Introduction
In this research note, we present the results of audit studies that were conducted with constituency offices of provincial and federal elected representatives across Canada.Footnote 1 Audit studies are particularly useful for investigating situations of discrimination in the context of constituency service (Butler and Crabtree, Reference Butler and Crabtree2021). In this research design, fictive constituents contact their representative by email for help or to inquire about a governmental service (for example, housing or unemployment). Randomization assigns representatives to a constituent request from a minority/underprivileged group (experimental treatment) or not (control group). Given that the email solicitations are identical except for the short personal resume identifying the (non-)minority group of the constituent, researchers could compare response rates across the two groups and determine whether inequality in service responsiveness exists. This research design also permits identifying factors that may increase or undermine inequality in representatives’ responsiveness.
American scholars have extensively used audit studies to investigate possible discrimination in service responsiveness to constituents. In particular, they showed that inequality exists when constituents contact their elected representatives. For example, White constituents generally receive more responses than Black constituents when they contact their legislators while Black legislators tend to favour Black constituents (Broockman, Reference Broockman2013; Butler and Broockman, Reference Butler and Broockman2011)—these results have been confirmed in a meta-analysis (Costa, Reference Costa2017). Butler (Reference Butler2014) found similar results with respect to low-income constituents. Latino citizens also received fewer responses from legislators than non-Latino constituents (Mendez and Grose, Reference Mendez and Grose2018) and when street-level bureaucrats provide information to constituents about voting ID laws across American States (White et al., Reference White, Nathan and Faller2015). Similar results on ethnic discrimination were also corroborated when contacting American citizens for participating in a survey (Block Jr. et al., Reference Ray, Crabtree, Holbein and Quin Monson2021) and local politicians in Denmark (Dinesen et al., Reference Dinesen, Dahl and Schiøler2021). Two meta-analyses also reported similar evidence of ethnic discrimination when hiring in the job market (Quillian et al., Reference Quillian, Pager, Hexel and Midtbøen2017) and renting houses (Auspurg et al., Reference Auspurg, Schneck and Hinz2019). Surprisingly, studies that examined whether women are discriminated against when contacting their representatives found opposite results; generally, women receive more responses to their requests than men (Wiener Reference Wiener2021; Thomsen and Sanders, Reference Thomsen and Sanders2020; Rhinehart Reference Rhinehart2020; Magni and de Leon, Reference Magni and de Leon2021).
Elected representatives are one of the most direct links for Canadians when they need to interact with the government for service, administrative and policy issues. This connection between constituents and representatives is key to many Canadians and politicians, as emphasized by the work of the Special Committee on Electoral Reform (ERRE report, 2016). Importantly, constituency offices are expected to address citizens’ demands and requests equally, and that citizens have equal access to their representatives. This is especially the case with respect to constituency service, which is a nonpartisan area. To our knowledge, only two audit studies of politicians have been conducted in Canada, one examining gender discrimination (Dhima, Reference Dhima2022) and another on exclusive/shared jurisdiction areas (Loewen and MacKenzie, Reference Loewen and MacKenzie2019).
Our objective in this research note is to extend this scholarship in considering other individual characteristics that may be subject to discrimination in the provision of constituency service in Canada: ethnicity, LGBTQ+, and language. We complement these audit studies with survey experiments administered to candidates of the 2021 Canadian federal election and a representative sample of Canadian citizens to provide a benchmark for comparing the responsiveness of constituency offices. In the second part of the research note, we consider how different individual and contextual factors might condition inequality in service responsiveness.
Responsiveness in Service Requests
Research on representatives’ responsiveness to service requests from constituents is located at the intersection of Pitkin (Reference Pitkin1967) and Eulau and Karps's (Reference Eulau and Karps1977) concepts of substantive representation. For Pitkin (Reference Pitkin1967), substantive representation implies that representatives should act for and be responsive to citizens in the policy positions they take and in their actions. Eulau and Karps (Reference Eulau and Karps1977) extended Pitkin's concept of substantive representation in identifying four areas through which it may be achieved: policy, service, allocation and symbolic. As such, Eulau and Karps criticized prior empirical studies which mostly focused on policy responsiveness (for example, Miller and Stokes, Reference Miller and Stokes1963) and neglected the other dimensions.
