What Do Workers and the Public Want? Unions’ Social Value

Abstract

Using data from a national sample of American workers, the authors develop measures for “prosocial unionism”—the belief that unions contribute to the common good—and use regression analysis to determine its impact on public support for unions and on workers’ likelihood of supporting a union in a representation election in their workplace. Results show that the public’s support for unions is stronger when the public believes that unions act in the interests of all working people instead of just their members. The analyses also show that workers who believe unions have social benefits are significantly more likely to say they would vote “yes” in a union election than those who do not hold this belief. These findings imply that if unions address political and social justice goals that transcend the workplace, their legitimacy and their success in attracting public support and members may be enhanced and help stem the tide of shrinking union density.

Keywords

attitudes toward unions labor unions public opinion of unions motivations for union support social movement unionism social benefits of unions

Being part of a union is about so much more than just the individual benefits. It’s about being part of a collective movement creating a good society with shared prosperity, security and fairness for all.

—Frances O’Grady (2014)

The labor movement is at a turning point. Years of declining density have led scholars and organizers to draw grim conclusions but also to outline paths forward. Regarding the decline, according to one labor scholar, “Today the only thing big about ‘Big Labor’ is its problems” (Rosenfeld 2014: 1), and another finds the state of the traditional labor movement “dismal” (McCartin 2017: 64). Yet the creation of new institutions and new movements—such as worker centers and Bargaining for the Common Good—has rekindled the idea of a labor movement with a purpose that transcends the business–unionism model of the second half of the 20th century. Many labor scholars cheer these developments, but we know little of a scientific nature about how the general public and workers regard them.

Understanding the importance of unions as agents of social betterment, or to put it more strongly, as warriors for social justice, is critical for unions striving to organize and represent workers in a time of severe retrenchment (Clawson 2003; Milkman 2013). This article explores whether perceiving unions as dedicated to the common good—or as “prosocial”—positively sways public opinion; it also assesses whether workers value unions more for their prosocial actions or for the economic returns they bring members. More broadly, we ask whether the conceptualization of unions as working for the betterment of society as a whole is an impetus for supporting them in the 21st century. Scholars agree that unions have been in a defensive posture for several decades, and business-as-usual will no longer suffice; unions must tailor a message that convinces workers that unions are organizations they want to support and be part of. What should be the role of social betterment in that argument? Should US unions focus on “delivering the goods” as an organizing strategy or should they home in on social goals? This study can inform such decisions by highlighting the extent to which the social betterment orientation matters to the public and to workers contemplating union formation.

We address these questions by drawing on a national poll of the working-age population that is unusual in its inclusion of numerous items reflecting beliefs about unions. Our analysis uses regression analysis to consider the effect of a belief that unions act on behalf of the collective good, examining three outcomes: general attitude toward unions, approval or disapproval of unions, and the likelihood of voting “yes” in a representation election. We also compare the power of a belief that unions are prosocial to the power of a belief that they focus on helping members. In sum, we seek to determine whether an orientation to unions as defenders of the common good—as guardians of a larger social purpose—resonates with the public and with workers.

Some analysts argue that homing in on social goals is a necessity for organizing success and revitalizing labor more generally, and a rich body of case studies has illustrated how union organizing success has been predicated on moving beyond bread-and-butter issues to appeal to workers’ broader concerns with social justice. Yet, case studies represent islands of success in a sea of low levels of union density and power. By empirically testing the relative importance of prosocial union strategies in garnering the support of workers and the public, this study contributes to the conversation about how union revitalization can best be accomplished.

Moreover, a functioning democracy is predicated on levels of inequality that are not stratospheric, and unions are a major force in reducing these levels (Rosenfeld 2014). In an era of historically high levels of inequality, it is imperative to understand how one bulwark propping up democracy—unions—can be strengthened. To the extent that unions can be crafted to be more relevant and appealing, democracy will be strengthened.

Prior Literature

Theoretical debates on the question of whether workers and the citizenry care only about personal gains date back more than 100 years, with notable contributions including Perlman’s “psychology of the laboring man,” with its appeal to the workplace aspirations of “Tom, Dick, and Harry,” which pointed to the powerful role of self-interest (Perlman [1928] 1979: 237, 274), and Kimeldorf’s conclusion that unions “provided one of the most enduring links between class and protest for the past century” (1991: 100–101). After all, unions began and are still the backbone of the labor movement—inherently engaged in the political task of promoting the interests of the working class as a whole (Neufeld 1982). And they have traditionally dominated that pulpit: Rosenfeld (2014: 5) pointed out how “the labor movement has stood as the most prominent and effective voice for economic justice in the United States.”

Perceptions about the function of unions—whether they merely serve their own “vested interests” or instead wield a “sword of justice” to improve society (Flanders 1970: 13–23)—matter because those perceptions influence the commitment of workers to union campaigns, public policy toward unions, and the power they can muster via popular support (Bok and Dunlop 1970: 11–12). Considering the decline in union density, one scholar argued that “[Many] unions . . . have recognized that they can sustain or recapture a significant role only by forging effective links with the other components of civil society. It seems clear that part of the problem is an erosion of credible mobilizing rhetorics, of visions of a better future, of utopias. Building collective solidarity . . . is part of a battle of ideas” (Hyman 2001: 174; italics in the original). Scholars have suggested that US union failures in that battle have contributed to their exceptional decline; when compared to Canadian unions, which are generally similar except for a stronger focus on social issues, the US decline was steeper (Robinson 1994; Godard 2009).

Several successful union campaigns have relied on the notion of building collective solidarity in the service of social betterment (Milkman 2013; Rosenfeld 2014). A rich set of case studies has documented how labor has drawn on a dynamic model of social movement unionism that forges alliances with social justice or community organizations and that pursues goals beyond the workplace, often ones that cut across race and social class, such as immigration, police harassment, or public education. The 1990s witnessed the success of the Service Employees International Union’s Justice for Janitors campaign, for example, which placed immigrant rights center stage, arguing that allegiance springs from unions working toward larger community issues (Savage 1998; Clawson 2003; Fantasia and Voss 2004). The striking Chicago Teachers Union in 2012 couched their demands in prosocial terms by demanding increased funding for schools, and this altruistic framing was crucial for garnering the support of the parents and the community that led to success (Greenberg and Lewis 2017). Partnerships between unions and worker centers, which connect minority or immigrant communities’ concerns to workers’ rights, represent one avenue that has proven effective (Milkman and Ott 2014; Fine et al. 2018). Bargaining for the Common Good, a movement in which unions leverage their statutory right to bargain to make demands that include community-based ones, is another (McCartin 2017).

Indeed, a narrow business–union orientation focusing on members’ immediate economic interests and relying on staff rather than grassroots support can be less effective than social justice mobilization (Fantasia 1988; Lopez 2004; Greenberg and Lewis 2017; Ellem, Goods, and Todd 2020). In short, if unions address political and social justice goals that transcend the workplace, both their legitimacy in the public’s eye and their success in attracting members may arrest and even reverse the trend of shrinking union density.

Our knowledge base, however, is limited by the nature of case study evidence, which makes drawing generalizable conclusions difficult. We draw on a national sample to quantitatively assess the relative importance of instrumentalism and social betterment in the public’s and workers’ support for unions. If prosociality turns out to matter, unions could gain precious public support and new members by turning their attention in that direction.

