Abstract
Union provision of collective bargaining and political lobbying generates positive outcomes across society. A union default holds significant promise to revive the flagging fortunes of unions by enabling them to recruit many more members, extend their bargaining coverage, and place them in a stronger position to deliver these positive outcomes. Non-unionism is the default setting in employment arrangements. A union default would reverse this, so employees are defaulted into membership in the first instance, but with a right to opt-out. In this article the authors test whether framing unions as providers of public goods increases worker support for a union default and intention to stay in a union, if defaulted. The authors find that workers are significantly more likely to support a union default and stay in union membership in scenarios involving positive framing for unions.
Introduction
Unions provide services such as collective bargaining and political lobbying that generate positive societal outcomes. First, research from various countries indicates unions consistently raise wages (Bryson, 2014; Farber et al., 2018), and by as much as 30–40% for low-waged workers (Card, 2001: 305), and their impact on fringe benefits is even larger (Kristal et al., 2020). Second, studies show unions reduce income inequality by compressing pay within and between organizations (Card et al., 2020; Huang et al., 2017; Metcalf et al., 2001) and by successfully lobbying governments for social programmes primarily benefiting the poor (Ahlquist, 2017; VanHeuvelen and Brady, 2022). Third, research confirms unions play an effective role in monitoring and enforcing employer compliance with laws governing, for example, health and safety, minimum wages, holidays and discrimination (Morantz, 2017). Fourth, recent international evidence from large samples also suggests unions improve the well-being of workers through, for example, their efforts to regulate workloads, working hours, overtime and job security (Blanchflower and Bryson, 2020). Longitudinal research also confirms that, when workers move from non-union to unionized jobs (or the reverse), their sense of well-being generally rises (falls) (Bryson and White, 2016). These positive outcomes of unions are public goods, insomuch as non-members and others in society share in the benefits without usually paying union fees (Freeman and Medoff, 1984; Madland, 2021).
A union default holds significant promise as a new employment relations policy that could revive the flagging fortunes of unions by enabling them to recruit many more members, extend their bargaining coverage at enterprise and sectoral levels, and place them in a stronger position to deliver the positive outcomes identified above, including at the national level of political exchange (Harcourt et al., 2019). In Anglophone countries in particular, where bargaining coverage has little statutory basis, the extension of collective bargaining could facilitate creation of sectoral bargaining as well as a return of some form of corporatism.
Worldwide, non-unionism is currently the default setting in employment arrangements. A union default would reverse this so that employees are defaulted to union membership in the first instance, but with a right to opt-out. It, thus, protects and advances both the freedom to associate and not associate (Harcourt et al., 2019, 2021). This makes the union default very different from the closed (union) shop, where workers are compelled to become and remain union members prior to or shortly after commencing employment. With a default, by contrast, continued union membership always remains an individual worker choice.
Prior research explored how a union default would work in practice, and could be adapted to suit different employment relations regimes, including those in Britain and the US (Harcourt et al., 2021). New Zealand (NZ) studies have shown a default would attract majority support and prompt nearly two-thirds of workers to remain in unions, after being defaulted (Harcourt et al., 2020, 2021). Notwithstanding some differences in density rates and some variations in institutional architecture, the union default is likely, we believe, to be comparably well received in other Anglophone countries, where preferences for union representation are similar to those in NZ (Freeman et al., 2007). Nevertheless, with nearly a third of workers still saying they would opt-out, unions in some industries and occupations would still struggle to establish a critical membership mass in order to effectively execute their public goods functions. We contend that, as long as workers conceive the choice to join/remain (or not join/remain) in a union in narrow, individualistic terms, involving only costs and benefits to themselves, they are more likely to opt-out and free-ride on the benefits of union representation. However, if workers are instead encouraged to focus upon their broader work interests, in terms of achieving the four aforementioned public goods outcomes, they are more likely to support the union default and remain union members.
New Zealand was chosen as the setting for this study for a number of reasons. First, the NZ Labour government is legislatively re-collectivizing employment relations through industry-wide collective agreements, so-called ‘fair pay agreements’, to be trialled first in low-wage sectors such as supermarkets and private security (Kent, 2021). From 1894 to 1991, the state established a centralized system of pay determination via occupational and industry awards, initially supported by arbitration. That system was entirely dismantled by the Employment Contracts Act 1991, precipitating a virtual collapse of bargaining in the private sector and a fall in density from almost 50% to 20% (Charlwood and Haynes, 2008). Further decline was arrested, though never reversed, by the Employment Relations Act 2000. Second, and relatedly, Labour has demonstrated willingness to consider the default as a complement to these efforts to revive union representation. 1 Third, the NZ Council of Trade Unions, which has a close relationship to Labour, has also expressed enthusiasm for the default. Fourth, Labour faces no major barriers to implementing employment relations reforms: NZ is a unicameral, unitary state with no written constitution, and Labour currently holds a majority of parliamentary seats. Thus, proposed laws cannot be rejected by an upper house within the legislature (e.g. a Senate filibuster); the laws apply to all New Zealanders and not just a federal jurisdiction; and they cannot be challenged constitutionally. Fifth, NZ evidence suggests that more than two-thirds of adults have favourable views of unions (New Zealand Election Survey, 2014, 2017), and half the workforce desires union representation (Freeman et al., 2007). No doubt, resurging interest in unions reflects the poor wages and working conditions prevailing under individualized employment relations. Real wage increases have trailed far behind productivity growth, and labour’s share of income fell precipitously after 1991 before partially recovering in the early 2000s (Rosenberg, 2016). Declining union representation in the 1990s was mainly responsible for dramatically widening income inequality (Charlwood and Haynes, 2008) and transferring a greater share of income to the top 10% than in any other developed country (Jaumotte and Buitron, 2015).
We test whether framing unions as providers of public goods, the aforementioned four societal outcomes, increases worker support for a union default and intention to stay in a union, if defaulted. 2 We find workers are significantly more likely, with some provisos, to support a default and stay in union membership (e.g. not opt-out) in scenarios involving this sort of positive framing for unions. We argue that, when attention is drawn to superordinate interests via such framing, workers are more likely to socially identify with unions and union members and are, thus, more likely to see it as appropriate to support the default and stay in unions.