Of particular interest for our study, service responsiveness relates to the constituency service and case work that elected representatives perform for their constituents, as well as representatives’ responses to constituents’ requests. Service responsiveness also stresses the importance of communication in the representative-constituent relationship, which is a core component in most models of representation and accountability (Mansbridge, Reference Mansbridge2003; Pitkin, Reference Pitkin1967; Eulau and Karps, Reference Eulau and Karps1977; Dryzek, Reference Dryzek2009). To gauge citizen preferences and interests, representatives should contact and interact with their constituents. At the same time, constituents may also signal their preferences, needs and demands in contacting representatives. Citizens, interest groups or local organizations often contact representatives for help with problems they face in the constituency or with respect to administrative issues (for example, social benefits, employment insurance or immigration). An elected representative thus often acts as an ombudsperson in trying to help solve these personal and local problems (Eagles et al., Reference Eagles, Gidengil and Bastedo2014). This is true for both the provincial and federal levels of government in Canada (Franks, Reference Franks2007).
Canadian studies underline the importance of constituency service for citizens (ERRE report, 2016) and Members of Parliament (MPs) (Docherty, Reference Docherty and Nevitte2002). In fact, MPs in the House of Commons rank constituency service as one of their most important responsibilities (Eagles, Reference Eagles1998; Docherty, Reference Docherty and Nevitte2002). Interestingly, Butler et al. (Reference Butler, Karpowitz and Pope2012) found in the American context that elected officials respond more to service requests than to policy requests. Single-member districts plurality electoral systems common to US Congressional and Canadian elections create strong electoral incentives for such service responsiveness because elected officials are individually accountable to their constituents (Breunig et al., Reference Breunig, Grossman and Hänni2022).
While evidence of policy (Soroka et al., Reference Soroka, Penner and Blidook2009; Blidook, Reference Blidook2012; Koop and Loewen, Reference Koop and Loewen2010) and service responsiveness (Eagles, Reference Eagles1998; Franks, Reference Franks2007; Koop et al., Reference Koop, Bastedo and Blidook2018; Koop, Reference Koop2016) from Canadian legislators is growing, only two studies, to our knowledge, conducted specific audit studies with Canadian elected officials. First, Loewen and MacKenzie (Reference Loewen and MacKenzie2019) used federal institutions to investigate whether MPs and Members of Legislative Assemblies (MLAs) respond similarly to constituents’ requests that are exclusive to the federal or provincial jurisdiction, or to requests that are shared between the two levels of government. Their results indicate that Canadian representatives appear to be responsive to requests that are exclusive to their own jurisdiction or shared between the two levels of government, but are slightly less responsive to requests outside their own jurisdiction. The second audit study examined whether female political aspirants are discriminated against when contacting Canadian politicians (local mayors, MLAs and MPs) for advice about how to become a politician (Dhima, Reference Dhima2022). Corroborating American results cited in the introduction, women received more answers than men, and more so when answered by women and left-wing politicians.
The Study
Our study extends the previous scholarship in considering whether Canadians constituency offices respond to enquiries from ethnic minority constituents and members of the LGBTQ+ community in a similar way to their opposite constituents (that is, non-ethnic minority constituents and non-members of the LGBTQ+ community, respectively). We are also interested in examining whether individuals employing one of the two official languages may get discriminated against when contacting their elected officials. Normatively, we wish constituency offices to be equally responsive to different groups of constituents in considering their demands and requests. This is especially the case in constituency service where citizen solicitations are generally nonpartisan, nonpolitical and mostly oriented towards governmental programs (for example, Human Resources and Skills Development Canada, Health Canada, Citizenship and Immigration, and the Canada Revenue Agency). Formally, we also hope for equal responsiveness towards the two official languages. The Canadian Official Languages Act (article 2a) “[ensures] respect for English and French as the official languages of Canada and [ensures] equality of status and equal rights and privileges as to their use in all federal institutions, in particular with respect to their use in parliamentary proceedings, in legislative and other instruments, in the administration of justice, in communicating with or providing services to the public and in carrying out the work of federal institutions” (Canada, 2024).