A preponderance of US studies about motivations for supporting unions center on unpacking the components of an instrumental orientation—What is the role of wages? Benefits? Job security?—as if these were workers’ sole motivations. An example of this orientation comes from an interview with a British Trades Union Congress organizer who ruefully likened some workers’ views of the union to “the fourth emergency service: First you have the police, the ambulance and the fire brigade, then the union. Just as you hope you never need to call the police, you hope you don’t need the union” (Tapia 2013: 679). Research indeed shows that people who perceive unions as instrumental in helping workers achieve better wages, benefits, and working conditions are more inclined to vote for union representation (e.g., Kochan 1979).

By contrast, few quantitative studies have examined how unions can “win” on social-betterment grounds, and fewer still consider the relative importance of unions’ instrumental and social-betterment orientations. This article draws on a relatively rich data set to address two concerns. The first is whether and to what extent the public’s views of unions’ social betterment purpose are associated with their decision to support or not support unions. The second is whether unions’ social betterment purpose relates to workers’ likelihood of voting for a union in a representation election.

Only two studies offer findings that touch on the notion of social betterment. Drawing on two data sets, one from 1984 and another from 1988, Fiorito (1992) analyzed the relative influence of an instrumental and a prosocial orientation on how employees would vote in a representation election. This earlier measure of “altruism” captures our notion of prosocial: “Unions work to get legislation that helps all working people, whether they are union members or not” (1992: 29), and it was a key predictor of a “yes” vote. Another study, also from an earlier era and better known for its emphasis on workers’ pragmatic or instrumentalist approach to union support, nonetheless offered support for the importance of prosocial factors: The most frequently offered reason for voting against representation was that “unions were out more for their own self-protection than for the good of society in general” (Kochan 1979: 25). Both studies, while on-point for our purposes, are dated, and since the 1970s and 1980s, inequality has substantially increased. In 2005, when our data were collected, inequality was reaching its peak before the Great Recession, and that highly unequal context continues to grow at an even faster pace today. So the question of what would impel support from the public and from workers in times of sharp inequality is timely and worth considering.

The preceding discussion leads to our hypotheses. First, we expect that people rating unions more highly on their prosocial benefits are more likely to have higher union-approval ratings. Second, we expect that workers who rate unions more highly on prosocial benefits are more likely to vote “yes” in a union-representation election. In both cases, we expect that the perception that unions are prosocial is at least as important as the perception that they are instrumental.

Data, Measures, and Methods

A telephone survey of Americans ages 18 and older was conducted in February 2005 by Peter D. Hart Research Associates on behalf of the AFL-CIO, and the latter group provided access to the data at our request. Participants were solicited from across the nation through random-digit dialing (RDD) for an interview lasting approximately 25 minutes. The sample was drawn in the following manner: 300 geographic points were randomly selected proportionate to the population of each region and, within each region, by size of place. Individuals were selected in accordance with a probability sample design that gives all telephone numbers (both listed and unlisted) an equal chance of inclusion. One adult, 18 years old or over, from each household was included, selected by a systematic procedure to provide a balance of respondents by sex. The AFL-CIO sponsorship was revealed only upon respondents’ request (as is standard practice among survey research professionals).

In regard to the sample, we are missing information on the response rate, and our contact at Hart Research Associates explained that records from this 2005 data collection project are no longer available. Yet, we see no reason to suspect that standard protocols were not followed. Hart Research Associates is a leading national polling firm in operation since the early 1970s that, in addition to corporate and nonprofit clients, regularly contracts with NBC News and the Wall Street Journal to conduct public-opinion polling. The Hart partner who designed and directed the poll confirmed that the response rate was unlikely to differ from the 17% rate that Gallup (Marken 2018) reported as the average response rate for telephone surveys in 2005.

The representativeness of the final samples is illustrated by comparing their characteristics with those from the Census (U.S. Census Bureau 2006, 2007) and Current Population Survey (CPS) tallies for 2005 (U.S. Bureau of Labor Statistics 2006). To gauge the representativeness of our “general population” sample, we compared the Hart data to Census data for persons age 18 and older. As Appendix Table A.1 shows, Hart respondents were 47% male (48% in the Census data), 35% southern residents (36%), and had similar distributions for age and schooling. A similar comparison for our “wage and salary worker” sample, detailed in Appendix Table A.2, reveals that Hart respondents were 52% male (52% in the CPS), 34% southern (36%), 81% full-time (82%), 13% union members (13%), and closely followed CPS and Census characteristics for age and education. Hart and CPS tallies also aligned on race-ethnicity, public-/private-sector employment, and occupation, after taking into account classification systems that differed, particularly in the case of race/ethnicity. Since the Hart samples are from a national public opinion poll, not an organizationally endorsed survey, and because of the similarity between the Hart samples and the US population and US workforce, we have some confidence in the samples’ representativeness.

Materially, the data set includes more than 20 items assessing beliefs, attitudes, and feelings about unions and their role in society, including credible indicators of our key constructs. We scrutinized the Hart questions and identified items we could use to form conceptually distinct measures of prosocial unionism and instrumental unionism, as well as a measure of general attitude toward unions and an item, union approval, comparable to the long-running Gallup poll question. We also identified measures or proxies for a wide range of control variables.

Measures

Dependent Variables

We examined three dependent variables: general attitude toward unions, union approval, and pro-union voting intention in a hypothetical representation election. General attitude is a multi-item scale formed by combining items that sampled a broad range of the construct domain (see McShane 1986). We focused our measure on items that referenced unions (rather than their leaders) and assessed general feelings or attitudes about them. Items (see Table 1) touched on positive or negative feelings toward unions, whether the respondent typically sides with workers and unions or with management and business in disputes, whether unions help or discourage individual efforts, whether unions are innovative or old-fashioned, and whether members make important decisions or simply follow union leader dictates. The items used various response scales, so we standardized items before summing them to form our general attitude measure. The resulting measure has a mean of 0, standard deviation of 0.69, a slight negative skew (skewness = −0.23), and a somewhat flat distribution (kurtosis = −0.89).

Table 1.

