Introduction
The shifting landscape of party-union relationships has rekindled interest in the vote intentions of union members in Canada. Multi-partisan endorsements by unions have become more common as formal, exclusive ties to the NDP have loosened over time (Savage, Reference Savage2021). Liberal, Conservative, Bloquiste, New Democrat, and Green candidates have all managed to secure formal union backing in recent years, a clear indication that the partisan breadth of union support has widened. Moreover, as unions become more overtly transactional in their approach to electoral politics, parties have seemingly become more interested in wooing union leaders in pursuit of coveted endorsements (Savage and Westlake, Reference Savage and Westlake2025) and union members’ votes.
Of course, union endorsements and union votes are different things. This article is concerned with the latter. Through a survey of union members, we gathered data to analyze their vote intentions with a view to better understanding the changing nature of party-union relationships, the voting patterns of union members, and the diversity of vote intentions within the union movement. Our focus on the latter allows us to build on the quantitative findings of Keith Archer, whose decades-old study of trade union affiliation to the NDP found not only that union members were more likely than the average voter to support the NDP, but also that members of NDP-affiliated unions were even more likely than union members in general to back the party (Archer, Reference Archer1990). Our unique data set allows us to extend the analysis to differentiate beyond NDP affiliated and non-affiliated union members, to public and private sector union members, and construction and non-construction trade union members. The differences among these union groupings are significant and worthy of analysis.
Though our survey data only covers union members outside of Quebec, we also draw on data from the Canadian Election Study (CES) to complement our analysis. Our findings reconfirm some long-held assumptions about labour politics and challenge others. Specifically, we found that while union membership continues to make voters more likely than the average voter to support the NDP, intersecting factors such as union type, gender, education, age, and income render that support less uniform than is often assumed, and therefore opens strategic spaces for other parties to make headway among specific pockets of union voters. In particular, our findings suggest the Conservatives have made inroads with members of construction unions as well as those with trade and college qualifications.
Literature review and theory
Union member vote choice in context
While the Canadian Labour Congress (CLC), most provincial federations of labour, and many of the unions affiliated to these organizations were founding partners of the NDP at its 1961 launch, New Democrats have never been able to rely on union members as a voting bloc capable of vaulting the federal party into government. A considerable body of research on organized labour and the NDP has highlighted the gap between union leaders’ and members’ support for the party. In 1976, Robert Laxer noted that while many national union leaders were NDP supporters and activists, most local unions in Canada either remained non-partisan or provided only “perfunctory” support to the party (Laxer, Reference Laxer1976, 263). Similarly, Desmond Morton, a party insider, argued that “the few unions that found the courage and the cash to survey their own members’ attitudes soon discovered that few of them had any allegiance to the labour movement’s political or social goals nor even to their own elected leaders” (Morton, Reference Morton1998, 315).
Researchers have made sense of the disconnect between union voters and partisan choice by pointing to various contextual factors in Canadian federal politics. Throughout the 20th century, federal politics in Canada primarily revolved around ethnocultural, religious, linguistic, and constitutional debates, particularly concerning Quebec’s position within the country (Johnston, Reference Johnston2017). The contentious nature of these issues has divided working-class voters, complicating the NDP’s electoral appeal and thwarting the party’s efforts to realign Canadian politics along a left/right axis (Brodie and Jenson, Reference Brodie and Jenson1988). Moreover, doubts about the NDP’s electoral viability have led some union members to strategically vote for the Liberals, even if they may have preferred the New Democrats, thus weakening the NDP (Johnston, Reference Johnston2017).
Both the NDP and the labour movement have undergone significant changes since the 1960s. The NDP was the creation of male-dominated private sector industrial unions that were representative of the labour movement in the immediate post-war period. While the share of workers covered by union contracts has remained relatively steady for many decades, the rise of public sector unionism and the growing proportion of women entering the workforce has significantly altered the composition of the labour movement. For example, women now constitute a majority of union members in Canada and union members in the public sector now outnumber those in the private sector (Ross and Savage, Reference Ross and Savage2023, 188, 184).
Nevertheless, research findings have quite consistently indicated that union membership increases the likelihood of a vote for the NDP. Richard Johnston’s research demonstrates a consistent, positive relationship between union membership and an increased likelihood of voting for the NDP (and its forerunner the Co-operative Commonwealth Federation). Studies of federal elections in 1965, 1968, 1974 and 1979 revealed that union members were approximately twice as likely as non-union members to support the NDP (Archer, Reference Archer1990, 56-58; Archer, Reference Archer1985, 357). Johnston, Blais, Brady, and Crête found that union households were 8 points more likely than a non-union household to vote NDP in 1988 federal election (Johnston et al., Reference Johnston, Blais, Brady and Crête1992, 89-90) and Erickson and Laycock found that being in a union household was associated with individuals being around 1.6 times more likely to support the NDP than non-union voters in the 2004 election, a trend continuing in 2006 and 2008 (Erickson and Laycock, Reference Erickson, Laycock, Laycock and Erickson2015, 50-51). In more recent research, Polacko, Kiss and Graefe found that the level of disproportionate support for the NDP from union households had remained mostly stable, but that “there is some tentative evidence of decline in support for the NDP by voters in union households, with support dropping in 2015 and 2019 and reaching near historic lows in the latter” (Polacko et al., Reference Polacko, Kiss and Graefe2022, 678).