Normative expectations, however, often differ from the state of affairs. As reported in our review of the scholarship on audit studies, minority groups (especially Black, ethnic minorities and low-income individuals) are often discriminated against when contacting their elected representatives. This could be the result of prejudice and negative out-groups attitudes combined with a perceived lack of electoral incentives associated with these groups (Butler, Reference Butler2014). Importantly, evidence of systematic discrimination still plagues Canadian society and political institutions (Godley, Reference Godley2018; Thompson, Reference Thompson2008). For example, racial economic inequality (Banting and Thompson, Reference Banting and Thompson2021), police racial profiling (Cole, Reference Cole2020) and discrimination in hiring (Banerjee et al., Reference Banerjee, Reitz and Oreopoulos2018) are persistent in Canada. Because of the respective minority status of the groups we study, we expect to observe similar discrimination. This is that constituency offices will be less responsive to requests from ethnic minority constituents, members of the LGBTQ+ community, and francophones (in English Canada) and anglophones (in Quebec). While discrimination against ethnic minorities and LGTBQ+ individuals might be driven by explicit or implicit prejudice (Greenwald and Banaji, Reference Greenwald and Banaji1995), we acknowledge that the mechanism might be different with respect to language. In this case, the absence of bilingual political staffers and thus expertise in the two official languages could also be the main source of unequal responsiveness and discrimination in the delivery of constituency services.
Note that staff members of Canadian elected representatives generally oversee constituents’ requests and communications. This is an issue in audit studies targeting professionalized constituency offices like those of Canadian MPs and MLAs (Butler, Reference Butler2014). For this reason, we refer to the responsiveness of constituency offices. We prefer this terminology to the responsiveness of staff members since elected officials are still responsible for the behaviours of their staff when they deliver constituency services. Importantly, staffers generally respond to constituents’ emails on behalf of their elected MPs or MLAs.Footnote 2 To address this issue, we also consider candidates running in elections (potential elected officials) so that we can directly assess their behaviour. We accomplish this by conducting survey experiments on candidates.
Methodology
Audit studies
To investigate situations of inequality in service responsiveness, we conducted two separate audit studies. In the first study conducted in June 2019, we randomized whether the constituent contacting the representatives was from the LGBTQ+ community or not. In the second study conducted in December 2020, we randomized whether the constituent was from an ethnic minority community or not. Both field experiments targeted about 1,000 constituency offices across Canada—338 offices of MPs at the federal level and 650 MLA offices across all provincial legislatures.Footnote 3 When contacting the offices of MPs, the language treatment was superposed to these previous treatments: half received the email request in French and half in English. For provincial MLAs, offices of elected officials from Quebec received emails in French while elected officials outside Quebec received emails in English. Enquiries from fictive constituents were made by email, which is one of the standard approaches in the discipline (Butler, Reference Butler2014). Only one email was sent to each representative. Blocked randomization for legislature and right-wing parties was applied to ensure balanced treatments (see the results of balance tests in the Appendix).