Variable and Item Descriptions, Coding, and Descriptive Statistics

Variables / Items	Summary description	Item(s)	Mean	SD
General union attitude	Feelings toward labor unions, labor-management dispute leanings, perceived union democracy, innovativeness, etc. Cronbach α = 0.82	Scale formed from standardized versions of the 6 items below (Gen1 – Gen6)	0.00	0.69
Gen1	Positive or negative feeling toward labor unions 1=Very negative, 2=Somewhat negative, 3=Neutral or don’t know name/ not sure, 4=Somewhat positive, 5=Very positive	I’m going to read you the names of several institutions. I’d like you to rate your feelings toward each one [Labor unions].	3.28	1.13
Gen2	Tendency to favor business or labor union during a dispute. Coding: 1=Business, 2=Middle/neither side (volunteered) or not sure, 3=Labor union	When you hear about a dispute between a business and a labor union, which side do you generally tend to favor – the business or the labor union? Form A	2.19	0.77
Gen3	Tendency to favor business or workers union during a dispute. Coding: 1=Business, 2=Middle/neither side (volunteered) or not sure, 3=Labor union	When you hear about a dispute between a business and its workers, which side do you generally tend to favor – the business or the workers? Form B	2.36	0.70
Gen4	Common stem: I’m going to read you several pairs of phrases, and I’d like you to tell me in each case . . . which better applies to labor unions today? [Read Item]. Coding: 1=Statement B, 2=Mixed, both, or not sure, 3=Statement A	Statement A: Help individuals make the most of themselves [OR] Statement B: Discourage individual effort and initiative.	2.13	0.91
Gen5		Statement A: Are innovative and change with the times [OR] Statement B: Are old-fashioned and backward-looking in their approach.	2.10	0.89
Gen6		Statement A: Let the members make important decisions [OR] Statement B: Make decisions and tell members what to do.	1.94	0.91
Union approval	General approval or disapproval of labor unions, 1=disapprove, 2=not sure, 3=approve	In general, do you approve or disapprove of labor unions?	2.44	0.81
Pro-union voting intention	Intention to vote for a union in a hypothetical representation election, 1=definitely against, 2=probably against, 3=probably for, 4=definitely for	If an election were held tomorrow to decide whether your workplace would have a union or not, do you think you would definitely vote for forming a union, probably vote for forming a union, probably vote against forming a union, or definitely vote against forming a union?	2.54	1.11
Prosocial unionism	Perception that unions help society, the poor, all workers, etc. (7-item scale). Cronbach α = 0.87	Scale formed from standardized versions of the 7 items below, Psoc1 – Psoc5, Psoc8, and Psoc11	0.00	0.64
Psoc1	Common stem: I’m going to read you several pairs of phrases, and I’d like you to tell me in each case which better applies to labor unions today? [Read Item]. Coding: 1=Statement B, 2=Mixed, both, or not sure, 3=Statement A	Statement A: Help the poor and disadvantaged in our society. Statement B: No longer help the poor and disadvantaged in our society.	2.09	0.87
Psoc2		Statement A: Are concerned about all working people [OR] Statement B: Are concerned only about their members?	1.75	0.91
Psoc3	Common stem: When it comes to [Read Item], do you think the impact of labor unions today is generally positive, neutral, or generally negative? Coding: 1=Negative, 2=Neutral or not sure, 3=Positive	Workers’ rights on the job. Form A	2.46	0.66
Psoc4		Giving working people a voice in the political process. Form B	2.24	0.77
Psoc5		The situation for workers in jobs with low wages and no benefits. Form B	2.15	0.77
Psoc8		Income and benefits of workers who are not in unions. Form B	2.00	0.70
Psoc11	Belief that more union representation would be good or bad for the country. 1=Bad, 2=Depends, some of both, or not sure, 3=Good	Do you think that it would be good for the country or bad for the country if more workers had union representation? Form A	2.29	0.84
Instrumental unionism	Perception that unions make a difference, provide value for dues, are effective for members, etc. Cronbach α = 0.80	Scale formed from standardized versions of the 5 items below	0.00	0.71
Inst1	Perception that unions are effective for members, 1=Not very effective, 2= Just somewhat effective, 3=Not sure (volunteered), 4=Fairly effective, 5=Very effective	Overall, how effective do you think labor unions are these days in improving wages, benefits, and working conditions for their members—very effective, fairly effective, just somewhat effective, or not very effective?	2.86	1.31
Inst2	Common stem: I am going to read you several pairs of phrases, and I’d like you to tell me . . . [Read Item]—which better applies to labor unions today? Coding: 1=Statement B, 2=Mixed/both or Not sure, 3=Statement A	Statement A: Can make a difference for workers today. Statement B: Are not really relevant or important anymore.	2.50	0.80
Inst3		Statement A: Deliver good representation for the dues they charge. Statement B: Charge too much in dues for what they deliver.	1.99	0.87
Inst4	Common stem: When it comes to [Read Item], do you think the impact of labor unions today is generally positive, neutral, or generally negative? Coding: 1=Negative, 2=Neutral or not sure, 3=Positive	Union members’ incomes and benefits. Form A	2.35	0.71
Inst5		Giving workers more say in the workplace. Form A	2.28	0.73
Controls
Feelings toward employer	Feelings toward employer. Coding: 1=very negative, 2= somewhat negative, 3=neutral or not sure, 4=somewhat positive, 5=very positive	I’m going to read you the names of several institutions. I’d like you to rate your feelings toward each one as either very positive, somewhat positive, neutral, somewhat negative, or very negative . . . the company or organization that you work for?	4.04	1.07
Parent union membership	=3 if either parent was a union member, =2 if not sure, =1 if no	When you were growing up, did either or both of your parents belong to a labor union?	1.70	0.92
Family or friends	=3 if family member or close friend is a union member, =2 if not sure, =1 if no	Do you have any family members or close friends who currently belong to a labor union?	2.14	0.96
Union voting efficacy	=1 if employers oppose union, =2 if not sure, =3 if employers take no position, =4 if employers support the union (volunteered response)	As you may know, elections are sometimes held in a workplace to determine if the employees will be represented by a union. When such elections are held, do employers generally oppose the union and try to convince employees to vote no, or do employers generally take no position and let the employees decide on their own?	1.76	0.91
Professional or managerial	Occupation. Coding: =1 if occupation is professional or managerial, 0 otherwise	What type of work do you do? Response options were Professional, manager; White-collar worker; Blue-collar worker; Farmer, Rancher; Self-employed/Own business; Other, Not sure, and Not employed.	0.30	0.46
Blue-collar	=1 if occupation is blue-collar, 0 otherwise	What type of work do you do?	0.33	0.47
Part time	Part-time employment. Coding=1 if employed part-time, =0 if employed full-time	Do you work on a full-time basis or a part-time basis?	0.20	0.40
Public employee	Public employee, including public schools. Coding=1 if a public employee, 0 otherwise	Would you tell me whether you are a public employee—that is, are employed by federal, state, or local government, a public school system, or some other governmental agency?	0.27	0.44
Supervisor	Supervisory status. Coding: =1 if respondent supervises other workers, 0 otherwise	Do you personally supervise two or more employees at your job?	0.32	0.47
Family income	Family annual income. Coding 1=Less than $10,000, 2=Between $10,000 and $20,000, 3=Between $20,000 and $30,000 . . . 8=More than $100,000	If you added together the yearly income of all the members of your family who were living at home last year, would the total be less than $10,000, between . . .	4.68	2.16
Individual earnings	Respondent annual earnings. Coding 1=Less than $10,000, 2=Between $10,000 and $20,000, 3=Between $20,000 and $30,000 . . . 8=More than $100,000	About how much did you earn from your work last year—less than $10,000, between $10,000 and $20,000, . . .	3.67	1.94
Schooling	9-point ordinal scale for years of schooling completed, 1=grade school, 2=some high school, 3=high school graduate, 4=some college, no degree, 5=vocational training/2-year college, 6=4-year college/bachelor’s degree, 7=some postgraduate work, no degree, 8=2-3 years postgraduate work/master’s degree, 9=doctoral/law degree	What is the last grade that you completed in school?	4.53	1.92
Age group	12-point ordinal scale for age group. Coding: 1=ages 18–24, 2=25–29 . . . 11=70–74, 12=age 75 and over	How old are you? (or if refused: Well, would you tell me which age group you belong to?)	5.65	3.22
Democratic Party	7-point ordinal scale for political leaning to Democratic Party (1=Strong Republican, 2=Not very strong Republican, 3=Independent/lean Republican, 4=Strictly independent, not sure, or nothing, 5=Independent/lean Democrat, 6=Not very strong Democrat, 7=Strong Democrat)	Generally speaking, do you think of yourself as a Republican, a Democrat, an independent, or something else?	4.16	2.03
Hispanic	Hispanic/Latino/Spanish-speaking identification. Coding: 1=Hispanic, Latino, or Spanish-speaking background, 0=other background	For statistical purposes only, are you from a Hispanic, Latino, or Spanish-speaking background? [If not], what is your race—white, Black, Asian, or something else?	0.16	0.37
Black	1=Black racial background, 0=other background	See above	0.16	0.37
Asian-Other	1=Asian or other racial background, 0=not Asian or other background	See above	0.04	0.18
Southern region	1=Southern residence, 0=Other region residence	Based on standard state and region codes recorded by interviewer	0.35	0.48
Gender	Gender. Coding: 1=Female, 0=Male	Recorded by interviewer	0.53	0.50
Union member	Union membership. Coding: 1=Union member, 0=Not a union member	Are you a member of a labor union?	0.12	0.32

Notes: Variable names appear in italics. Item names appear in standard typeface.