Despite the historically positive correlation described above, NDP support among union voters did not surpass 25 per cent in any federal election study conducted between 1968 and 1984 (Archer, Reference Archer1990). Figure 1 estimates union vote share by party using data from the CES for union households.Footnote 1 While the NDP consistently does well with respondents from union households relative to its overall vote share, the party has never won a plurality of the union vote in the rest of Canada and only in 2011 won a clear plurality of the union vote in Quebec.Footnote 2
The 2011 federal election stands out as the notable exception for the NDP with regard to support from union households. During that campaign, CES data found that 35 per cent of union members outside of Quebec supported the NDP, surpassing the 21 per cent who backed the Liberals. The CES data suggests that in Quebec nearly half of union households supported the NDP in 2011 (48%) compared to around 27 per cent for the Bloc Québécois and less than 10 per cent for the Liberals. Ipsos-Reid data found that 40 per cent of union members in Quebec supported the NDP (McGrane, Reference McGrane2019, 245).
Public opinion polls released between elections consistently reinforce the NDP’s disproportionate support among unions members. A string of polls by Ekos and Abacus Data between 2015 and 2023 revealed that union members were more likely than others to indicate support for the NDP (Ekos, 2023). This did not always mean that the NDP was the preferred electoral choice of union members. Liberals or Conservatives were often the first choice of union members, even if union membership makes voters more likely than the average voter to indicate support for the NDP.
Party Vote Share Among Union HouseholdsFootnote 3, Footnote 4 .
Whether the NDP’s failure to win a majority of union voters is a product of strategic voting, the party system, the party’s public policy priorities, the salience of issues unrelated to class, or its record in office at the provincial level is a matter of debate. However, the changing landscape of party-union relationships appears to be making the NDP’s task even more difficult. While exclusive partisan relationships between unions and the NDP have steadily declined, ad hoc electoral alliances between unions and Liberals (and to a lesser extent Conservatives) have gained steam (Ross and Savage, Reference Ross and Savage2024; Savage, Reference Savage2010). It is in this dynamic environment that we surveyed the voting intentions of union members.
Determinants of union members’ vote choice
In this article we consider two sets of variables that shape union members’ vote choices: union characteristics and the demographics of union members. These are important variables to consider because unions are heterogeneous organizations with varied cultures, ideological traditions, and memberships. This diversity opens up strategic opportunities for parties to engage in voter segmentation and make electoral pitches that appeal to different types of unions as well as union members who share similar demographic characteristics.
We start with an expectation that union members will be more likely than others to indicate support for the NDP. While the Liberal Party pivoted to become more pro-union under the leadership of Justin Trudeau, the NDP remains the labour movement’s most reliable ally on the public policy front, consistently supporting labour movement priorities and opposing incursions on workers’ rights in the form of back-to-work legislation. The NDP’s decision to prioritize the union movement’s longstanding demand for a ban on replacement workers as part of its 2022 confidence and supply deal with the Liberal government stands out as a particularly compelling example of this dynamic. While shared political interests between the NDP and the union movement are evident, historic ties should not be discounted (Pilon et al. Reference Pilon, Ross and Savage2011; Savage, Reference Savage2021). An entire generation of labour activists were exposed to union education programs that consistently and exclusively trumpeted the NDP as the political arm of the labour movement. Moreover, unions continue to be an important source of NDP candidates and volunteers, and the composition of the NDP’s parliamentary caucus and governing council has always featured a much larger share of labour movement operatives and union members than is found in the Liberal or Conservative parties (Geddes, Reference Geddes2011). Even activists from unions like CAW/Unifor, which broke formal partisan ties to the NDP in 2006, still routinely and almost exclusively run as NDP candidates in federal and provincial election campaigns (Ross and Savage, Reference Ross and Savage2024). In other words, a weakening of institutional ties has not entirely extinguished personal and ideological bonds between union activists and the NDP (Jansen and Young, Reference Jansen and Young2009; Pilon et al. Reference Pilon, Ross and Savage2011).
While we expect the NDP to have disproportionate support among union members, we also expect the party’s appeal to vary depending on union type. There is good reason to believe that vote choice between members of public sector unions and members of private sector unions will differ based on political-economic structure, history, and the different relationships that public and private sector unions have with government.Footnote 5 In the case of public sector workers, government is also the employer. Presumably, this gives public sector workers more of an incentive to affect electoral outcomes. Indeed, comparative European research has found that health, education, and public service production workers are more likely than private sector workers to vote for left-wing parties (Tepe, Reference Tepe2012). However, public administration workers surveyed as part of the same study did not display political attitudes or behaviours that differed significantly from private sector workers. Thus, while there is significant diversity within the public sector along occupational lines, workers in the sector as a whole tilt to the left with regard to political attitudes and vote choice.
Our expectation is that, on the whole, public sector unions members in Canada would likely prefer governments that show a greater commitment to defending and enhancing public services, not only out of self-interest but also for the benefit of the publics they serve. Conversely, public sector union members are presumably less likely to support parties seeking to shrink the public sector and restrict public sector union rights.