When developing the email messages, we opted for a service rather than a policy request since legislators are generally more responsive to service requests (Butler et al., Reference Butler, Karpowitz and Pope2012). In reviewing the messages used in previous audit studies, we did not find clear guidelines aside from trying to create realistic and impartial requests. We indeed found a variety of messages such as questions pertaining to unemployment benefits (Broockman, Reference Broockman2013); steps to get citizenship (Butler et al., Reference Butler, Karpowitz and Pope2012); scholarships for college (Landgrave, Reference Landgrave2021); declaring a home sale in tax returns (Butler, Reference Butler2014); or student loans and waiting lists for family doctors (Loewen and MacKenzie, Reference Loewen and MacKenzie2019). We thus decided to develop our own original messages that we believe are realistic in the Canadian federal/provincial contexts and are relatively easy to answer. Importantly, we aimed to craft messages typical of service requests, genuinely nonpolitical, nonpartisan and asking for advice with respect to a governmental service. We do not expect, therefore, our communications to be necessarily a representative sample of all types of email requests received in constituency offices. This may influence our assessment, and we discuss this issue in our conclusion.
For the LGBTQ+ treatment, the email asked about the procedure to be recognized as common-law partners for income taxes (see full email verbatim in the Appendix). The LGBTQ+ email specifically asked: Are there any differences for same-sex partners like us? While income tax returns are an exclusive federal responsibility (outside Quebec), it is well-known that same-sex partners have the same rights as heterosexual couples with respect to tax returns.
For the ethnicity treatment, the email asked about the procedure for homeschooling one's children. In the ethnicity treatment, the only difference in the verbatim was the name of the fictive constituent for which we randomly assigned one of four sounding ethnic minority names (Wei Lam, Qiang Chen, Mohammed Bensouda or Arjun Mukherjee)—see the Appendix for the procedure we followed to choose these names.Footnote 4
Our outcome variable of interest is whether the office of the elected official responds substantively to the constituent email (1) or not (0). A positive response (1) includes an appropriate answer to the email (that is, same-sex couples can follow the same procedure as heterosexual couples when completing their taxes; parents must contact their school boards) or a redirection towards the government websites with the exact information (that is, the relevant website page of the Canada Revenue Agency; a provincial or organization website that details the steps to follow in one's province—there are slight variations across provinces). Because education is a policy domain exclusive to provinces and the expected answer is less common knowledge than the one about income tax returns, we also coded a MP’s redirection of the constituent towards the provincial MLA with full contact information as a positive answer. Two research assistants coded the representatives’ answers correspondingly and showed an intercoder agreement of about 95 per cent. The principal investigator verified each of the coders’ disagreements to make a final decision.
Candidates and citizens’ survey experiments
To complement these audit studies, we conducted one online survey with candidates of the 2021 Canadian federal election from the five main parties represented in the House of Commons (N=325) and another survey with 5,762 Canadians representative of the voting age population.Footnote 5 We conducted these survey experiments for two reasons. First, as discussed above, staff members in constituency offices are generally those who have the responsibility to respond to constituents’ emails. Conducting a survey experiment with candidates from the 2021 federal election allows us to analyse for discrimination among individuals who could potentially be elected and eventually in charge of constituency offices. We could then evaluate if (non-)discrimination at the candidacy stage is transposed in actual constituency work. Second, in focusing only on political elites, typical audit studies cannot evaluate whether this group discriminates more or less than the citizens they represent. As such, our research design, which mobilized both typical audit studies as well as candidate and citizen survey experiments, allows us to draw some comparisons with respect to the presence of discrimination across the three groups. For example, constituency office and candidates may (or may not) exacerbate possible discrimination found among the public. Studying this citizenry-candidacy-office linkage is important to better contextualize possible discrimination among political elites.
As such, to mimic the experimental treatments of our audit studies, we asked candidates and citizens to answer the following question:
Constituents contact their federal representatives for several reasons and in different languages. On a scale from 0 to 10, where 0 means “not important at all” and 10 means “extremely important,” indicate the importance of answering the following constituents’ emails for Members of Parliament.