Union approval uses responses to a question very similar to one that Gallup polls have posed since the 1930s—“In general, do you approve or disapprove of labor unions?”—with the response options “Approve” (66%), “Disapprove” (20%), and “Not sure” (14%). While arguably tapping the same construct domain as our general attitude measure, this question has its own 80-years-plus history that encourages considering it independently.

Pro-union voting intention relies on a question that asks how respondents would cast their vote in a hypothetical union certification election at their current workplace (Table 1 shows exact wording). Responses were recorded on a 4-point scale, on which (after recoding) 1 = definitely vote against forming a union (26% of the sample), 2 = probably vote against . . . (17%), 3 = probably vote for . . . (34%), and 4 = definitely vote for . . . (23%). We limited the analysis for this variable to non-union wage and salary workers (neither self-employed nor farmers or ranchers). This measure and sample inclusion criteria are similar to those used in numerous studies of union voting intentions, including Kochan, Yang, Kimball, and Kelly’s 2019 study about the effect of perceptions of a lack of worker voice relative to desired voice (“voice gap”).

Independent Variables

We created multi-item scales for two independent variables. Items for prosocial unionism tap beliefs about whether unions help the poor and disadvantaged in society, are concerned about all workers or just their members, improve worker rights on the job, give workers voice in the political process, help workers with low wages and no benefits, improve income and benefits for non-union workers, and whether it would be good or bad for the country if more workers had union representation. Because of varying response scales, we standardized all items before summing and dividing by the number of items.

The instrumental unionism measure taps into beliefs about union effectiveness for members in improving pay, benefits, and working conditions, whether unions make a difference for workers or are irrelevant, and whether unions deliver good representation for dues, improve member incomes, and increase worker voice in the workplace. Various response scales were used, so we standardized items prior to summing to form the instrumental unionism measure. (See Table 1 for item wording for all measures.)

Several variables control for individual characteristics and are used in all equations: education, age, ethnic/racial group identification, southern-region residence, and gender. We also include political leanings (e.g., Republican versus Democrat). Two additional controls are derived from the “subjective norms” construct of Ajzen’s (1991) theory of planned behavior, which implies that others’ opinions are influential. Parent union membership and family or friends assess union membership status among those groups and serve as proxies reflecting an assumption that these people are likely to encourage others to support unions.

Two measures are included only in the general population sample used to model general attitude and union approval. A union membership variable is restricted to these models because our pro-union voting intent model, as in most union voting intent studies, includes only non-members. Family income is included only in these models in place of the individual earnings measure used in the employed sample.

Other control variables apply only to the pro-union voting intent model, including ones that are employment-based, such as binary variables denoting broad occupational categories (professional/managerial and blue-collar), part-time status, public-sector employment, supervisory status, and an ordinal variable for earnings. Prior studies have found an important role for job satisfaction or related measures (e.g., Kochan 1979), and our data set offered a measure of feelings toward one’s employer (feelings toward employer), arguably more relevant than job satisfaction (Martinez and Fiorito 2009). Union voting efficacy is a proxy for another Ajzen concept, that of perceived control—in this case, a sense that one has control over the ability to successfully vote for union representation. We assume that respondents expecting employer opposition, which is commonplace and influential in the United States (Godard 2009), will foresee futility in achieving the goal of union representation compared to those anticipating employer neutrality.

Missing Data Replacement

Most questions were asked of all participants, but a split-form design, in which half the respondents completed one version of the questionnaire (Form A) and the other half completed another (Form B), was used extensively for items we identified for general attitude, prosocial unionism, and instrumental unionism. The general attitude measure comprised four common items and one additional item each from Form A and Form B. Prosocial unionism comprised two common items, two items unique to Form A, and three items unique to Form B; instrumental unionism comprised three common items and two asked only in Form A. We used the “person mean” (or Mean_person) technique (Roth, Switzer, and Switzer 1999), which makes maximum use of actual responses and provides reasonable estimates of missing values. The technique uses a respondent’s available items for a given scale, calculates a mean value from these, and imputes that mean in place of missing items for that person. Roth et al. described this technique as “most effective” because “imputation techniques used at the item level preserve and use a great deal of information by not deleting real data and using existing data to estimate missing scores” (1999: 228). Person-based mean imputation is analogous to a regression-based imputation technique in which unit weights rather than regression weights are used, so that it “acts like a robust, unit-weighted regression approach” (Roth et al. 1999: 229). Both person mean and regression-based imputation at the item level conform to the “fundamental” principle of missing data treatments: “use all the available data” (Newman 2009: 11). While all missing data replacement techniques can introduce bias, Roth et al.’s Monte Carlo study found the person-mean imputation technique produced “nearly imperceptible levels of bias” (1999: 227). We also experimented with regression-based imputation and found similar results, as reported below.

Statistical Methods

The analysis proceeds as follows. We first report the results of confirmatory factor analyses (CFAs) to assess the consistency of the data on union beliefs with our conceptual scheme, which specifies three distinct dimensions of beliefs and attitudes about unions among the 18 items used to form general attitude, prosocial unionism, and instrumental unionism. Bivariate correlations showing relations among selected variables are then presented, followed by ordinary least squares (OLS) regression to estimate a model for general attitude, and by ordered logistic regression to estimate models for union approval and pro-union voting intent. Finally, we report results from robustness checks.

Results

Strong correlations among prosocial unionism, general attitude, and instrumental unionism raise questions about discriminant validity. That is, do these measures of union beliefs and attitudes represent three separate constructs? We assigned items to constructs on conceptual grounds, namely item content relative to our constructs. But we also offer empirical support for our item assignments and the scales we formed from the 18 items on union beliefs and attitudes.

At the outset, we note that the three scales each achieved an internal consistency reliability that was better than the conventional α = 0.70 minimum standard, with values of 0.82 for general attitude, 0.87 for prosocial unionism, and 0.80 for instrumental unionism. These values may be inflated slightly by the person-mean missing-data replacement technique, but the degree of inflation is probably small: Roth et al. (1999: 215–16) reported that the person-mean technique reproduced alphas within 0.02 of actual values under missing data circumstances similar to ours, that is, approximately 30% missing data for the 18 items.

We conducted CFAs for one-, two-, and three-factor solutions as a direct assessment of dimensionality or discriminant validity among our three multi-item belief and attitude measures. We found that the best-fitting model was the three-factor model. Table 2 presents these comparisons and information on four commonly cited fit criteria for CFA.

Table 2.