As the federalist party most committed to the expansion of public services and the preservation of union rights, one might expect the NDP to be the preferred choice of public sector union members in the rest of Canada. Of course, that hypothesis is complicated by history and the party’s track record in government provincially. The NDP was essentially the creation of private sector unions, and the party’s institutional ties to public sector unions have historically been weak. NDP-union friction has been most evident in the public sector in provinces where the NDP forms government (Evans, Reference Evans, Ross and Savage2012). Essentially, labour expects NDP governments to be model, pro-union, employers and when the party lets down its public sector union supporters, those supporters do not easily forgive and forget (Savage and Smith, Reference Savage, Smith, Ross and Savage2013).
The Liberal Party has an equally complex, yet different, relationship with public sector unions. Historically, the party’s responsiveness to the needs, wants, and wishes of public sector unions has varied based on the dominant ideological paradigm of the period. In the 1990s Chrétien era, the Liberals oversaw significant cuts to public services and incursions on workers’ rights that alienated the labour movement (Panitch and Swartz, Reference Panitch and Swartz2003), but over the course of the 2000s the Liberals have shifted to the left, with members and voters who increasingly see themselves as left wing (Johnston, Reference Johnston, Hopkins and Sides2015). Under Justin Trudeau, the Liberals worked to build bridges with public sector unions through the repeal of anti-union legislation and significant investments in new social programs. However, they have also resorted to back-to-work legislation to extinguish public sector strikes, thus alienating segments of the union movement (Ross, Reference Ross, Macdonald, Scott and Trew2024).
Given this mixed bag, the expected level of support for Liberals and the NDP from public—as opposed to private—sector union members is difficult to predict. Thus, we have two competing hypotheses related to NDP support among private and public sector union members.
What is less contentious is the assumption that public sector union members are less likely to support Conservatives than their private sector union counterparts (Tepe, Reference Tepe2012). Public sector unions see the Conservatives as the party least committed to the preservation of public services and trade unions rights (Savage and Smith, Reference Savage, Smith, Ross and Savage2013). While the Conservatives have made unprecedented efforts in recent years to win over blue-collar working-class voters in the private sector, these overtures have rarely extended to the public sector.
H1a: Members of public sector unions will be less likely to support the Conservatives than members of private sector unions.
H1b: Members of public sector unions will be more likely to support the NDP than members of private sector unions.
H1c: Members of public sector unions will be less likely to support the NDP than members of private sector unions.
We also expect unions’ affiliations to the NDP to shape their members’ support for the party. The NDP’s Constitution provides for institutional affiliation by labour organizations. While most unions do not have institutional ties to the NDP, those that do proudly endorse the party at election time and send labour delegates to NDP conventions. NDP affiliated unions routinely engage in activities and initiatives designed to increase the likelihood their members will vote NDP. Keith Archer showed that in the 1979 federal election 30 per cent of members of NDP-affiliated unions reported voting for the party compared to just 20 per cent of non-affiliated union members (Archer, Reference Archer1985).
H1d: Members of unions affiliated to the NDP should be more likely to support the NDP than members of unions that are not affiliated to the NDP.
Finally, we expect construction unions to be distinct. Construction unions sometimes have allied interests with employers that do not match the adversarial nature of the labour relations relationship in other sectors of the economy. This divergence is further reflected in political differences between construction and non-construction unions. Construction unions have never really shared formal institutional ties to the NDP. In fact, building and construction trades unions briefly broke away from the Canadian Labour Congress in 1980s, in part, over the organization’s support for the NDP (Rose, Reference Rose1982).
Admittedly, building and construction trades unions, along with the broader labour movement, strongly opposed Conservative attempts to undermine unions’ ability to organize, bargain collectively, and engage in political action around the turn of the 21st century, and mobilized significant resources to defeat Conservative candidates. In recent years, however, we have witnessed a rapprochement between Conservatives and construction unions. The party has abandoned overt efforts to dismantle unions and shown greater interest in consulting and working with these unions. In exchange, construction unions have pulled back from attacking the Conservatives, and in some cases, have endorsed the party.
Under the leadership of Pierre Poilievre, the party worked hard to reposition itself as union friendly. For example, despite strong opposition from the business community, the party raised eyebrows by trumpeting support for free and fair collective bargaining and by supporting a ban on replacement workers in federally regulated industries (Tumilty, Reference Tumilty2024). Conservatives courted the votes of construction union members by positioning themselves as champions of private sector growth while advocating for public sector restraint. Specifically, the Conservative strategy is aimed at attracting blue-collar union voters traditionally opposed to the party’s anti-union policies through populist messaging that lays the blame for economic insecurity on “greedy elites” and a “bloated” public sector (Savage, Reference Savage, Collier and Malloy2024).
This strategic repositioning vis-à-vis organized labour and working-class voters includes populist appeals to deliver “powerful paycheques” by cutting statutory deductions and taxes, therefore easing the cost of living for working-class people (Global News, 2023). Notably, Conservatives address working-class concerns over economic inequality without blaming capitalism. This aligns well with the dominant pro-growth mentality of building and construction unions. Moreover, in their 2022 study, Polacko et al., found that “cultural issues such as moral traditionalism and anti-immigration are now significant drivers of working-class Conservative support” (Polacko et al., Reference Polacko, Kiss and Graefe2022, 663). Thus, the overwhelmingly male membership, cultural conservatism (see for example Prescod, Reference Prescod2020), and pro-development mentality of construction unions make them a natural target for the Conservatives.