We begin by examining the nature of unions as providers of public goods. From here, the influence of framing on how individuals perceive public goods is considered, and then applied to the subjects of unions and a union default system. Next, we present our fieldwork in terms of its methods and results before discussing them and drawing out implications for wider generalizability.
Unions, framing and public goods
Unions as providers of public goods
The quartet of aforementioned outcomes are largely public goods, being both non-rivalrous and non-exclusive (Olson, 1965; Pulsipher, 1966). They may be non-exclusive to non-members for several reasons. First, the natural properties of a particular benefit may make it almost impossible to deny non-members access to it. For instance, collectively bargained improvements in health and safety cannot be restricted to union members when union and non-union staff work alongside each other (Freeman and Medoff, 1984). Second, employers often voluntarily extend whatever a union has negotiated to non-union staff in the same organization or in competitor organizations, either to avoid unionization, maintain simplicity and uniformity in rules and procedures, or preserve a sense of fairness necessary for cooperative working relations among staff (Barnes, 2005). In NZ, employers commonly offer individual agreements to non-union staff that are virtually the same as collective agreements with unionized staff, and ‘pass on’ any union-negotiated pay increases to non-union staff doing the same work (Barnes, 2005). In the US, non-union firms have long raised and flattened pay rates to somewhat mirror those in nearby unionized firms, particularly in heavily unionized regions (Denice and Rosenfeld, 2018; Western and Rosenfeld, 2011). Third, employment relations laws can compel extension of terms and conditions in a collective agreement to non-union workers in the same or other organizations. In the US and Canada, for example, exclusive representation requires the extension of a collective agreement to all non-union workers in the union’s bargaining unit (Pulsipher, 1966).
The four outcomes are also almost always non-rivalrous, for non-member enjoyment does not detract from member enjoyment of benefits. Indeed, member and non-member enjoyment are often complementary. For instance, if non-union firms match the pay of unionized competitors, it reduces downward pressure upon union wages. Wages rates, removed from competition, are easier to sustain over time.
The four outcomes are also public goods because individual workers acting alone have little incentive to bargain for a change in terms and conditions if aggregate benefits flow mostly to their colleagues. Moreover, in the absence of union action, an employer faces little pressure to address or accommodate one worker’s complaint. Even if open to such an accommodation, an employer might still resist making concessions, not knowing whether these would please other workers in the same circumstances. Much the same would apply to individual lobbying of governments for new legislation. Lastly, individual workers are unlikely to challenge controversial management actions for fear of provoking retribution (Freeman and Medoff, 1984).
Free-riding versus cooperating
In the absence of compulsion, individuals may freely choose to either cooperate with others by contributing to the production of a public good or free-ride by not contributing. Whether they contribute or not, they still receive the benefits public goods provide (Olson, 1965). Much the same applies to unions. While unions seek to address this issue by providing benefits (like insurance discounts) only to members, this does not fundamentally alter the public goods character of most union benefits.
Standard neo-classical economic theory predicts rational, utility-maximizing individuals will always free-ride in public goods situations (Becker, 1974; Kolstad, 2011; Olson, 1965; Sugden, 1993). Widespread free-riding creates a collective action problem, where failure to cooperate in providing financial support leads to under-production of the public good (Olson, 1965). Thus, free-riding poses a serious threat to establishing and maintaining union representation, even though collective benefits greatly exceed collective costs of membership dues (Freeman and Medoff, 1984; Pulsipher, 1966). However, field studies (e.g. Gerkey, 2013) and laboratory experiments (Bernold et al., 2015; Brewer and Kramer, 1985; Kramer and Brewer, 1984) suggest more nuanced, less individualistic views of motivations and behaviour. Some individuals invariably free-ride, but most do not. Individuals can place group welfare above their own immediate self-interest in public goods situations. Some are unconditional cooperators, consistently contributing to providing public goods. Most are conditional cooperators, choosing to contribute in some circumstances but not others (Fischbacher and Gachter, 2010; Fischbacher et al., 2001; Kolstad, 2011).
Social identity theory, framing and cooperation
Humans are social beings with both private (self) and social (group) preferences which they often subconsciously and habitually prioritize in different ways at different times (Sugden, 1993). Social groups are important to people because they fulfil various needs, including for belongingness and distinctiveness, as well as providing access to power and resources (Van Bavel and Pereira, 2018). Social identity theory posits people are driven by identities as both individuals and members of overlapping social groups (Tajfel and Turner, 1986; Van Bavel and Pereira, 2018). Moreover, ‘the relevance of any given identity varies with social context. When a particular membership is made salient . . . people are more likely to behave and experience emotions in ways that are congruent with the activated social identity’ (Van Bavel and Pereira, 2018: 214–215).
The salience of a particular social identity and associated social preferences to some extent depend upon social context framing (Brewer and Kramer, 1985; Gerlach and Jaeger, 2016; Gold and Sugden, 2007; Kramer and Brewer, 1984). Framing is the process by which individuals ‘select some aspects of a perceived reality and make them more salient in a communication context, in such a way as to promote a particular problem definition, causal interpretation, moral evaluation, and/or treatment recommendation for the item described’ (Entman, 1993: 53). Framing differences have been empirically linked to variation in levels of cooperation, involving the same public good (Brewer and Kramer, 1985; Kramer and Brewer, 1984). For instance, cooperation across participants is less prevalent in social dilemma experiments such as the prisoner’s dilemma or the Wall Street game rather than a community or environment game (Bernold et al., 2015). A frame can draw attention to superordinate goals and interests that unite people through a broader, more inclusive, social identity, prompting their cooperation in mutually contributing to public goods (Gerlach and Jaeger, 2016; Van Bavel and Pereira, 2018). For instance, social identity framing increased various forms of support for green energy in a US study of university students (Segrian, 2014). Contrastingly, a frame can highlight goal or worldview differences that inflame parochial divisions, prompting individuals to withhold support for a public good (Gerlach and Jaeger, 2016; Van Bavel and Pereira, 2018). Correspondingly, studies show greater public support for various defaults if these are believed to be in the public interest or of benefit to many or most people (Junghans et al., 2015; Reisch and Sunstein, 2016). Furthermore, if default setters (e.g. policy-makers) are believed to be motivated by a desire to improve the lives of the many rather than the few, people are more likely to support a particular default (Bang et al., 2018; Junghans et al., 2015; Osman et al., 2018; Reisch and Sunstein, 2016; Sunstein, 2016a; Sunstein et al., 2018).