Confirmatory Factor Analysis for Union Belief and Attitude Items

Model	Factors	X²	df	X²/df	CFI	SRMR	RMSEA	X² Change
Model 1	Three-Factor:PU, IU, GA	773.4	132	5.9	0.94	0.04	0.06	—
Model 2	Two-Factor:GA, PU + IU	1309.3	134	9.8	0.89	0.05	0.08	535.9***
Model 3	Two-Factor:IU, GA + PU	1114.2	134	8.3	0.91	0.05	0.07	340.8***
Model 4	Two-Factor:PU, GA + IU	1230.7	134	9.2	0.90	0.05	0.08	457.3***
Model 5	One-Factor:PU + IU + GA	1603.5	135	11.9	0.86	0.05	0.09	830.1***

Notes: N = 1,382. CFI, comparative fit index; df, degrees of freedom; GA, general attitude; IU, instrumental unionism; PU, prosocial unionism; RMSEA, root mean square error of approximation; SRMR, standardized root mean square residual.

***

p < .01.

Dozens of CFA fit measures exist, and Kline (2005: 133–42) recommended that, at a minimum, four should be reported: model X², comparative fit index (CFI), standardized root mean square residual (SRMR), and root mean square error of approximation (RMSEA). The Cornell Statistical Consulting Unit (n.d.) recommends a cutoff of p > .05 for model X² (a badness-of-fit measure), 0.90 or higher for CFI (a goodness-of-fit measure), and 0.08 or lower for SRMR and RMSEA (also badness-of-fit measures). The model X² suggests a poor fit for all models, as is typical with large samples such as ours, and thus we focus on the other three recommended measures, which are less sensitive to sample size. First, the CFI of 0.94 for the three-factor model (our a priori specification) suggests a good fit, whereas the two-factor and one-factor models fare less well (CFIs < 0.92). The RMSEA is a parsimony-sensitive badness-of-fit criterion in that it is adjusted for model complexity, making comparisons across nested models more meaningful, and experts suggest that an RMSEA of 0.05 or lower is a “close” fit and that values up to 0.08 “represent reasonable errors of approximation” (Jöreskog and Sörbom 1993: 124). Our three-factor model fares best on this test (RMSEA = 0.06), as is also the case for the SRMR test (0.04).

Two additional perspectives on CFA model fit are provided via the sequential chi-square difference test (SCDT) and the sign and statistical significance of item loadings on factors. Both of these fit criteria include formal tests for statistical significance. The SCDT provides a test for comparative fit among nested models, and the results indicate that the three-factor model provides a significantly better fit (p < .01) than the one-factor and two-factor models. We also note that each item in the three-factor model loaded significantly (p < .05 or better) on its respective factor with the expected sign. All things considered, the three-factor model provides a reasonably good fit and the best fit among the models in Table 2.

Table 3 presents bivariate correlations for dependent variables (general attitude, union approval, and pro-union voting intent), prosocial unionism, instrumental unionism, and feelings toward employer.

Table 3.

Correlations among Selected Variables

	General attitude	Union approval	Pro-union voting intent	Prosocial unionism	Instrumental unionism
General attitude
Union approval	0.62***
Pro-union voting intent	0.69***	0.60***
Prosocial unionism	0.68***	0.55***	0.60***
Instrumental unionism	0.64***	0.56***	0.57***	0.63***
Feelings toward employer	−0.13***	−0.07**	−0.25***	−0.04	−0.03

Notes: N = 1,382 for correlations except N = 548 for correlations involving pro-union voting intent (difference in N attributable mainly to excluding union members and non-employed persons) and N = 982 for correlations involving feelings toward employer other than pro-union voting intent (difference in N attributable mainly to this question being restricted to employed persons).

***

Statistically significant at p < .01 or better, two-tailed test. ** Statistically significant at p < .05 or better.

The correlation matrix reveals preliminary support for the hypothesized effects for prosocial unionism. The perception that unions work for the greater good, prosocial unionism, was positively associated with all three dependent variables at the p < .01 level (general attitude r = 0.68; approval r = 0.55; pro-union vote intent r = 0.60).

We now turn to our main analysis. Table 4 presents regression results for each dependent variable. Our review of residual plots revealed no notable problems, and details on these analyses are available from the authors.

Table 4.

Regression Results for General Attitude, Approval, and Union Voting Intent

	General attitude(standardized betas)	Union approval(parameter estimates)	Pro-union voting intent(parameter estimates)
Prosocial unionism	0.42***	1.40***	1.41***
Instrumental unionism	0.32***	1.22***	0.98***
Feelings toward employer			−0.54***
Parent union membership	0.05**	0.27***	−0.07
Family or friends	0.00	0.18**	0.14
Union voting efficacy			−0.07
Professional or managerial			0.27
Blue-collar			0.45**
Part time			−0.07
Public employee			0.64***
Supervisor			−0.19
Family income	−0.04	−0.06
Individual earnings			−0.04
Schooling	−0.07***	0.01	−0.13**
Age group	−0.03*	−0.01	0.02
Democratic Party	0.16***	0.14***	0.12**
Hispanic	−0.02	−0.20	0.31
Black	0.04**	−0.12	0.53**
Asian–Other	0.07	0.73	0.25
Southern region	−0.00	−0.31*	0.14
Gender	0.06***	−0.17	−0.01
Union member	0.08***	1.07***
Estimation method	OLS	Ordered logistic	Ordered logistic
N	1,092	1,092	548
Fit statistics	R ²=0.61, F_{(14,1077 df)}=119.3***	R ²=0.39 (Cox-Snell), X²_14df=54.6***	R ²=.049 (Cox-Snell), X²_20df=365.1***

Notes: Intercepts not shown. For significance testing in the general attitude equation, two-tailed t-tests were used for unstandardized coefficients. For significance testing in the union approval and pro-union voting intent equations, Wald chi-square tests were used for logistic regression coefficients. OLS, ordinary least squares.

***

, **, and *: Statistically significant at the .01, .05, and .10 levels, respectively.

Considering first the general attitude equation, we see that prosocial unionism has a positive and statistically significant effect (p < .01). The standardized betas indicate that its effect size (β = 0.42) can be considered large (Cohen 1992). It is also considerably larger than influences captured by other predictors in the equation, including instrumental unionism (β = 0.32; F-test for difference in coefficients significant at p < .01).

The union approval equation shows a similar result, with a positive and statistically significant effect for prosocial unionism (unstandardized b = 1.40, p < .01; odds ratio = 4.07). Compared to other predictors, except perhaps for Instrumental unionism (b = 1.22, p < .01; odds ratio of 3.40) and Union member (b = 1.07, p < .01; odds ratio = 2.92), this effect size stands out.

Finally, using only non-member wage and salary workers, the pro-union voting intent equation shows similarly strong results for prosocial unionism (b = 1.41, p < .01), with an odds ratio of 4.09. Once again, compared to all other predictors in the model, this effect size is impressive. The next largest effect is for instrumental unionism (b = 0.98, p < .01) with an odds ratio of 2.65. The difference in coefficients (b) is significant at the .10 level.

Although odds ratios provide a way to compare effect sizes in a logistic estimation context, predicted probabilities allow for a more intuitive interpretation. As shown for union approval in Table 5, panel A, and for pro-union voting intent in panel B, the predicted probabilities change meaningfully in response to changes in prosocial unionism. Compared to a baseline of all predictors at their mean values, a one standard deviation increase in prosocial unionism pushes the predicted percent approving of unions from 74% to 92%, while the predicted percent disapproving falls from 10% to 3%. For pro-union voting intent, a one standard deviation increase in prosocial unionism boosts the predicted percent “definitely voting for” the union from 13% to 37% and reduces the predicted percent “definitely voting against” from 16% to 4%.