The success of these Conservative overtures to labour is likely best observed in Ontario provincial politics. After nearly two decades of strong and organized construction union opposition to the Ontario Progressive Conservatives (PC) based on the party’s overtly anti-union policy positions, the governing Tories secured the support of eight building and construction trades unions in the 2022 provincial election. PC efforts to woo organized labour were dubbed a “labour charm offensive” by the media (The Agenda, 2021). Unlike some of his predecessors, Ford did not share the same rigid ideological opposition to unions. While Ford introduced government-imposed wage restraint legislation on most of the public sector in 2019, he did not set out to dismantle unions, and his government routinely met with key construction union leaders (Hauen, Reference Hauen2022). His Labour Minister, Monte McNaughton, played a key role in establishing positive relations between construction unions and the government, which ultimately led to a flurry of union endorsements during the lead-up to the election. The unprecedented number of PC labour endorsements fed a media narrative of growing union support for the government despite strong opposition from the larger public sector unions upset with PC wage restraint legislation (Savage, Reference Savage, Collier and Malloy2024). Most of those unions, along with the Ontario Federation of Labour, were solidly lined up behind the opposition NDP. Given the dynamics in Ontario, and continued efforts by the federal Conservatives to make inroads among constructions workers, we expect to see disproportionate support for the Conservatives among construction union members.
H1e: Members of construction unions should be more likely to support the Conservatives than members of other unions.
It is important to consider that union members have multiple identities that are likely to shape vote choice. Indeed, union membership may not actually motivate their vote choice at all. We thus consider several demographic variables that we expect to shape how union members vote. We expect the trends among union voters to reflect those of other Canadians, though baseline support for the NDP should be higher for union members.
The first demographic variables we consider are regional. Canadian elections always have a regional component, and we expect regionalism to be a factor for members of unions as for the rest of the population. Historically the Conservatives and the NDP have been particularly strong in Western Canada. This trend can be attributed to a number of factors. The Conservatives have been dominant in the West for over half a century and benefit from their past links to the Reform Party and its mobilization of Western Canadian populism in the 1990s. Conservative opposition to environmental measures that impact fossil fuel development also helps the party in Alberta and Saskatchewan. The main challenge for the Conservatives in Western Canada has come from the NDP, which also benefits from historic strength in the region. Though provincial politics in Canada are distinct from federal politics, both parties (and particularly the NDP) likely benefit from the organizational support of their provincial counterparts in the region (Chhibber and Kollman, Reference Chhibber and Kollman2004). Conversely, Liberal parties are marginal in provincial politics in Western Canada. We thus expect to see union members in Western Canada to be more supportive of the Conservatives and the NDP than union members in Ontario where the Liberal brand is stronger
H2a: Union members in Western Canada should be more likely to vote for both the Conservatives and the NDP than union members in Ontario.
In line with broader research on vote choice, we expect gender to matter to union members’ vote choice. There is clear evidence in the Canadian context that women are more likely than men to hold left-leaning views on a variety issues (Gidengil, Reference Gidengil2022). Historically, women have been less likely than men to vote for parties on the right (Goodyear-Grant, Reference Goodyear-Grant, Bittner and Koop2013). In recent years a clearer gap has developed between the Conservatives and other major parties, with polls suggesting that the Liberals and NDP do better with women than with men (Fournier Reference Fournier2020). The NDP also benefits from its longstanding support of and links to feminist causes including abortion rights and pay and employment equity (Young, Reference Young2000). The trends in vote share across the general population in Canada suggest that women union members should be more likely to vote for the Liberals and the NDP than the Conservatives.
H2b: Women union members will be more likely to vote for the Liberals and the NDP than they will be to vote for the Conservatives.
There is a broad trend across wealthy democracies in which younger voters tend to hold more socially liberal values (Norris and Inglehart, Reference Norris and Inglehart2019). We expect this to be true of union members in Canada as well. The NDP’s long history of social justice activism and progressive social values likely resonates more with younger union voters. Many unions themselves have emerged as leading social justice voices for progressive causes including LGBTQ+ rights, Indigenous sovereignty, and support for the anti-war movement (Rayside, Reference Rayside, Hunt and Rayside2007; Ross and Savage, Reference Savage2021). Moreover, public opinion polling routinely shows that younger voters are more likely to support the party than older generations of voters (Angus Reid Institute, 2023; Colletto, Reference Colletto2016; Leger, 2024). These findings have been confirmed by CES data, though the impact of age on the likelihood of voting NDP is inconsistent across elections (Gidengil et al., Reference Gidengil, Nevitte, Blais, Everitt and Fournier2012; Fournier et al., Reference Fournier, Cutler, Soroka, Stolle and Bélanger2013). On balance, however, we expect the NDP to receive more support from younger voters in unions.
H2c: Younger union members should be more likely to support the NDP than older union members.
In line with recent research on class and Canadian politics, we expect both income and education to influence union members’ vote choice. The Liberals tend to benefit from the support of Canadians with university degrees, while the NDP tends to benefit from the support of lower income voters because of the party’s emphasis on wealth redistribution and income inequality (Kiss et al., Reference Kiss, Polacko and Graefe2023; Westlake, Reference Westlake, Robbins-Kanter, Koop and Troup2025). We have no reason to believe such patterns would be broken by virtue of union membership.