So, which social identity is salient at any given time affects which preferences are prioritized and, thus, whether people contribute to public goods. To further elucidate, with activation of a more exclusive social identity, an individual might perceive some beneficiaries of the same public good as belonging to an outgroup and, thus, less deserving of the public good. In this context, self-interest would be more likely prioritized over group interest in a decision to contribute to the public good. More free-riding would be expected. Contrastingly, with activation of a more inclusive social identity, an individual might perceive some beneficiaries of the same public good as belonging to the individual’s ingroup, and as therefore more deserving of the public good. 3 In this context, group interest would be more likely prioritized over self-interest. Less free-riding would be expected (Gerlach and Jaeger, 2016).
Framing, social identity and unions
Media can negatively or positively frame unions, encouraging or discouraging identification with them (Brimeyer et al., 2016). In negative frames, unions are ‘vested interests’ (Flanders, 1970), only motivated to advance members’ and leaders’ welfare (Brimeyer et al., 2016). For example, members belong to privileged groups of predominantly highly-paid, older, white, blue-collar men (Gibney et al., 2018). Worse still, union leaders are corrupt, powerful and undemocratic (Puette, 1992). Collective bargaining monopolizes the labour market, coerces unjustified pay increases, and imposes adverse consequences on employers (e.g. reduced competitiveness), consumers (e.g. price hikes), and other workers (e.g. unemployment) (Brimeyer et al., 2016; Kane and Newman, 2017).
In positive frames, unions are ‘swords of justice’ (Flanders, 1970), defending and advancing interests of all workers, especially the low-waged and/or mistreated (Brimeyer et al., 2016). Members and leaders are ordinary, albeit fallible, people, with the latter answerable to, and elected by, the former. Collective bargaining is just one tool among several used to pressure employers and governments into making concessions that improve all workers’ and their families’ lots. Positive frames are largely consistent with unions’ public goods character, even though they are less common in the media, at least in the US (Brimeyer et al., 2016; Fiorito and Padavic, 2022; Kane and Newman, 2017).
Frames influence the extent to which individuals identify with unions. One study examined reactions to three experimental scenarios, 4 where unions were framed neutrally, negatively but reasonably allowing for genuinely negative effects, or very negatively and unreasonably using class-based anti-union rhetoric (CAR) typical of Fox News (Kane and Newman, 2017). Exposure to the very negative CAR frame diminished perceived similarity to union members, perceived deservingness of union members, and support for laws to enable collective bargaining and strikes. Kane and Newman (2017: 24) concluded ‘the broad political alignment of socio-economic groups can potentially be undone when the members of one group are portrayed as socio-economically dissimilar to the members of the other groups’. Though not focused on framing effects, other studies (e.g. Fiorito and Padavic, 2022; McGrane and Berdahl, 2015) also found ‘public attitudes towards unions are more positive when the labour movement is perceived to be working for the betterment of all society and more negative when unions are perceived to be pursuing the sectional interests of their own members’ (McGrane and Berdahl, 2015: 62). More (less) favourable views of unions are also positively (negatively) associated with support for collective bargaining and intention to vote for a union in representation elections (Fiorito and Padavic, 2022). Whether such positive or negative framing occurs often depends upon the effectiveness of employers in promulgating their own discourses internally and externally. Internally, employers have the right to frame messages to their employees via their own organizational communication channels. Externally, they are usually well-resourced and networked, and the media are usually receptive to business discourses.
Positive framing and the union default
Our study tests whether positive framing increases worker support for a union default and intention to stay in a union when defaulted. In our experimental scenarios, the union default is framed positively by stressing four public goods aims: increasing pay, decreasing inequality, increasing legal compliance and enhancing worker well-being. Drawing attention to such superordinate aims should encourage more workers to socially identify with unions (Gerlach and Jaeger, 2016; Van Bavel and Pereira, 2018). Positive framing is, therefore, expected to increase union popularity and workers’ willingness to remain members, consistent with both union framing (Kane and Newman, 2017) and default popularity (Junghans et al., 2015; Reisch and Sunstein, 2016) research. However, informing individuals a default is effective may not be sufficiently persuasive; some may want to know whether there is evidence of such effectiveness before deciding to support the default and/or stay in union membership. Thus, our study also tests, in another experimental scenario, whether alluding to evidence of unions’ four public goods effects has an even larger effect on worker support for the default and propensity to stay in membership. Past default studies already suggest supplying more information about default effectiveness (e.g. evidence) helps to increase public support (Davidai and Shafir, 2018; Sunstein, 2016b).
We contend the impact of positive framing is likely to depend upon political party affiliation. The four public goods aims of unions are likely less appealing to those on the political right than left. For instance, research suggests right-wing voters are more tolerant and accepting of income inequality (Grunhage and Reuter, 2020). They may also be more concerned about business profits than wages, too much regulation than too little compliance, and see well-being as a private matter unamenable to public policy influence. Thus, any attempt to highlight that unions can achieve the four public goods aims, even if believed, is less likely to be valued.