In sequence, the respondents then read the same email as those used in the audit studies and were randomly assigned to the treatment (minority alias) or control group (non-minority alias). To reproduce the language treatment, we added a new message in French and in English asking for references about government programs or local organizations that could help the constituent start a small business.Footnote 6
For these survey experiments, our dependent variable is the score assigned by the respondent on the 0–10 importance scale. For the sake of simplicity and clarity in the visualization of the next results, we multiplied the citizens/candidates scale by 10 to make it correspond to the 0–100 scale associated with the response rates of representatives.
Empirical Strategy
We use OLS regressions to analyze the data. Compared to a logit regression model, OLS regressions simplify the interpretation of the coefficients that directly indicate the average difference in response rates across categories of the independent variable. The substantive results are also comparable to results generated by logit regressions (Angrist and Pischke, Reference Angrist and Pischke2009). In the audit studies models, we controlled for right-wing parties and for region because of associated block randomization, which helps to minimize the standard errors (Gerber and Green, Reference Gerber and Green2012). For each treatment, we estimated an additive model where the outcome variable is regressed on the treatment variable and other covariates for audit studies models. All regression results are presented in the Appendix. In the next section, we present graphically the main results of our analyses.
Results
In Figure 1, we first present the response rates of MLAs and MPs constituency offices under each treatment. We also display the importance candidates and citizens assigned on average to each email scenario. The dots represent the average outcome for a given group/treatment and the vertical lines indicate the 95 per cent confidence intervals. In terms of constituency offices, we notice that MP constituency offices appear to be more responsive (average of 50–55%) than those of MLAs (average of 34–39%) in both experiments. We must recognize, however, that the high levels of responsiveness found in MP constituency offices in the ethnicity experiment is due to our coding choice of processing redirection towards MLAs as a responsive answer (90% of these positive answers). The results also confirm that MLA offices were comfortable answering the email on income tax returns since their response rate is higher than for the email on homeschooling (that is, the message used under the ethnicity treatment). Interestingly, it appears that candidates ranked answering the three emails as more important than citizens. The email request about governmental programs for starting a business (right-panel) also generated more reactivity. A possibility here, however, is that the language treatment raised respondents’ awareness of our research objectives and triggered a social desirability bias. If this was the case, responsiveness to the non-mother tongue language would be higher in the next analyses, but this did not occur (see Figure 3).
Figure 1. Average of the outcome variables across studies
In Figure 2, we present the average treatment effects (ATE) for each study. The results contradict our expectations that ethnic minority individuals and those of the LGBTQ+ community are discriminated against when contacting their elected representatives in Canada. Both MLA and MP constituency offices respond similarly to the email request from senders associated with a minority group and a non-minority group (see the blue and red horizontal lines representing the 95% confidence intervals that overlap the vertical line at zero). Interestingly, results from the survey experiments indicate that candidates actually discriminated positively in favour of ethnic minority aliases and members of the LGBTQ+ community. In each treatment, candidates prioritized the importance of responding to the ethnic minority and LGBTQ+ emails by about 8 points and 17 points, respectively. Those differences are statistically significant at the 0.1 and 0.05 levels. Citizens also appear to advantage LGBTQ+ senders (by about 9 points), but not ethnic minority senders.
Figure 2. Summary of the ATE across studies
While the results associated with the ethnicity and LGBTQ+ treatments highlight an absence of discrimination against those two minority groups, the results at the bottom of Figure 2 underline that emails sent to MPs in French were less answered than emails sent in English (the differences are 19 points in the ethnicity treatment—red empty circle—and 16 points in the LGBTQ+ treatment—red filled circle). Citizens also discriminated against the French email in the survey experiment, but candidates did not.