H2d: Union members with university degrees should be more likely to support the Liberals than those without.
H2e: The lower a union member’s income, the more likely they should be to support the NDP.
Data and Methods
To examine union members’ vote-choice we conducted a survey of union members in Canada using the survey firm Ekos.Footnote 6 The survey included three waves, one in September 2023, one in January 2024, and one in April 2024. All respondents were members of unions drawn at random from Ekos’ online panel.Footnote 7 Because there are a limited number of union members in the panel there are some respondents included in multiple waves of the survey. Over the three waves we had 1953 unique respondents. Of those respondents, 1127 (57%) were surveyed only once, 562 (29%) were surveyed in two of the waves, and 264 (14%) responded to all three waves. Our analysis relies on the pooled survey responses, using the most recent response for each respondent. We also conducted a robustness check by analyzing each wave of the survey separately, included in appendix B.Footnote 8 As a data quality check, we excluded those respondents who answered multiple waves and reported ages that suggested they had become younger.Footnote 9
We exclude respondents from Quebec from our analysis for three reasons. First, the link between unions and parties in Quebec is complicated by the presence of the Bloc Québécois as a party with important ties to nationalist segments of the Quebec labour movement (Graefe, Reference Graefe, Ross and Savage2021; Collombat and Lafrance, Reference Collombat and Lafrance2022). Second, the impact of class on vote choice tends to be weaker in Quebec because of the high salience of questions around nationalism (Rivard and McGrandle, Reference Rivard, McGrandle, Robbins-Kanter and Koop2025). Most significantly, however, our data set does not have a variable for respondents’ preferred language, and this is a critical variable to understanding vote choice in the province.
In the survey we ask union members both who they would vote for if the election were held today and, for those who are undecided, whether there is a party they are leaning towards. In the main analysis in the article, we group together those that have decided to vote for a party with those who are leaning towards a party. In online appendix C we run a robustness check excluding all undecideds, including those leaning towards a particular party. The results of this analysis are similar to those in the main body of the article. We exclude respondents who were not leaning towards any party or who supported a party other than the largest three.Footnote 10
We analyze two groups of explanatory variables. The first address union type. We have a binary variable categorizing respondents as members of either public or private sector unions. We use a second question asking respondents to identify their specific union to generate binary variables for whether a respondent is in an NDP affiliated union and whether they are in a construction union. Table 1 shows which unions we classified as NDP affiliated and construction unions.
NDP Affiliated and Construction UnionsFootnote 11
We include a group of demographic variables in our analysis. These include variables for region, gender, age, income, and education level. We break respondents into four regions: Atlantic Canada, Ontario, Western Canada, and the North. We break gender into three categories, men, women, and non-binary. Because we did not have enough respondents from the North or who identified as non-binary to have meaningful estimates for average marginal effects, we suppressed both from the average marginal effects included in the main body of the article (average marginal effects for both are available in the online appendix). We break age into 5 categories (18-34, 35-44, 45-54, 55-64, and 65+) and income into 12 categories. We have 6 income categories rising in $9 999 increments from 0 to $59 999. We have an additional 3 categories rising by $19 999 increments from $60 000 to $119 999. We break education into three categories: those who have no more than a high school education, those with a highest level of education below the level of a bachelor’s degree such as a college degree, apprenticeship, or trade certificate, and those with a university degree at the bachelor’s level or higher.
Because we have three waves of survey responses and support for parties can change over time, we include controls for the last wave of the survey each respondent responded to. For readability these are suppressed in the main body of the article but included in appendix A. To account for possible correlations between the error terms among respondents that answered the same survey wave, we cluster errors by the last survey wave answered.
To account for differences between the demographic distribution of respondents in the survey and the demographics of union members more generally, we applied survey weights to our analysis. The sample was weighted to match the gender, regional, age, and educational distribution of the union population in Canada. We also weighted to account for discrepancies between the proportion of respondents that belong to public sector unions and the proportion that belong to private sector unions. The decision to include or exclude union sector from the weighting calculation does not affect results, as shown in appendix D.
Our analysis uses multinomial logistic regression models, as is appropriate for cases where the dependent variable is vote choice. This matches the approach of Gidengil et al., Fournier et al., and others (Gidengil et al., Reference Gidengil, Nevitte, Blais, Everitt and Fournier2012; Fournier et al., Reference Fournier, Cutler, Soroka, Stolle and Bélanger2013). For ease of interpretation, we show figures with average marginal effects capturing the estimated effect that each of our explanatory variables has on vote choice in the main body of the article. We include full regression tables with average marginal effects, standard errors, and R2s in appendix A.
We supplement our regression analysis with descriptive statistics that show each party’s vote share. We also include regression analysis on CES data that shows that union members have been more likely to vote for the NDP than non-union members. Regression tables for the CES regression analysis are available in online appendix A.
Finally, it is important to acknowledge that our article involves a considerable number of tests, which may increase the likelihood of false positives. However, many of our findings are broadly consistent with our expectations and with analyses conducted on the voting behaviour of the Canadian population as a whole.