Even without a framing intervention, a US study shows party affiliation has a major effect upon default preferences (Tannenbaum et al., 2017). Democrats were more likely to prefer defaults involving ‘food stamps’ for the poor and a ‘safe sex’ educational programme for school children, whereas Republicans were more likely to prefer defaults involving ‘tax breaks’ for capital gains income and an ‘intelligent design’ programme for school children. Since unions are normally aligned to left-of-centre parties (e.g. Democrats, Labour, Social Democrats) and advocate left-wing policies (e.g. social programmes, progressive taxation, more regulation), we would expect right-wing voters to be less likely both to endorse a default and remain in membership. Indeed, an NZ study confirmed National voters (conservatives) were less likely to approve of the default idea than non-voters and other voters, including Labour voters (Harcourt et al., 2020). Evidence from public goods and trust games also reveals right-wing voters have less pro-social views. They are less trusting of others, more suspicious of ideas about ‘the common interest’, more likely to see life as a ‘competitive jungle’ and, therefore, less likely to contribute to public goods (Grunhage and Reuter, 2020). As a result, even in the absence of any framing, we would expect right-wing (left-wing) voters to be less (more) likely to both support the default and stay in a union, if defaulted. We would also expect positive framing to have weaker (stronger), positive effects on right-wing (left-wing) intentions to support the default and stay in a union.
Methods
Data collection
Data were collected through an online survey, involving a random sample of NZ-domiciled adult employees via the Qualtrics Company. Qualtrics tries to ensure its samples are representative of larger populations via traditional, actively managed, double-opt-in market research panels recruited through a network of panel routers or aggregators. Nevertheless, females (52.6%), union members (28.5%), public sector workers (30.3%) and Labour Party voters (50.6) were over-sampled, while younger workers and non-voters (11%) were under-sampled. 5 Potential respondents were randomly invited to participate via email, text, in-app notification, or by signing in to the response portal, which also outlined the expected time to complete the survey and its title. Anyone accepting the invitation was guided to material on participant information and then informed consent. The former required acknowledging reading the document and the latter accepting before beginning the survey. At the start of the survey, respondents were asked screening questions on age, employment status and residency status to check their fit with the participation criteria (e.g. 18+, employee, legally living in NZ). Any answers indicating respondents were younger than 18, not employees, or not NZ citizens, permanent residents or current visa holders automatically resulted in survey termination. Invitations remained active until 1047 qualifying participants’ completed responses were collected. Respondents were paid by Qualtrics following survey completion.
Procedure
Respondents were randomly allocated to one of three different groups: one control group and two experimental groups. The control group comprised 610 respondents. The first experimental group comprised 220, all of whom were exposed to Scenario #1; the second experimental group comprised 217, all of whom were exposed to Scenario #2 (see below).
Variables
The variables were derived from the answers to survey questions. All survey questions required respondents to select an answer prior to proceeding to the next question. 6 There are two dichotomous dependent variables. For the first, respondents were asked to imagine a situation where they had just been employed and auto-enrolled (defaulted) into membership as a requirement of a new law. Respondents were then asked whether they would ‘stay in’ the union, coded ‘1’, or ‘opt-out’, coded ‘0’. For the second, respondents were asked to indicate their view of auto-enrolment in unions, where the options were ‘I support automatic enrolment in unions’, coded ‘1’, or ‘I oppose automatic enrolment in unions’, coded ‘0’.
There are 14 independent variables. Two are of primary research interest, with each indicating one of the two experimental scenarios. Each scenario, when present, was described right after respondents were asked to imagine they had just been hired and auto-enrolled in a union. For the first scenario, coded ‘1’ for the variable ‘Aims with evidence’, respondents were asked to ‘imagine that a law requiring automatic enrolment in a union would have four main aims for NZ:
Reduce income inequality across society;
Raise wages, especially for low-waged workers;
Ensure employer compliance with employment laws in NZ;
Improve employee well-being and work–life balance.’
Then, before they were asked whether they would ‘stay in’/‘opt-out’ of membership, respondents were also supplied partial citations, relating to research demonstrating unions have been important to achieving the above aims. 7 For the second scenario, coded ‘1’ for the variable ‘Aims no evidence’, respondents were provided word-for-word the same information about the four aims but supplied no information about research linking unions to the achievement of such goals. For both scenario variables, a ‘0’ coding indicated respondents were part of the control group and, therefore, not exposed to any information about aims or associated evidence prior to indicating they would ‘stay in’/‘opt-out’ of membership.
Two additional independent variables are used to capture respondents’ voting behaviour. ‘National Party’ was coded ‘1’ if respondents indicated they voted for the National Party in the 2020 NZ general election. National is a conservative, pro-business party with some antipathy towards unions. ‘Labour Party’ was coded ‘1’ if respondents indicated they voted for the governing Labour Party in this election. Labour has traditionally been social democratic and still maintains ties to unions. Both of these variables were coded ‘0’ for all other responses (e.g. voted for a minor party or did not vote). Thus, National voters were expected to be less likely to support a default and stay in membership if auto-enrolled, when compared to minor party voters and non-voters. The opposite was expected for Labour voters.
Four interaction variables were also included to assess whether effects of the two scenarios were moderated by whether the respondent was a National or Labour voter. ‘Labour × 4 aims with evidence’ was coded ‘1’ for all Labour voters exposed to the first scenario and ‘0’ otherwise. ‘Labour × 4 aims no evidence’ was coded ‘1’ for all Labour voters exposed to the second scenario and ‘0’ otherwise. ‘National × 4 aims with evidence’ was coded ‘1’ for National voters exposed to the first scenario and ‘0’ otherwise. ‘National × 4 aims no evidence’ was coded ‘1’ for all National voters exposed to the second scenario and ‘0’ otherwise.
There are also 10 other independent variables included to control for other factors that might influence views of a default. ‘Union member’ was coded ‘1’ if respondents answered ‘yes’ to the question ‘Are you a member of a union at your workplace?’ and coded ‘0’ if they answered ‘no’. Union members, perhaps more than any other group, were expected to be more likely to favour a default as the option most aligned with their own preferences and to stay in the union if automatically enrolled.