A note on social desirability bias
Some readers may be concerned that a possible social desirability bias (SDB) is affecting the results of the survey experiments. SDB refers to respondents’ willingness to offer answers appearing socially appropriate and/or to avoid embarrassments (Holbrook and Krosnick, Reference Holbrook and Krosnick2009). We doubt that the results are severed by this issue for several reasons. On the one hand, we think that our research design diminishes SDB in assuring respondents’ anonymity. By design, respondents were also only displayed one survey question at a time and never saw the messages from minority and majority groups simultaneously, as it is typical in feeling thermometer questions that this might trigger SDB. Moreover, the fictive emails were about nonpolitical administrative/service requests that should have made it difficult for respondents to detect our research motives associated with measuring discrimination. Furthermore, the results do not offer strong evidence of SDB given the observed variation across the importance scores assigned to each minority email by candidates and citizens, respectively. For example, if SDB was at work and respondents wanted to appear as non-discriminating or even favouring minority groups, we assume that the language survey question—probably the item that might have triggered greater suspicion—would not have generated non-discrimination from candidates and yet discrimination from citizens. Finally, while other experimental designs such as conjoint experiments could have maybe decreased SDB (Horiuchi et al., Reference Horiuchi, Markovich and Yamamoto2022), it would have jeopardized our objective of comparing the citizens-candidacy-offices results.
Determinants of Inequality in Service Responsiveness
We examined several individual and contextual factors that might presumably condition responsiveness to constituents’ requests. For example, youth (Weiss, Reference Weiss2020) and left-wing parties (Bakker et al., Reference Bakker, Jolly and Polk2012; Marks et al., Reference Marks, Hooghe, Nelson and Edwards2006) are generally more open to ethnic minorities and social diversity, which might foster responsiveness to the minority groups we study. In terms of the political context, elected officials who are in electorally competitive ridings may have more incentives to be responsive to their constituents (Dropp and Peskowitz, Reference Dropp and Peskowitz2012; Giger et al., Reference Giger, Lanz and De Vries2020). None of these conditional factors—specified in our registered pre-analysis plan—proved to be systematically significant in our field and survey experiments. For this reason, full results and their associated methodology are presented in the Appendix of this research note.Footnote 7
We also consider the relationship between descriptive and substantive representation, which has been reported in previous audit studies (for example, the meta-analysis by Costa, Reference Costa2017). Descriptive representation means that elected representatives share similar individual characteristics as the individuals they represent, such as ethnicity, gender, religion, and class. Descriptive representation may represent an important condition for improving the substantive representation of underprivileged groups (Phillips, Reference Phillips1998). In our analyses, we did not find that constituency offices with LGBTQ+ or ethnic minority representatives and corresponding citizens/candidates prioritized the emails from LGBTQ+ and ethnic minority aliases more than others (see Tables A5 and A6 in the Appendix).Footnote 8 This contrasts with the American findings showing that Black and Latino representatives generally respond more to communication from Black (Broockman, Reference Broockman2013) and Latino constituents (Mendez and Grose, Reference Mendez and Grose2018), respectively.
A reason for these null results may be that staff members do not necessarily share the identities of their representative. Note, however, that this issue pertains to American studies as well, which indeed found a connection between descriptive and substantive representation as reported above. Even if staffers do not necessarily share the same LGBTQ+ and ethnic identities as their representatives, staffers of LGBTQ+ or ethnic minority representatives are presumably more sensitized and sympathetic to those communities and may act upon those considerations. Importantly, Grose (Reference Grose2011) showed that having Black legislators in the American context is a key predictor of hiring Black staffers who more likely self-identify with the Black community. Such comparable data does not exist in the Canadian context with respect to the groups we study, but we had assumed that a similar hiring process was favoured and could have led to the expected results. This is, however, what we found with respect to the language of the elected officials.
In Figure 3, we examine whether constituency offices of representatives speaking a given official language to be more (less) responsive to constituents of their own (other) language group when being contacted by constituents. The graph summarizes the results of interactions between MPs, citizens and candidates’ language and the language of the emails (full results are presented in section 4 of the Appendix). We identified MPs preferred official language based on the information provided on the Parliament's official website and those of citizens/candidates based on the language of the survey questionnaire they choose. The graph reports the average marginal effects of the French email treatment when an English (left-panel) and French respondent (right-panel) received the request.