Results and Analysis
Union membership and party vote share
Analysis of union member support for parties in our data has the NDP at relatively high levels. We find that between 36 per cent and 40 per cent of union members support the NDP, depending on the month the respondent last answered the survey. This is a relatively high estimate of support for the NDP compared to other polls conducted around the same time.Footnote 12
A number of factors may account for our high estimates for the NDP. First, we only look at responses from outside Quebec, whereas the surveys cited consider both Quebec and the rest of Canada. When we include respondents from Quebec, our estimate for the NDP’s vote share falls to between 30 per cent (for those polled solely in the first wave) and 37 per cent (for those whose most recent response was the second wave or third wave of the survey).
Second, it is important to note that our confidence ranges for both the Conservatives and NDP are quite large. In fact, in two of the three waves, there is no statistically significant difference between the likelihood of a respondent supporting the Conservatives and the likelihood of a respondent supporting the NDP.Footnote 13 As a result, one should not overstate the implications of NDP lead among union members in our survey.
Third, it is possible that sampling bias is leading us to overestimate NDP support. We try to correct for this by using weights,Footnote 14 but we can only correct for differences between our sample and known demographics such as gender, education, and membership in a public or private sector union. We cannot correct for biases that might exist when there are unknown factors that reduce a respondent’s likelihood of responding to a survey. For example, if those that have low level of trust are less likely to answer our survey and more likely to vote for Conservative, our survey will underestimate Conservative support. Without knowing the relative levels of trust of survey respondents and non-respondents, there is no way to correct for this.
Finally, it has been well established that questions that lead respondents to think more about particular beliefs or social identities can influence responses to subsequent survey questions (Zaller, Reference Zaller1992). It is possible that respondents were unintentionally primed to think about their union identity by being asked questions about union membership. Since union membership and support for labour has often been connected to greater support for the NDP, priming respondents to think about their union membership may lead them to be more likely to report support for the NDP.
While our baseline estimate for union member support for the NDP is higher than one might expect, we have no reason to believe this will bias the tests of our hypotheses. We expect the large confidence ranges we find in our analysis in Figure 2 to be accounted for by the standard errors in our regression analysis. If we have inadvertently primed respondents to think about their identities as members of unions, that prime should affect all respondents to the survey. Thus, it should not bias our estimates of the impact that membership in different types of unions or demographics has on the relative likelihood that union members support different parties.
Vote Choice by Union Members Across Survey Waves (ROC).
We cannot use our survey to determine how union members (and those that were part of union households between 2008 and 2015) vote in relation to non-union members because our survey was focused only on members of unions. To determine the relative likelihood of union member support for the NDP, we turn to CES data. Figure 3 shows the average marginal effect of union membership on the likelihood of an individual supporting the NDP across the five most recent elections, controlling for a number of other demographic variables that affect vote choice.Footnote 15 In each election, being in a union household makes one more likely to vote for the NDP than those who are not in union households. The NDP’s disproportionate support among unionized voters, however, has declined over time. Members of union households were 19 percentage points more likely to vote NDP than non-union members in 2008 and only about 7 percentage points more likely to do so in 2019 and 2021.
Average Marginal Effects of Union Membership in Vote Choice.
While the CES data gives us good reason to believe that union members are still more likely than the average voter to support the NDP, it is important to keep in mind that unions are not monolithic organizations. Members of unions may be pushed in different political directions based on union type. Union members also have other identities that may lead them to be either more or less supportive of the NDP. We consider these factors in the next two sub-sections.
Union type and union members’ vote choice
We begin our analysis of the determinants of union member support of political parties by examining the relationships between union type and vote choice. We first look at whether one is a member of public or private sector union. With regard to H1a, as expected, Figure 4 shows that the estimated effect of membership in a public sector union on vote for the Conservatives is negative (and positive for the Liberals). Neither, however, is statistically significant. We find no evidence for H1b that members of public (rather than private) sector unions should be more likely to support the NDP. Rather we find evidence for H1c, that members of public sector unions should be less likely than members of private sector unions to support the NDP. However, while the negative effect that membership in a public (as opposed to private) sector union has on support for the NDP is statistically significant, at 3 percentage points it is not very large.
Effect on Support for Party (ROC).
Given the general composition of public sector unions in Canada, it is not possible to differentiate between health, education, public service production, and public administration occupations based on union affiliation. However, the concentration of a large number of public sector workers in the Public Service Alliance of Canada, a union that has reliably promoted anti-Conservative strategic voting, may provide some analytical clues to make sense of the data (Savage, Reference Savage2021; Collombat and Picard, Reference Collombat, Picard, Marland and Giasson2020). Concerns about the federal NDP’s electoral viability may be prompting some public sector union members to vote strategically for the Liberals in an attempt to block the election of Conservatives. In the context of Canadian federal elections, some have pointed to union-backed strategic voting efforts to help account for the NDP’s relatively poor electoral performance, even among union members themselves (Savage, Reference Savage2021).
Despite the move towards strategic voting, some unions remain steadfast in their support for the NDP. In line with H1d, our findings indicate that workers who belong to unions affiliated to the party are more likely to support the NDP compared to those in unions that are not formally affiliated to the party. Members of NDP affiliated unions are about 10 percentage points more likely to vote for the NDP than respondents that were members of other unions. While the composition of unions affiliated to the NDP has changed since the 1980s, our findings about the effects of affiliation on vote choice are consistent with those of Archer (Reference Archer1985, 360).