The variable ‘Manager’ was coded ‘1’ if respondents selected ‘Manager’ as the ‘occupational category that best describes your main job as an employee’, and coded ‘0’ if respondents chose one of the other nine occupational categories in the Australia New Zealand Standard Classification of Occupations. 8 Managers, as employer representatives, were expected to be less likely to support the default and stay in if automatically enrolled, as compared to other occupational groups. ‘Income’ is an ordinal variable measuring respondents’ individual income from all sources, including, for example, wages, salaries, tips, interest, dividends, scholarships, government benefits and child support. Respondents were asked to select the income category best matching their total income. The categories start with ‘loss’ (coded ‘1’), rise in increments of $10,000 for the first $100,000 of income (e.g. ‘$0–$10,000’, coded ‘2’), and rise in increments of $20,000 for the second $100,000 (e.g. ‘$100,001–$120,000’ [coded ‘12’]). Thereafter, total incomes of $200,001–$300,000 were coded ‘17’, and finally total incomes above $300,001 were coded ‘18’. ‘Age’ is an interval variable measuring respondents’ reported age in years. It was coded ‘1’ for ‘under 18’, increasing one unit for every year older than 17, reaching a maximum of ‘74’ for ages 90 and above. ‘NZ European’ was coded ‘1’ if respondents indicated the ethnic group best describing themselves included NZ European, and ‘0’ where ‘NZ European’ was not chosen. 9 ‘Staff less than 20’ was coded ‘1’ if respondents indicated fewer than 20 staff presently worked for their main employer, and ‘0’ if they indicated 20 or more staff presently worked for their main employer. 10 ‘Regular part-time’ was coded ‘1’ if respondents indicated they worked part-time in their main job, defined in NZ and on the survey as fewer than 30 hours per week. ‘Regular full-time’ was coded ‘1’ if respondents indicated they worked full-time in their main job, defined in NZ and on the survey as 30 or more hours per week. Both variables were coded ‘0’ if respondents indicated employment in their main job was either ‘casual’, ‘fixed-term’ or ‘triangular/agency’ instead. ‘Public sector’ was coded ‘1’ if respondents indicated they were primarily employed in the public sector or government, and ‘0’ if they indicated they were primarily employed instead by either the private or non-profit/charity sectors. Finally, ‘Female’ was coded ‘1’ if respondents indicated their gender was female, and ‘0’ otherwise (e.g. male, other, or did not answer the question).
Data collection involved just one method, a survey. There was consequently some risk of common method bias. To assess this bias, we ran a Harman single factor test, using an exploratory factor analysis with a maximum likelihood extraction. We found that a single factor explains only 19.482% of the variance in the independent variables. A number below 50% suggests variable values are not biased in a particular direction because only the one method was used to collect the data.
In the survey, we guarded against common method bias in various ways. First, the dependent variable questions were spaced far apart and given no particular importance. Second, the independent variable questions were placed in no particular order and also interspersed throughout the survey (e.g. questions about age and employment status were near the start; questions about voting and gender near the end). Third, efforts were made to ensure that no particular answer to any given question was ever more socially desirable than another; answers to questions were also randomized. Fourth, there was some time delay between each question. Fifth, respondents were given no hint that the survey involved experiments regarding framing, and that they might be assigned to a control group or one of two experimental groups.
Data analysis
This study used SPSS binomial logit to estimate three maximum likelihood models for each of the two dependent variables: ‘Staying in the union’ and ‘Support for a union default’. Each binomial logit model took the following form:
With Models 1, 2 and 3, ‘event’ meant respondents would stay in the union (coded ‘1’), whereas ‘no event’ (coded ‘0’) meant respondents would opt-out, if defaulted into membership. With Models 1a, 2a and 3a, ‘event’ meant respondents supported a union default (coded ‘1’), whereas ‘no event’ meant respondents opposed it (coded ‘0’). The probability of ‘event’ divided by the probability of ‘no event’ denoted an odds ratio. In the case of Models 1, 2 and 3, this ratio was the odds of staying in the union versus opting out. In the case of Models 1a, 2a and 3a, this ratio was the odds of supporting versus opposing a union default.
The log of the odds ratio for all six models was predicted by the independent variables. The parameter estimates, or βs, in each of the six models showed the expected change in the log odds for each independent variable. Models 1 and 1a respectively estimated the expected change in the odds of staying in versus opting out and supporting versus opposing the default on the basis of the two scenarios, as indicated by ‘Aims with evidence’ and ‘Aims no evidence’. Models 2 and 2a respectively did the same, but included all 14 independent variables. Models 3 and 3a respectively did the same, but included the 14 independent and four interaction variables.
Results
Table 1 shows the percentage staying in and supporting the default for the control group only (e.g. in the absence of any effects from the experimental conditions). It shows that most respondents (65.7%) would stay in membership if defaulted and also supported (56.1%) a default. Table 1 also indicates percentages staying in membership and supporting a default for each subgroup associated with the binary variables. At this stage prior to the logistical regression analyses, several observations are noteworthy. First, a majority of individuals in all groups expressed an intention to stay in membership if defaulted. This is even true of managers (55.0%) and National voters (53.5%). Second, a majority (or large minority) of all groups also supported the default. Third, and in particular, 57.3% of non-union employees indicated they would remain members if defaulted and 50.1% supported the default.
Percentages of control group and subgroups: Staying in the union and Support for the union default (except for non-binary variables).
The ‘0 value’ for ‘male’ refers mainly to males but also includes individuals of other genders and respondents who did not disclose their gender identity.
Tables 2 to 5 outline the findings for the models. Model 1 in Table 2 and Models 2 and 3 in Table 4 were used to predict the odds of staying in membership versus opting out, the first dependent variable. Models 1a in Table 3 and Models 2a and 3a in Table 5 were used to predict the odds of supporting versus opposing the default, the second dependent variable. Models 1 and 1a were simpler models, where only the two experimental scenarios, ‘Aims with evidence’ and ‘Aims no evidence’, were used to predict the two dependent variables. Models 2 and 2a used the same scenario variables plus the remaining 12 independent variables to respectively predict the first and second dependent variables. Models 3 and 3a used the same scenario variables, the same 12 independent variables and the four interaction variables to respectively predict the first and second dependent variables. The parameter estimates for each independent variable are provided in columns, labelled for each model in Tables 2 to 5, with the standard error in parentheses. The adjacent column to the right lists the odds ratio change associated with each parameter estimate. For Models 1, 2 and 3, positive (negative), statistically significant parameter estimates suggest an increase (decrease) in the odds of ‘staying in the union’ as opposed to ‘opting out’. Likewise, for Models 1a, 2a and 3a, positive (negative), statistically significant parameter estimates suggest an increase (decrease) in the odds of ‘support for a default’ versus ‘opposition to a default’. These changes in odds are in relation to the odds implied by the constant parameter estimate for each model.
