argument. Specifically, the results demonstrated that the presence of employment equity and sexual harassment clauses were significant predictors of anti-discrimination language, both in terms of coverage and specific coverage. Second, for labour leaders seeking to make inroads in the sexual orientation arena, the present analyses suggest that bargaining units with collective agreement clauses pertaining to equality issues (such as employment equity and sexual harassment) would be a natural choice. Alternatively, labour leaders may seek to take several rounds of bargaining to implement sexual orientation clauses. In earlier rounds of bargaining, union leaders could ensure that supporting equality issues (e.g., employment equity, sexual harassment) are covered in the collective agreement, and then in subsequent rounds start to focus on sexual orientation directly.

Fifth, this paper supports Kervin’s (1992) argument that both qualitative and quantitative research is needed in social sciences. More specifically, the paper shows how the quantitative analyses conducted in the present study were able to test many of the key relationships found in past qualitative works (e.g., Hunt 1997; Brown 1998).

However, the hypotheses related to same-sex benefits could not be tested given the very small number of agreements that provided such benefits. With a larger database, and more variation in this variable, these relationships can be tested in the future. The recent trend to recognize same-sex marriage indicates that Canadians’ views towards sexual orientation are becoming more positive. Perhaps as Canadians’ views on the issues of sexual orientation continue to become more positive, same-sex clauses will become more common in collective agreements. As such, it would seem appropriate to re-examine these relationships in a few years as the HRDC data set should have incremental variation in this key variable at that time.

There were two unanticipated findings in this study. First, the presence of disability clauses was negatively related to sexual orientation coverage. Second, bargaining unit size, which was significant in the crosstabs analyses, was not significant in either of the binary logistic regressions. As previously discussed, a potential explanation of the disability finding is that many of these disability clauses pertained to workplace injuries and transfer options that are more common in private sector, manufacturing, mining and construction collective agreements. Given the significant effect of the public sector on sexual orientation clauses, this may be a confound in the analysis. Similarly, the null finding concerning bargaining unit size may be a factor of the data set as all bargaining units were relatively large (e.g., the smallest was 200 and the mean was 1992). Moreover, there may be a confound as most public sector unions are large. Hence, in the crosstabs analysis, the variable concerning large bargaining units may have been picking up some of the variation that was actually caused by the public sector.

There are five limitations to this work. First is the generalizability of these findings, especially those related to same-sex benefits. The small number of agreements in this sample with such benefits presents a limitation in this study. As such, the hypotheses related to same-sex benefits need to be reexamined using a different data set. In addition, as previously stated, the primary purpose of this study was to examine non-discrimination clauses in collective agreements to assess the extent to which they provided protection based on sexual orientation. As such, the study used a sample limited to collective agreements containing some language concerning freedom from discrimination, whether based on union involvement or human rights issues. Now that this study presents evidence concerning the relationships between several key variables (e.g., public sector, bargaining unit size, equality provisions) and sexual orientation collective agreement provisions, research should next examine whether these relationships hold true, or increase in significance, using a sample of collective agreements that includes agreements with no freedom from discrimination language.

A second potential limitation of this study is that the presence of anti-discrimination and/or same-sex benefits provisions may have been under-represented versus the ‘true’ availability of these provisions. For example, an employer and union may agree to include sexual orientation in non-discrimination language and/or provide same-sex benefits, even if it is not specifically mentioned in the collective agreement. Alternatively, the parties may agree to provide the provisions for sexual orientation, but place them in human resource policies not contained in collective agreements.[1]

A third potential limitation of this study is that the HRDC electronic system was used to identify collective agreements with discrimination language or same-sex benefits language. As such, any errors in the database could impact these results. A manual research of all agreements in the database would have confirmed the accuracy of the coding, particularly as it related to same-sex benefits. Unfortunately, the geographic distance between the researcher and the library precluded this manual search. However, none of the agreements that the HRDC library coded as having discrimination language were found to have been incorrectly coded when the research assistant went to photocopy discrimination language (e.g., all agreements coded as having discrimination language were found to have such language). This would suggest that the system was accurate and that agreements were correctly coded.

A fourth potential limitation of this study is that multicolinearity existed. Specifically, the literature review suggests that unions with a high percentage of females in the social sciences disciplines would be most supportive of sexual orientation issues. Unions that meet this description are largely located in the public sector. Unfortunately, I could not statistically check for this multicolinearity issue given that the HRDC library did not code agreements based on the percentage of bargaining unit members that were male/female or employed in a social science discipline. Hence, future quantitative research in this area may wish to use a survey that measures these two variables versus an analysis of collective agreement provisions.

A fifth limitation of this study is that I used an analysis of collective agreement provisions to examine the responses of labour to issues concerning sexual orientation. This decision was consistent with past research concerning labour’s responses to issues regarding sexual orientation (Brown 1998; Hunt 1997). It was also consistent with the view that collective agreements are “a benchmark for union ability” to gain employee rights in the workplace; thus, they represent the tangible results of union membership (Giles and Starkman 1995: 368). However, these agreements do reflect negotiations between employers and unions. Thus, we cannot ignore the fact that employer willingness to agree to these provisions as well as internal inputs (e.g., power, values, and goals of the actors) would influence whether or not sexual orientation provisions were provided in the final collective agreement. This limitation also points out one of the weaknesses of quantitative research in that it does not permit a detailed examination of the context and of the events being examined (Kervin 1992). As such, a future qualitative study could be conducted that examines the role of both labour and management (as well as their relative internal inputs) in an effort to understand how sexual orientation clauses become included (or are not included) in collective agreements.