The Liberals appear to be the party that struggles most with respondents from NDP affiliated unions. Respondents were about 11 percentage points less likely to indicate support for the Liberals than respondents who were members of unions not affiliated to the NDP. We find little evidence that support for the Conservatives is influenced by whether a respondent’s union is affiliated to the NDP.
Finally, in line with our expectations in H1e, we find that respondents who are members of construction unions are more likely to vote Conservative by quite a substantial margin. We estimate that members of construction unions are about 23 percentage points more likely to support Conservatives than members of other unions. This finding is consistent with the broader trend of increased manual working-class support for the Conservative Party observed by Polacko et al. (Reference Polacko, Kiss and Graefe2022).
The NDP suffers most from the Conservative Party’s disproportionate support among construction union members. We estimate that members of construction unions are about 22 percentage points less likely to vote for the NDP than members of other unions. By contrast, support for the Liberals does not appear to be clearly affected by a respondent’s membership in a construction union.
Overall, our analysis suggests that some characteristics of a union matter to members’ vote choice while others do not. While we find little difference in vote choice between public and private sector union members overall, construction union members as a subset of private sector union members, are much more likely than other union members to support the Conservatives and much less likely than other union members to back the NDP.
By contrast, our results suggest that unions formally affiliated to the NDP positively reinforce support for the party among their members. Notably, the increased likelihood of members of NDP-affiliated unions voting for the NDP corresponds with a slightly greater decrease in the likelihood of those members supporting the Liberals. Meanwhile, NDP affiliation seems to have little effect on the likelihood of members voting Conservative. This suggests that union affiliation to the NDP primarily influences voters who are leaning towards the Liberals or the NDP, drawing them more towards the NDP rather than affecting members across the union regardless of their initial ideological and partisan preferences.
Demographics and union member vote choice
We find evidence that union members’ vote choices reflect some of the broader trends in the electorate, but also that there are important ways in which union members are unique. This is particularly true with respect to the regional effects shown in Figure 4. In line with our expectations, the Liberals do worse with union members in Western Canada than they do with members in Ontario. Union members in Western Canada are about 11 percentage points less likely to vote Liberal than those in Ontario. It is the NDP, however, that most clearly benefits from Liberal weakness among union members in Western Canada. Union members in Western Canada are about 17 percentage points more likely to vote NDP than union members in Ontario.
Interestingly, we find no evidence that union members in Western Canada are more likely to vote Conservative than union members in Ontario. This result is surprising given the electoral dominance of Conservatives in Western Canada. While we cannot confidently explain this with the data we have, we can offer a plausible explanation. It is possible that Western Canada is more polarized along class lines than the rest of Canada with union members more supportive of the NDP and non-union members more supportive of the Conservatives.Footnote 16 It may also be the case that a disproportionate share of union members in Western Canada live in places where the Conservatives are less dominant (Vancouver, Winnipeg, Edmonton, and Victoria) (Ross and Savage, Reference Ross and Savage2023, 187). The fact that we find no evidence that union members in Western Canada are more likely to vote for Conservatives than union members from Ontario suggests a need for further research into the way region shapes union member vote choice.
As expected, we also found that gender matters to union members’ support for political parties. In line with H2b and the existing literature on gender and vote choice among Canadians more generally, women in unions are less likely to vote Conservatives than men in unions. We estimate that women in unions are about 8 percentage points less likely to support the Conservatives than men in unions. In contrast, women in unions are about 10 percentage points more likely to vote for the NDP than men in unions. We find little evidence that gender has an effect on Liberal support among union members.
Our results indicate a significant gender gap in party support among union members, reflecting a broader trend observed in the electorate. Remarkably, this gender gap persists even when accounting for the Conservatives’ stronger appeal among members of construction unions. This suggests that the Conservative Party’s resonance with male union members extends beyond traditional male-dominated sectors like construction. However, without conducting targeted inquiries into the values held by union members, our ability to fully unpack these dynamics within this article remains limited. Future research is needed to delve deeper into the complexities surrounding gender and vote choice among union members.
Also noteworthy is the fact that the NDP appears to be the primary beneficiary of women union members’ reluctance to support the Conservative Party. This raises questions about the electoral choices women in unions make compared to women in the general electorate. While women voters broadly tend to favour both the Liberals and the NDP (Fournier, Reference Fournier2020), our data reveal that women in unions appear predominantly inclined to support the NDP. We cannot, however, offer a straightforward explanation for this disparity, underscoring the need for further research into how union membership influences women’s vote choice.
Our findings with respect to age match our expectations in H2c. Younger union members are more likely to support the NDP, while older union members are more likely to support the Conservatives and the Liberals. A union member one age category older than another is 9 percentage points less likely to support the NDP, about 7 percentage points more likely to support the Liberals, and about 3 percentage points more likely to support the Conservatives. This means that we expect that a union member between the ages of 35 and 44 would be 9 percentage more likely to support the NDP, 7 percentage points less likely to support the Liberals, and 3 percentage points less likely to support the Conservatives than a union member between the ages of 45 and 54. This indicates that age-related preferences among union members align with broader trends observed.