Predicting staying in the union: Parameter estimates (standard errors).
Predicting support for the union default: Parameter estimates (standard errors).
Predicting staying in the unions: Coefficient estimates (standard errors).
Predicting support for the union default: Coefficient estimates (standard errors).
Model 1 in Table 2
Table 2 shows the chances of respondents saying they would maintain membership, if defaulted, are higher in both experimental scenarios. The parameter estimate for the constant in Model 1 is positive and statistically significant. It suggests, for the control group, that the odds of ‘staying in’ are nearly double those of ‘opting out’ (1.919:1), equivalent to a 65.7% probability of ‘staying in’ (i.e. 1.919/2.919). The parameter estimates for ‘Aims with evidence’ and ‘Aims no evidence’ are also positive and statistically significant. In the ‘Aims with evidence’ scenario, the odds of ‘staying in’ increase from just under 2:1 to just over 5:2 (i.e. 1.919 × 1.359 = 2.607:1), equivalent to a rise in probability from 65.7% to 72.3% (i.e. 2.607/3.607). In the ‘Aims no evidence’ scenario, the odds of ‘staying in’ increase much more from just under 2:1 to nearly 5:1 (i.e. 1.919 × 2.455 = 4.711:1), equivalent to a rise in probability from 65.7% to 82.5% (i.e. 4.711/5.711).
Model 1a in Table 3
Table 3 shows the chances of respondents saying they would support rather than oppose a default are only higher in the scenario where the four aims are articulated but no evidence is given. The parameter estimate for the constant in Model 1a is positive and statistically significant. It suggests, for the control group, that the odds of supporting rather than opposing a default are roughly 5:4 (i.e. 1.276:1), equivalent to a 56.1% probability of ‘support’ (i.e. 1.276/2.276). The parameter estimate for ‘Aims with evidence’ is not statistically significant, suggesting no impact of this scenario on the propensity to support a default. In contrast, the positive, statistically significant parameter estimate for ‘Aims no evidence’ indicates that the second scenario increases the odds of ‘support’ for a default from approximately 5:4 to almost 2:1 (1.276 × 1.546 = 1.972:1), equivalent to an increase in probability from 56.1% to 66.4% (i.e. 1.972/2.972).
Models 2 and 3 in Table 4
Table 4 outlines the findings for Models 2 and 3, predicting ‘staying in’ membership if defaulted. Results for Model 2 show that, once various factors are controlled for, the parameter estimate for ‘Aims with evidence’ is no longer statistically significant. Favourable variation in the added independent variables could, therefore, explain the higher propensity to ‘stay in’ in the ‘Aims with evidence’ group. However, even with controls included, the parameter estimate for ‘Aims no evidence’ remains highly statistically significant, large and positive, with an implied odds ratio increase for ‘staying in’ versus ‘opting out’ of 2.604 with all 14 independent variables versus 2.455 with only the two scenario variables.
The parameter estimates are positive and statistically significant for ‘Union member’ and ‘Female’, but negative and statistically significant for ‘Manager’, ‘Income’, ‘Age’, ‘NZ European’ and ‘Staff less than 20’. Thus, union members and women were more likely to say they would ‘stay in’ membership if defaulted than other groups (e.g. non-union employees, men). Conversely, National Party voters, managers, higher income earners, full-time workers, older individuals, NZ Europeans and those employed by organizations with fewer than 20 staff were less likely to say they would ‘stay in’ membership if defaulted than other groups (e.g. minority party voters and non-voters, non-managers, low-income earners, non-regular employees, younger employees, non-Europeans and other Europeans, employees working for an employer with 20 or more staff).
Model 3 includes the same independent variables as Model 2 but with the interaction variables added. 11 Parameter estimates and statistical significance levels for the independent variables are similar to those in Model 2. First, the parameter estimate for ‘Aims with evidence’ is still statistically insignificant, as in Model 2, but the parameter estimate for the interaction variable ‘Labour × aims with evidence’ is statistically significant and positive. The parameter estimate for the interaction variable ‘National × aims with evidence’ is, however, statistically insignificant. Together, these three variables indicate that there was only a positive effect of announcing the four aims, in tandem with the empirical evidence, on the staying in intentions of respondents who voted Labour and not on National voters or minority voters/non-voters. Second, the parameter estimate for ‘Aims no evidence’ is positive and statistically significant, but somewhat larger than in Model 2, whereas the parameter estimate for the interaction variable ‘National × aims no evidence’ is statistically significant and negative. The parameter estimate for the interaction variable ‘Labour × aims no evidence’ is, however, statistically insignificant. Together, these three variables show that the announcement of the four aims, this time without evidence, increased the chances of ‘staying in’ for all respondents (e.g. Labour voters, minority party voters, non-voters), though the effect was much smaller for National voters. Essentially, the positive overall effect of disclosing the four aims was largely, but not completely offset, for National voters only.
Models 2a and 3a in Table 5
Table 5 outlines the findings for Models 2a and 3a, predicting ‘support’ for the default. Results for Model 2a show that the parameter estimate for ‘Aims with evidence’ remains statistically insignificant, while that for ‘Aims no evidence’ is still statistically significant and positive. Both results changed little with the addition of more independent variables. The parameter estimates for ‘Labour Party’, ‘Union member’ and ‘Female’ are all statistically significant and positive. Hence, Labour voters, union members and women were more likely to support the union default than other groups (e.g. minority party voters/non-voters, non-union employees, males). The parameter estimates for ‘National Party’, ‘Manager’, ‘Age’, ‘NZ European’, ‘Regular part-time’, ‘Regular full-time’ and ‘Public sector’ are all statistically significant and negative. Hence, National voters, managers, older individuals, NZ Europeans, full-time and part-time employees and public sector employees were less likely to support a default than other groups (e.g. minority party voters/non-voters, non-managers, younger employees, non-Europeans and other Europeans, non-regular employees, private-sector/non-profit employees).