Figure 3. Average marginal effects of French emails for English and French MPs/candidates/citizens
Constituency offices of MPs whose preferred language of communication is English were less likely to respond to the French email by about 23–26 percentage points (these differences are statistically significant at the 0.001 level). A similar pattern emerges in the survey experiment of the citizen survey, but not in the candidate survey. Constituency offices of MPs having French as their preferred language are slightly more likely to respond to emails sent in French and thus discriminate against English emails (dots to the right of zero indicates that French respondents answer the French emails more and thus answer the English emails less), but the differences are more marginal and not statistically significant at the 0.05 level. The small number of francophone MPs (65–77) in each study might explain the lack of statistical significance. Yet, a robustness check indicates that the difference is statistically significant at the 0.1 level when we combine MPs from the ethnicity and LGBTQ+ studies.Footnote 9 The smaller effect among francophone respondents is probably due to the greater rate of English-French bilingualism among francophones (32% among Quebec francophones) than anglophones (5% for anglophones outside of Quebec) (Statistics Canada, Reference Canada2021).
Finally, for several reasons, constituency offices may also privilege minority groups that are more present in their ridings. First, there is no doubt that constituency services are conceived by elected officials and their staff members as an important tool for helping with their re-election; especially in a single-member district electoral system like in Canada (Grose Reference Grose2011; André et al., Reference André, Bradbury and Depauw2014; Heitshusen et al., Reference Heitshusen, Young and Wood2005; Dropp and Peskowitz, Reference Dropp and Peskowitz2012). Consequently, constituency services and communications may be skewed towards communities and groups that appear as more electorally advantageous. As such, electoral incentives associated to a minority group are proportional to the size of the group presence (Grose, Reference Grose2011). Second, contact theory predicts a decrease in individual prejudices towards out-groups as contact increases across groups, which is a function of the group's presence (Allport, Reference Allport1954; Pettigrew, Reference Pettigrew1998). Third, representatives should become, presumably, more sensitive to the realities of a group as its size increases in the riding. With respect to language, the odds of having representatives and political staffers that speak one of the official languages also increases when the size of either the francophone or anglophone community grows in that riding.
While the exact mechanism connecting minority group presence to service responsiveness needs to be better understood, previous audit studies confirm that the offices of representatives become more responsive as the size of a minority group increases in a district (Einstein and Glick, Reference Einstein and Glick2016; Rivera-Burgos and Rubio, Reference Rivera-Burgos and Rubio2024). As we report in the Appendix (see section 5), this relationship between the descriptive representation of minority groups in ridings and their service representation does not appear to be statistically significant with respect to ethnic minority communities neither in the audit studies nor the survey experiments.Footnote 10 As such, we only display below how the presence of French language speakers in a riding significantly condition the response rate of constituency offices to French emails.
In Figure 4, we display the results of the interactions between the percentage of the riding's population that knows French (data from Statistics Canada) and the language of the sender (full results are presented in section 5 of the Appendix).Footnote 11 The presence of French language speakers in a riding proved to significantly condition the response rate of MPs’ constituency offices (the interaction is significant at the 0.01 level), candidates (p-value of the interaction term equals 0.14) and citizens (interaction significant at the 0.01 level).Footnote 12 Figure 4 plots the average marginal effect (AME) of the French email for each group of respondents, given the proportion of the population knowing French in federal electoral districts.Footnote 13 MPs’ constituency offices (panels A and B) are less likely to respond to the request sent in French when there are few francophones living in their riding. For example, the average response rate to the French email is about 24–28 percentage points lower than to the English email when there is about 10 per cent of the population who speaks French in a riding (this is about 20% of the districts across Canada). Similarly, citizens rank the French email less importantly by about 16 points under the same context. The results also indicate that the AME of the treatment decreases in magnitude and comes closer to zero as the proportion of francophones increases in a riding. As the proportion of francophones exceeds 50 per cent, the AME becomes positive and statistically significant (in the ethnicity experiment and citizen study), which indicates that respondents also discriminate against English requests when living in a francophone riding. Note that the AME is also statistically significant (p<0.05) under francophone ridings when we combine the ethnicity and LGBTQ+ studies given the small number of representatives living in such districts in each study.