Consistent with H2d, we find that income matters to the vote choice of union members, though the effects of income are not large. We estimate that moving up an income category correlates with a 2 percentage point decrease in the likelihood that a respondent supports the NDP and about a 1 percentage point increase in the likelihood of voting Conservative. Income does not have a statistically significant effect on the likelihood of a union member supporting the Liberals.
Without overstating the importance of relatively small effects, our findings suggest a stratification of class interests within unions. In short, lower-income union members would stand to benefit more from the welfare state investments promoted by social democratic parties, thus drawing them closer to a party like the NDP, while higher-income union members may be more interested in the tax cuts routinely promised by conservative parties, thus increasing the likelihood of a vote for the Conservatives.
Finally, and in line with H2e, we find that education matters to vote choice. As compared to respondents with no more than a high school education, the Conservatives do better with union members that have a college or similar education and worse with voters with a university degree. Union members with a college qualification are about 9 percentage points more likely to vote Conservative than union members that have no more than a high school education, while union members that have a university degree are 18 percentage points less likely to support the Conservatives than union members with no more than a high school education. Not shown in Figure 4 (to make the figure easier to read) is the finding that union members with a university education are much less likely to support the Conservatives than union members with a college or similar qualification (27 percentage points less likely to support the Conservatives).
This finding appears to reinforce our earlier discussion of disproportionate Conservative support among members of construction unions, who are more likely to hold trade qualifications as opposed to university degrees. This suggests that the type of union job and the level of education associated with it matters a great deal to the Conservatives’ appeal to unionized voters.
The relationship between education and union member vote choice is nuanced for both the Liberals and the NDP. Figure 4 indicates that there is no statistically significant difference in the likelihood of union members with university educations voting for the Liberals or NDP compared to those with only high school educations, although the estimates suggest positive relationships. However, there are statistically significant differences between the likelihood of union members with university degrees voting for either party compared to those with college or similar qualifications. Union members with university degrees are 12 percentage points more likely to vote Liberal and 15 percentage points more likely to vote NDP than those with college or similar qualifications. This suggests that, while university educated union members prefer the Liberals or the NDP, they do not have a clear preference when choosing between the two parties, reflecting a developing trend in the broader electorate.
Overall, our analysis of demographics tells a complex story. In some ways, with respect to age, gender, income, and education, demographics shape union members’ support for parties in ways that are similar to the rest of the population. However, region shapes union members’ partisan support in ways that are distinct from Canadians as whole. The Conservatives do not appear to have the same kind of disproportionate support among union members from Western Canada that they appear to have among the Canadian population as a whole. Strikingly, education stands out as the demographic variable that has the largest impact on union members’ support for parties. The difference in support for the Conservatives among union members with college or similar qualifications compared to the support for the Liberals and the NDP among union members with university degrees is substantial. The 27-percentage point increase in the likelihood of college-educated union members voting Conservative, as opposed to university-educated union members, is the most significant effect among all variables in our analysis, including those related to union type. These findings are indicative of an ability on the part of the Conservatives to appeal to particular segments of organized labour and suggests a continued need for researchers to treat union members as a politically diverse group.
Conclusion
Acknowledging the complex and multifaceted nature of labour politics is key to understanding how union membership can influence electoral behaviour and outcomes. Our findings demonstrate that while union membership, in general, makes voters more likely to vote NDP, various intersecting factors such as union type, gender, education, age, and income introduce important vote choice nuances that suggest less political uniformity among union members than is often assumed. The NDP shows a clear but small level of disproportionate support among lower income union members, younger union members, and members of NDP-affiliated unions. In contrast, Conservatives enjoy disproportionate support among members of construction unions and among college-educated union members. Both the Liberals and the NDP enjoy disproportionate support among university-educated union members, at least compared to those with college and similar educational qualifications. In short, the voting intentions of union members in Canada are multifaceted and open clear strategic possibilities for parties competing for votes among different segments of organized labour.
The NDP continues to enjoy significant levels of support among NDP affiliated union members. However, the party certainly cannot take these votes for granted. Although some union and NDP activists dismissed Pierre Poilievre’s overtures to blue-collar voters as “cosplaying a working-class guy, (Burke et al., Reference Burke, McKenna and Ryan2024)” growing support for Conservatives among particular union and working-class voters is real (Polacko et al. Reference Polacko, Kiss and Graefe2022). Elevated levels of support for the Conservatives among members of construction unions and college-educated union members suggests a need to further examine the relationship between the Conservatives and these groups of voters which are becoming a key electoral constituency for the party.
Finally, it is worth acknowledging some of the limitations of this research. While the survey was conducted in three waves over six months, it does capture a time horizon that featured comparatively elevated support for the Conservatives and depressed support for the Liberals among voters more generally. Studies of union members’ political support that reach across longer time horizons may yield different results given changing circumstances and political dynamics. However, embedded within the broader literature, some of which stretches back decades, we are confident that our findings reveal important insights about continuity and change in the vote choice of unions members in Canada.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0008423925100449
Acknowledgements
The authors are grateful to the anonymous reviewers for their comments on the article.
Competing interests
The authors declare that they have no conflicts of interests related to this publication.
Open access