Model 3a is the same as Model 2a, but for the incorporation of the four interaction variables. Parameter estimates and statistical significance levels for the control variables are little different in Models 2a and 3a. The parameter estimate for ‘Aims with evidence’ is now statistically significant although negative, but that for the interaction variable ‘Labour × aims with evidence’ is even larger, statistically significant and positive. The parameter estimate for ‘National × aims with evidence’ is, however, statistically insignificant. Together, these three variables indicate that announcing the four aims and substantiating evidence generally reduces the likelihood of supporting the union default across the various groups (e.g. National voters, minor party voters, non-voters), except for Labour voters. Essentially, the general negative effect of this announcement on support for the default was more than offset for Labour voters, rendering them the only voter/non-voter group more likely to support the default. The parameter estimate for the ‘Aims no evidence’ variable is statistically significant and positive, as before, but neither of the parameter estimates for the ‘Labour × aims no evidence’ and ‘National × aims no evidence’ variables is statistically significant. Thus, Model 3 still shows a large, positive main effect of announcing the four aims, absent any evidence, on the probability of supporting a union default across all groups, Labour voters, National voters and minority party voters/non-voters.
Discussion
Our results indicate there is considerable potential support for a union default among NZ employees. In our control group, 56% supported a default as against 59% in an earlier study of 1471 NZ adults 12 (Harcourt at al., 2020). Higher levels of support were found for union members and women, and lower for National Party voters, regular part-time and full-time employees, older employees, managers, the public sector and NZ Europeans. These results are also broadly consistent with patterns observed in the earlier study of NZ adults. An even higher percentage of respondents (65% of the control group) said they would maintain membership if auto-enrolled. Once again, union members and women were more likely to say they would remain, but National voters, managers and older workers were less likely to say they would. The ‘support’ and ‘staying in’ results were dissimilar in two respects. Employees at organizations with fewer than 20 staff were just as likely to support a default as employees in larger organizations, but were still more likely to say they would opt-out. Lower and higher income employees were just as likely to support a default, but those on higher incomes were more likely to say they would opt-out. These seeming inconsistencies might reflect stronger concerns among some groups (e.g. small business employees, higher income employees) about management retribution for remaining members (e.g. loss of promotion opportunities for higher income employees).
Our most important results confirm that positive framing increases both support for the default and intention to remain members following auto-enrolment. The announcement of the four public goods aims but with no evidence had a positive effect on the likelihood of supporting the default across the entire sample, equivalent to raising the probability of support from 56.1% to 71.7% (i.e. with an odds ratio change of 1.98). Its effect on intention to remain in a union was even stronger, but only for respondents who did not vote National, equivalent to raising the probability of maintaining membership from 65% to 86% (i.e. with an odds ratio change of 3.42). The effect on National voters was still positive, albeit much smaller (3.422 × 0.353 = 1.21 odds ratio change). Overall, these results are consistent with prior research, showing defaults are more popular when they have aims beneficial to much/most of the wider community (Junghans et al., 2015; Reisch and Sunstein, 2016). They are also consistent with studies revealing participants socially identify more easily with other public goods beneficiaries when superordinate goals are emphasized (Gerlach and Jaeger, 2016; Van Bavel and Pereira, 2018). Finally, they are consistent with research showing these effects are somewhat smaller for right-wing voters, more averse to notions of common interests (Grunhage and Reuter, 2020).
The same announcement, but with mention of supporting empirical studies, had a more differentiated impact. In this scenario, respondents who did not vote Labour (e.g. National voters, minor party voters, non-voters) in the 2020 election were actually less likely to support a default, equivalent to lowering the probability of support from 56.1% to 42.8% (i.e. with an odds ratio change of 0.586), but they were no less likely to maintain membership if auto-enrolled. Thus, the overall impact of the positive framing for these groups was negative for support and neutral for staying in, contrary to expectation. In contrast, the same scenario increased support among Labour voters, equivalent to raising the probability from 56.1% to 62.7% (i.e. with an odds ratio change of 0.586 × 2.248 = 1.317), and their propensity to maintain membership, equivalent to raising the probability from 65 to 85% (i.e. with an odds ratio change of 0.952 × 3.151 = 2.999). Thus, the impact of the second scenario was similar to the first for Labour voters, at least with respect to their intention to maintain membership if defaulted (i.e. 85 versus 86% chance of staying in the union if defaulted).
The provision of evidence occasioned a negative response from National voters and minority party voters/non-voters not matched in Labour voters. Evidence from persuasion-related research in marketing suggests how this might happen (Artz and Tybout, 1999). Announcing the four public goods aims alone invites respondents to focus attention upon superordinate goals likely appealing to most of them, except possibly right-wing (e.g. National) voters, with positive repercussions for their intentions to support the default and remain in membership (Gerlach and Jaeger, 2016; Van Bavel and Pereira, 2018). Likewise, it invites them to favourably evaluate the intentions of the policy-makers, in establishing default membership for essentially noble reasons. However, the introduction of evidence shifts respondents’ focus and invites them to evaluate the empirical basis for the claims associated with the four goals. If respondents trust the source of the information, the evidence is still likely to be uncritically accepted with no adverse effect upon intention to support the default or maintain membership. For example, it is likely most Labour voters viewed the researchers as unbiased and, thus, readily accepted the evidence’s validity. In contrast, if respondents are sceptical about the source, they are more likely to closely scrutinize the evidence. If the source is an expert (e.g. an academic), they are also likely to expect any evidence to be quantitative (e.g. detailed and precise) and, thus, verifiable. Experts are expected to have access to such quantitative evidence, and so their failure to produce some risks being perceived as manipulation. In our particular case, no quantitative data were supplied (see endnote 7). It is, therefore, likely non-Labour respondents suspected the researchers of having a manipulative intent and so reacted by rejecting the default proposal (see Artz and Tybout, 1999).