Figure 4. AME of French emails across proportion of citizens in ridings knowing French
Discussion
We believe that our study complements the scholarship on audit studies conducted with legislators in considering two original treatments (LGBTQ+ and language) and replicating a common one (ethnicity) by auditing the constituency offices of Canadian elected representatives, which are under-investigated in this literature. We also replicate our field experiments with similar survey experiments conducted with candidates of the 2021 Canadian federal election and a representative sample of Canadian citizens.
In contrast with previous studies, our results indicate the absence of discrimination towards constituents from an ethnic minority group or who identify with the LGBTQ+ community. Candidates of the 2021 federal election were even more likely to prioritize those minority groups (this was also the case for citizens towards LGBTQ+ senders). Those results are positive and meet normative expectations with respect to service responsiveness from the constituency offices of Canadian legislators. A caveat, of course, is that this conclusion of no-discrimination is circumscribed to the specific emails we sent to constituency offices. As in other audit studies, we attempted to write typical service requests that were not specifically aimed at generating strong discrimination. As such, our objective was to investigate possible discrimination in typical constituent requests for nonpolitical and nonpartisan advice with respect to governmental services. In this specific setting we created, we found no evidence of discrimination towards LGBTQ+ and ethnic minority senders. We may presume comparable results with other similar emails, but this would need to be validated in future studies. In addition, it will be important to audit constituency offices based on more sensitive issues that may trigger greater discrimination such as police racial profiling, (il)legal immigration or the Reconciliation process, among other examples. As such, we do not claim that our emails are a representative sample of all the constituents’ communications being sent to constituency offices across Canada. It is highly likely that other types of requests would provoke obvious discrimination that would be consistent with what is observed in other spheres of Canadian society (Godley, Reference Godley2018; Thompson, Reference Thompson2008). These are important considerations to address in future research.
We found, however, systematic evidence that emails sent in French were less likely to be answered than those sent in English. Constituency offices of anglophone MPs and those representing ridings with a small proportion of francophones were significantly less likely to respond to French emails. A similar pattern was observed among offices of francophone MPs towards emails sent in English, but the level of discrimination was more moderate. The results of the survey experiments also indicate similar discrimination from citizens but not from candidates. This last result infers that a lack of bilingual staff in MPs’ constituency offices probably drives discrimination towards the two official languages in Canada.
We believe that different solutions could be implemented to address inequality towards the two official languages. First, MPs should privilege hiring bilingual staffers for positions that require screening emails and responding to the public—especially in districts with a significant francophone/anglophone minority population. Second, MPs from the same party/region could collaborate in redirecting emails written in the minority language of the region to a “shared” bilingual staffer who is designated to process those communications. Third, training for MPs and political staffers should educate these individuals about the language barrier many constituents face when contacting constituency offices. Awareness of discrimination has proved to be an effective tool for reducing prejudice (Hsieh et al., Reference Hsieh, Faulkner and Wickes2022). Fourth, translation options integrated to email providers are becoming more and more accessible. Such software should be available to MPs and their constituency office. MPs and staffers should also be trained to use them efficiently. More than 50 years have passed since the adoption of the Official Languages Act—several solutions now exist to facilitate bilingual communication between constituents and their constituency office in Canada.
Interestingly, we did not find substantial differences across the behaviours of citizens and political elites. In our analyses, citizens appeared slightly less favourable to minority groups than candidates, but did not necessarily discriminate more than constituency offices. Overall, our results suggest that staff members in constituency offices behave quite similarly to average citizens. This conclusion comports with the view that citizens and political elites’ decision-making are often much more similar than initially suggested (Kertzer, Reference Kertzer2022).
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0008423924000817
Open access