Table 5.

Illustration of Effect Sizes: Actual and Predicted Relative Frequencies from Table 4 Estimates for Union Approval (Panel A) and Pro-Union Voting Intent (Panel B)

Panel A	Union approval categories
	Disapprove (1) %	Not sure (2) %	Approve (3) %
Actual	20	14	66
Predicted with all independent variables at means	10	17	74
Predicted with +1.0 unit in prosocial unionism	3	5	92

Panel B	Voting intent categories
	Definitely against union (1) %	Probably against union (2) %	Probably for union (3) %	Definitely for union (4) %
Actual	26	17	34	23
Predicted with all independent variables at means	16	24	48	13
Predicted with +1.0 unit in prosocial unionism	4	10	49	37

Although the correlation between prosocial unionism and instrumental unionism (r = 0.63; see Table 3) suggests the possibility of multicollinearity problems, variance inflation factors (VIFs) under 2.0 and other results show that our sample has sufficient power to overcome this threat for all three equations. The tolerances for these two measures exceed 0.50, indicating that more than half the variance in each is unique variance not shared with other predictors in the model. Highly significant results for each demonstrate that the data provide a combination of sufficient unique variance in each independent variable and sufficient statistical power to detect distinct effects. This detection ability, not the level of correlation among predictors, is the real issue raised by multicollinearity (O’Brien 2007: 683): “If a regression coefficient is statistically significant even when there is a large amount of multi-collinearity—it is statistically significant in the ‘face of that collinearity.’”

Thus, it appears that self-interest (which our instrumental unionism measure taps) is far from the only motivation for holding positive general attitudes toward unions, approving of unions, or voting for unions in representation elections. Our results show that in all instances the prosocial unionism effect is at least as strong as the effect for instrumental unionism. The perception of prosocial unionism, that is, the belief that unions help all working people, is a powerful motivation that disposes people to hold a positive general attitude toward unions, to approve of them, and to vote for them in representation elections.

Results for control variables reveal no surprises, with findings roughly in accord with evidence from previous studies, although a few stand out. Results for parent union membership and for family or friends indicate that influential others’ views matter for union approval and disapproval, and parental membership matters for general attitude. Respondents with positive feelings toward their employers were much less likely to indicate a pro-union voting intention, confirming previous findings (e.g., Martinez and Fiorito 2009) and paralleling similar findings about job dissatisfaction (Budd 2017: 204–6). Blue-collar workers and public employees were also more likely to report a pro-union voting intention, the former perhaps reflecting traditional views on the appropriateness of unions for some worker groups rather than others and the latter perhaps reflecting muted cost-reduction pressures. The higher the educational level, the less likely were positive attitudes or votes for union representation, perhaps reflecting perceptions of individual bargaining power as obviating the need for collective action. People favoring the Democratic Party expressed more favorable union attitudes, greater likelihood of union approval, and a greater inclination to vote for union representation. Some race/ethnic effects were present, with Blacks evidencing a more positive general attitude to unions and a greater likelihood of voting for union representation (see also Gumber and Padavic 2020). Southerners were less likely to approve of unions. Women were more likely to hold a favorable general attitude. Finally, union members were more likely to have positive attitudes and more likely to express union approval.

Robustness Checks: Missing Data Replacement, Multicollinearity, Alternative Samples, and Alternative Measurement Scales

We now consider alternatives to the person-mean missing data replacement (MDR) technique and, given the fairly high correlation between prosocial unionism and instrumental unionism, we also address further the issue of multicollinearity. Accordingly, we conducted sensitivity checks by analyzing models with no MDR and regression-based imputation in addition to the person-mean MDR results reported above. The MDR and multicollinearity issues are intertwined in that the correlation between prosocial unionism and instrumental unionism depends partly on MDR, since using the full set of items for each scale requires MDR (as noted above). To address these issues, we specified the following alternative measures and sample restrictions and found that these variations produced very similar results to those reported above, and none produced meaningfully different results for the role of prosocial unionism.

Regression-based imputation (without an error component). This MDR technique is similar to the person-mean approach we used but relies on regression to estimate missing values based on modeling each item as a function of available items. Roth et al. (1999) referred to this technique and person-mean MDR as “very promising approaches” (1999: 211) and reported Monte Carlo evidence in support of this characterization. Results using this technique were similar to those that we report above.

Best single item. In the interests of simplicity, we chose one representative item (available for Form A and Form B; no MDR) that captured the central concept for the prosocial measure, the instrumentality measure, and for the general attitude toward unions measure. For prosocial unionism, we used item Psoc2 (see Table 1 for all item wordings), based on “Statement A: [Unions] Are concerned about all working people [OR] Statement B: Are concerned only about their members?” For instrumental unionism, we used Inst1: “Overall, how effective do you think labor unions are these days in improving wages, benefits, and working conditions for their members—very effective, fairly effective, just somewhat effective, or not very effective?” Finally, for general attitude we used item Gen1: “I’m going to read you the names of several institutions. I’d like you to rate your feelings toward each one [Labor unions].” Besides simplicity, another virtue of this approach is the relatively low correlation among the single-item measures: between instrumental unionism and prosocial unionism, 0.24; between instrumental unionism and general attitude, 0.38; and between prosocial unionism and general attitude, 0.31. Regression results were similar to those reported above.

Subsample results. We estimated separate models for Form A and Form B subsamples, creating scales with fewer items and lower reliabilities. Cronbach alphas ranged from 0.55 to 0.74 for instrumental unionism and from 0.63 to 0.65 for prosocial unionism. Positive effects (p < .01) for both prosocial unionism and instrumental unionism appeared in both subsamples, with the effects for prosocial unionism larger or not significantly different from those of instrumental unionism.

Union voting models. We ran these models excluding those who self-identified as supervisors (nearly one-third of our worker sample). This exclusion had no meaningful effect on the results.

General union attitude as a predictor. We believe that it tells us little about how beliefs about unions influence union voting intentions (see Montgomery’s [1989] reference to “emotional residue”), and it magnifies multicollinearity problems. At the suggestion of a reader, however, we added general union attitude as a predictor of pro-union vote intent and found that prosocial and instrumental unionism measures remain positive and significant, as is the general union attitude measure.

Expanded prosocial unionism scale. The data set offers the possibility of an expanded prosocial unionism scale, as it contains items that pertain to the economic as well as to the social movement qualities of unions. The additional four items assess whether unions generally have a positive impact on productivity and work quality, on America’s ability to compete, on the economy, and on consumer prices. Impacts of this sort can be seen as another dimension of the social benefits unions can foster (McKersie 2019: 42–43). Including these items in the prosocial unionism scale makes it slightly more reliable and produces similar regression results.

In sum, all alternative specifications yielded results similar to those we found in the main analysis, and in all cases prosocial unionism was a positive and significant predictor with a substantial effect size. Appendix Table A.3 provides summary results for these alternative specifications; detailed results are available upon request.

Discussion and Conclusion

Our findings are twofold. When people rate unions highly for their prosocial values, they are more likely to support unions generally. Whether the outcome is a positive attitude toward unions or approval of unions, the sentiment is stronger if they perceive unions as helping all working people instead of only their members. Workers are more likely to vote for a union in a union election if they view unions as promoting prosocial goals. Indeed, in Table 4’s voting model specification, the belief that unions act on behalf of all working people rather than just on behalf of their members had a higher impact than the pragmatic, instrumental reasons that most studies point to.