Our results do have caveats. Efforts were made to help ensure our findings were generalizable by recruiting a large sample (i.e. 1047 respondents), collecting data from a broad cross-section of NZ adults, randomly assigning individuals to the control and experimental groups, and controlling for the influence of mainly demographic factors on individual choices. However, some groups were over-represented in our sample (e.g. women, public sector workers, union members, Labour voters), whereas others were under-represented (e.g. younger workers, non-voters). It is, therefore, possible our data do not completely reflect the composition of views in the general adult population. This is to some degree inevitable even with a stratified sample. However, in defending our findings’ validity, we emphasize again their similarities with findings from the previous NZ study (Harcourt et al., 2020) and that, even among employee groups less enthusiastic about the default, the majority or at least a large minority supported it and most still indicated they would maintain membership if defaulted. For instance, in the control group, 60.8% of males, 61.6% of private sector employees, 57.3% of non-union employees and 53.5% of National voters said they would remain union members if defaulted. Likewise, 52.2% of males, 53.5% of private sector employees, 50.1% of non-union employees and 42.6% of National voters still supported the union default. Overwhelming opposition and opt-out intentions were not common to any of the groups under study. Even 55.0% of managers said they would remain members if defaulted, and 44.2% supported the union default. Furthermore, our sample was not obviously skewed towards those with more favourable views of the default, as some more pro-union groups (e.g. younger workers, non-voters) were under-sampled. We also ran statistical tests, involving moderator variables, to determine whether females, union members and public sector workers were more responsive to the experimental scenarios than other workers. In virtually all cases, we found they were not. 13 This suggests that over-sampling of these groups was not the cause of the experimental scenario effects observed.
The most serious limitation of our study is, of course, that it is based on a survey of workers’ views of a policy that does not yet exist. How workers might respond to an actual union default remains to be seen and could be different. For instance, workers might prove more supportive of a default once they have experienced and/or observed some of the benefits of union membership (e.g. higher wages, lower income inequality, increased employer legal compliance, increased worker well-being). Ideally, future researchers will have opportunities to study the union default following its implementation in a particular country (e.g. New Zealand) or sector (e.g. public sector). These studies will then have more definitive lessons for adoption in other countries and/or sectors.
Conclusion
We believe our article provides considerable analytical insight into processes needed for making good on the increasing recognition by mainstream bodies like the IMF and OECD that unions are important to preventing economic inequality and maintaining/improving living standards. We suggest a hard form of state regulation is required to achieve this. In this context, our research has significant implications for unions and their supporters. Positive framing would likely increase support for a default and intentions to maintain membership if defaulted. Such framing is likely to have a particularly large effect on voters of left-of-centre parties traditionally affiliated with unions (e.g. Labour, Social Democratic, Democratic parties).
Our findings have four implications. First, they provide a strong basis for rejecting a more right-wing view of unions, in which individual liberty and rights are elevated to paramount virtues and where the exercise of choice, including the decision to join a union, is presumed to be rightly and only predicated upon a comparison of the individual’s costs versus benefits (see Harcourt et al., 2021). 14 Our study confirms workers are more likely to remain in unions and support the default when exposed to framing that draws attention to the group benefits unions provide (e.g. their public goods). Second, it follows that any messaging around the default should emphasize unions’ public goods benefits, especially when targeted at groups likely to value such benefits. It is similarly important, in targeting others initially less convinced of unions’ efficacy (e.g. disillusioned non-voters and centrist voters), that any public goods framing also be accompanied by detailed supporting evidence of unions’ positive public goods impacts. Effectively reaching and influencing these segmented audiences is possible, as both recent election communications and modern advertising demonstrate. Third, allies of unions, comprising left-of-centre political parties, think-tanks, charities, pressure groups and community organizations, could similarly assist in promoting the default by incorporating the same sorts of public goods framing into their own messaging about unions. A broad, solid base of popular support would legitimize the default, providing a stronger rationale for its passage into law and deterring antagonistic governments from seriously considering its repeal. Fourth, and most importantly, a proposed programme of employment relations reform needs to incorporate complementary elements deliberately and obviously designed to enable unions to deliver public goods (see Madland, 2021). Though the details are beyond our scope here, this programme would involve a number of features intended to facilitate union organizing, even in hitherto hard-to-reach smaller firms, where many low-waged, vulnerable workers work. In particular, it would certify unions to collectively bargain for broader bargaining units, each encompassing the smaller firms within a given industry, region, or supply chain, rather than just single enterprises or establishments as with the current US employment relations system. A more inclusive system is needed to spread the benefits of union representation more evenly, and empower unions to play their various public goods roles.
No doubt, employment relations reforms would be difficult to bring about, because the political challenges would be major and come from several quarters (e.g. business opposition in the media, court litigation, elections at different levels of government in Australia, Canada and the US). However, moderation in any employment relations reform, such as tinkering with the National Labor Relations Act, is not the answer. More radical and broader change is needed to facilitate unionization of precarious workers and workers at smaller employers. Neither is an appeal based on individual rights (e.g. freedom of association for the individual) or benefits (e.g. higher wages in your firm). Our results show a default is more likely to be supported and prompt workers to remain members when it is promoted and designed to serve broader public purposes a wide range of people can relate to, and identify with, as common interests. And, crucially, earlier research suggests a default would be strongly supported by unions themselves. When interviewed, legal experts from 11 major NZ unions were all supportive of a default and had many ideas to make it work in practice (Harcourt et al., 2021). Contact with the NZ Council of Trade Unions and our experience of presenting to their affiliated unions also suggest there is across-the-board backing from union leaders. Indeed, there is strong interest from unions in finding ways to sell the proposal to the general public, and Labour voters in particular, via appropriate framing.
At present, prospects for introduction of a union default in Britain and the US are not looking as favourable. The Conservative government in Britain is indifferent or even hostile to unions. The US Democratic administration is friendlier to unions, but has experienced difficulties passing any pro-union legislation, most recently the Bill called Protecting the Right to Organize Act in the Senate. The situation in Australia, with a newly elected Labor government, is more promising but there is no commitment to radical change at this stage. We would, therefore, argue that necessary and sufficient conditions for consideration of a default system would respectively involve the election of left-of-centre parties to national office, and those parties being committed to radical reform of the labour relations system. How these nations proceed could depend on what happens first in New Zealand.
Footnotes
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: We are grateful to the New Zealand Law Foundation for providing the funding to undertake the data collection.