A few limitations need to be kept in mind when considering our results. The lack of a response rate that would allow us to establish more firmly the representativeness of our sample means that more-than-usual caution may be appropriate in generalizing beyond our samples. The fact that the data are from 2005 means that results may not reflect current beliefs among the general public and among workers, although we note that Kochan et al. (2019) found no diminution in workers’ interest in unionization between the 1970s and today. Finally, relying on cross-sectional data provides a limited basis for causal inference. Given that people strive for consistency, reliance on data collected at one point in time can produce an upward bias in relationships (common-method variance). Although we cannot dismiss the concerns raised by these limitations, we have addressed them to the extent possible, and we believe the data support reasonable inferences.

This study makes an empirical contribution by creating a scale representing prosocial unionism that improves on previous measures. It combines items that clearly tap into the notion of ideals of fairness and justice, it includes several such items representing a broader sampling of the construct domain, and it is reliable. Items assess agreement with statements such as “unions help the poor and disadvantaged in our society,”“[unions] are concerned about all working people,” and they ask if unions’ impact is positive or negative for the “income and benefits of workers who are not in unions.” In short, this scale appears to measure something directly relevant to questions about the social-good-creating elements of unions that workers and the public find appealing.

This study provides strong evidence that the “nobler sentiments” (Fiorito 1992: 21) fueling union support matter, and in doing so it makes an important theoretical contribution. Our finding that the general public and workers care—to a fairly large extent—about whether or not unions are engaged in universalist actions adds weight to theorists (e.g., Flanders 1970; Fantasia 1988; Kimeldorf 1991; Clawson 2003; Milkman 2013) who have argued that labor’s history of fighting for justice has been a source of unions’ appeal to the public and to potential members. Our findings highlight that this “nobler” impetus still matters, perhaps more than ever.

The very survival of the union movement may in fact hinge on it. McCartin (2017: 65) noted that “US workers will attain bargaining power only to the extent that their exercise of that power can credibly claim to be democratically accountable and in furtherance of the common good.” Others agree, pointing out that union decline partly stems from “a narrow economistic form of unionism lacking broader institutional legitimacy” (Godard 2009: 101). They may be right, and as noted above, campaigns emphasizing demands that transcend self-interest have met with some success.

In short, if unions address political and social justice goals that go beyond the workplace, their legitimacy and their success in attracting public support and members may be enhanced and help stem the tide of shrinking union density. The movement we are seeing toward a prosocial orientation hearkens back to the early years of the labor movement, when its demands also included social welfare. Then, as now, unions were instrumental in supporting democracy and in reducing inequality. Regarding the former, in addition to acting as a democratizing influence in workplaces (Godard and Frege 2013), unions increase voter turnout in elections (Lamare 2010) and encourage democratic participation in society more generally (Sojourner 2013).

Unions also are a major force in reducing levels of inequality (Rosenfeld 2014). To the extent that the labor movement can capitalize on the age-old social betterment character of unions, the greater its likelihood of success. Finally, at the political level, a point of attack by anti-union forces is the argument that unions are narrow self-interest groups and serve no public purpose. Our results show that many in the general public and many workers believe unions help build a better society, and they support unions for that reason. This outcome is a solid riposte to the “no social purpose” argument.

Footnotes

Appendix

Table A.3.

Summary of Robustness Check Results

Specification	Baseline: See Table 4 and text	Baseline using regression-based MDR	Best single item	Baseline for Form A sub-sample	Baseline for Form B sub-sample	Baseline with Gen’l Union Att IV	Baseline with expanded prosocial item pool
	Par. est.	Par. est.	Par. est.	Par. est.	Par. est.	Par. est.	Par. est.
DV = General union attitude (standardized betas)
Prosocial	0.42***	0.40***	0.17***	0.44***	0.44***		0.44***
Instrumental	0.32***	0.33***	0.27***	0.34***	0.27***		0.29***
PU-IU Corr.	0.65***	0.62***	0.24***	0.69***	0.57***		0.65***
R ²	0.61	0.61	0.31	0.66	0.57		0.61
N	1,092	1,092	1,092	547	545		1,092
Table 4 controls	Yes	Yes	Yes	Yes	Yes		Yes
Method	OLS	OLS	OLS	OLS	OLS		OLS
DV = Union approval
Prosocial	1.40***	1.72***	0.57***	1.72***	1.21***		1.70***
Instrumental	1.22***	1.37***	0.60***	1.15***	1.25***		1.12***
R ²	0.39	0.39	0.24	0.43	0.36		0.40
N	1,092	1,092	1,092	547	545		1,092
Table 4 controls	Yes	Yes	Yes	Yes	Yes		Yes
Method	Ord. Log.	Ord. Log.	Ord. Log.	Ord. Log.	Ord. Log.		Ord. Log.
DV = Pro-union voting intent
Prosocial	1.41***	1.73***	0.44***	1.57***	1.37***	0.81***	1.59***
Instrumental	0.98***	1.10***	0.63***	1.01***	0.90***	0.62***	0.92***
PU-IU Corr.	0.66***	0.65***	0.24***	0.72***	0.60***	0.65***	0.68***
Gen’l Union Att						1.50***
R ²	0.49	0.49	0.34	0.52	0.47	0.54	0.49
N	548	548	548	281	267	548	548
Table 4 controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Method	Ord. Log.	Ord. Log.	Ord. Log.	Ord. Log.	Ord. Log.	Ord. Log.	Ord. Log.

Notes: DV, Dependent Variable; Gen’l Union Att, General Union Attitude; Instrumental or IU, Instrumental Unionism; IV, Independent Variable; MDR, missing data replacement technique; OLS, Ordinary Least Squares; Ord. Log., Ordered Logistic; Par. est., Parameter Estimate; Prosocial or PU, Prosocial Unionism; PU-IU Corr., Simple correlation between Prosocial and Instrumental variables.

Acknowledgements

The authors thank Anne Barrett, Philip Roth, Carl Schmertmann, Miles Taylor, Chad Van Iddekinge, and Gang Wang for helpful comments and expertise on various issues, and Kimberly McClellan for incredible staff support. Valuable research assistance was provided by Arkin Buyukozturk, Andrew Keyes, and Nicholas Marangi. A thank you for useful feedback is also due to participants at the Labor and Employment Relations Association 70th Annual Meeting (2018), where an earlier version of this paper was presented. Finally, we are grateful to Sheldon Friedman and Guy Molyneux for making available the Hart survey data used in this study.

For information regarding the data and/or computer programs used for this study, please address correspondence to Jack Fiorito at jfiorito@fsu.edu.

ORCID iD

Jack Fiorito

References

Ajzen

Isek

. 1991. The theory of planned behavior. Organizational Behavior and Human Decision Processes 50: 179–211.

Bok

Derek C.

Dunlop

John T.

1970. Labor and the American Community. New York: Simon and Schuster.

Budd

John W.

2017. Labor Relations: Striking a Balance, 5th edition. New York: McGraw-Hill Irwin.

Clawson

Dan

. 2003. The Next Upsurge: Labor and the New Social Movements. Ithaca, NY: Cornell University Press.

Cohen

Jacob

. 1992. A power primer. Psychological Bulletin 112(1): 155–59.

Cornell Statistical Consulting Unit. n.d. Fit Statistics commonly reported for CFA and SEM. Handout available by request. Accessed at https://www.cscu.cornell.edu/news/handouts.php (December 2019).

Ellem

Bradon

Goods

Caleb

Todd

Patricia

. 2020. Rethinking power, strategy and renewal: Members and unions in crisis. British Journal of Industrial Relations 58(2): 424–46.

Fantasia

Rick

. 1988. Cultures of Solidarity: Consciousness, Action and Contemporary American Workers. Berkeley: University of California Press.

Fantasia

Rick

Voss

Kim

. 2004. Hard Work: Remaking American’s Labor Movement. Berkeley: University of California Press.

10.

Fine

Janice

Burnham

Linda

Griffith

Kati

Minsun

Narro

Victor

Pitts

Steven

. 2018. No One Size Fits All: Worker Organization, Policy, and Movement in a New Economic Age. Labor and Employment Relations Association Series. Champaign: University of Illinois Press.

11.

Fiorito

Jack

. 1992. Unionism and altruism. Labor Studies Journal 17(3): 19–34.

12.

Flanders

Allan

. 1970. Management and Unions: The Theory and Reform of Industrial Relations. London: Faber and Faber.

13.

Godard

John

. 2009. The exceptional decline of the American labor movement. Industrial and Labor Relations Review 63(1): 82–108.

14.

Godard

John

Frege

Carola

. 2013. Labor unions, alternative forms of representation, and the exercise of authority relations in U.S. workplaces. ILR Review 66(1): 142–68.

15.

Greenberg

Marian

Lewis

Penny

. 2017. The City Is the Factory: New Solidarities and Spatial Strategies in an Urban Age. Ithaca, NY: ILR Press, an imprint of Cornell University Press.

16.

Gumber

Clayton

Padavic

Irene

. 2020. Race differences in motivations for joining unions: The role of prosocial beliefs. Social Science Quarterly 101(2): 490–502.

17.

Hyman

Richard

. 2001. Understanding European Trade Unionism: Between Market, Class and Society. London: SAGE Publications.

18.

Jöreskog

Karl

Sörbom

Dag

. 1993. Structural Equation Modeling with the SIMPLIS Command Language. Hillsdale, NJ: Lawrence Erlbaum Associates.

19.

Kimeldorf

Howard

. 1991. Bringing unions back in (or why we need a new old labor history). Labor History 32: 91–103.

20.

Kline

Rex B.

2005. Principles and Practices of Structural Equation Modeling. Ithaca, NY: Cornell University Press.

21.

Kochan

Thomas A.

1979. How American workers view labor unions. Monthly Labor Review 102(4): 23–31.

22.

Kochan

Thomas A.

Yang

Duanyi

Kimball

William T.

Kelly

Erin L.

2019. Worker voice in America: Is there a gap between what workers expect and what they experience? ILR Review 72(1): 3–38.

23.

Lamare

J. Ryan

. 2010. Union influence on voter turnout: Results from three Los Angeles County elections. Industrial and Labor Relations Review 63(3): 454–70.

24.

Lopez

Steven

. 2004. Reorganizing the Rust Belt: An Inside Study of the American Labor Movement. Berkeley and Los Angeles: University of California Press.

25.

Marken

Stephanie.

2018. Still listening: The state of telephone surveys. Gallup Online Report, January 11. Accessed at https://news.gallup.com/opinion/methodology/225143/listening-state-telephone-surveys.aspx (January 2018).

26.

Martinez

Arthur D.

Fiorito

Jack

. 2009. General feelings toward unions and employers as predictors of union voting intent. Journal of Labor Research 30(2): 120–34.

27.

McCartin

Joseph A.

2017. Rejoinder. Labor: Studies in Working-Class History of the Americas 14(2): 61–67.

28.

McKersie

Robert B.

2019. A Field in Flux: Sixty Years of Industrial Relations. Ithaca, NY: Cornell University Press.

29.

McShane

Steven L.

1986. General union attitude: A construct validation. Journal of Labor Research 7(4): 403–17.

30.

Milkman

Ruth

. 2013. Back to the future? US labour in the new gilded age. British Journal of Industrial Relations 51(4): 645–65.

31.

Milkman

Ruth

Ott

(Eds.). 2014. New Labor in New York: Precarious Workers and the Future of the Labor Movement. Ithaca, NY: ILR Press, an imprint of Cornell University Press.

32.

Montgomery

Ruth

. 1989. The influence of attitudes and normative pressures on voting decisions in a union certification election. Industrial and Labor Relations Review 42(2): 262–79.

33.

Neufeld

Maurice F.

1982. The persistence of ideas in the American labor movement: The heritage of the 1830s. Industrial and Labor Relations Review 35(2): 207–20.

34.

Newman

Daniel A.

2009. Missing data techniques and low response rates: The role of systematic nonresponse parameters. In Lance

Charles E.

Vandenberg

Robert J.

(Eds.), Statistical and Methodological Myths and Urban Legends, pp. 7–36. New York: Routledge.

35.

O’Brien

Robert M.

2007. A caution regarding rules of thumb of variance inflation factors. Quality and Quantity 41: 673–90.

36.

O’Grady

Frances

. 2014. New TUC guide shows how unions improve jobs and build a fairer society. Trades Union Congress press release, September 6. Accessed at http://www.tuc.org.uk/union-issues/new-tuc-guide-shows-how-unions-improve-jobs-and-build-fairer-society (September 7, 2014).

37.

Perlman

Selig

. [1928] 1979. A Theory of the Labor Movement. New York: Macmillan. Reprint, Philadelphia: Porcupine Press.

38.

Robinson

Ian

. 1994. NAFTA, social unionism, and labour movement power in Canada and the United States. Relations Industrielles/Industrial Relations 49(4): 657–93.

39.

Rosenfeld

Jake

. 2014. What Unions No Longer Do. Cambridge, MA: Harvard University Press.

40.

Roth

Philip L.

Switzer

Fred S.

Switzer

Deborah M.

1999. Missing data in multiple item scales: A Monte Carlo analysis of missing data techniques. Organizational Research Methods 2(3): 211–32.

41.

Savage

Lydia

. 1998. Geographies of organizing: Justice for Janitors in Los Angeles. In Herod

Andrew

(Ed.), Organizing the Landscape: Geographical Perspectives on Labor Unionism, pp. 225–52. Minneapolis: University of Minnesota Press.

42.

Sojourner

Aaron J.

2013. Do unions promote members’ electoral office holding? Evidence from correlates of state legislatures’ occupational shares. ILR Review 66(2): 467–86.

43.

Tapia

Maite

. 2013. Marching to different tunes: Commitment and culture as mobilizing mechanisms of trade unions and community organizations. British Journal of Industrial Relations 51(4): 666–88.

44.

U.S. Bureau of Labor Statistics. 2006. Union members in 2005 (USDL 06-99). Accessed at https://www.bls.gov/news.release/archives/union2_01202006.pdf (July 8, 2018).

45.

U.S. Census Bureau. 2006. Table 5. Educational attainment of civilians 16 years and over in the labor force, by age, sex, race, and Hispanic origin: 2005. Accessed at https://www.census.gov/data/tables/2005/demo/educational-attainment/cps-detailed-tables.html (January 22, 2020).

46.

———. 2007. Age and sex composition in the United States: 2005. Accessed at: https://www.census.gov/data/tables/2005/demo/age-and-sex/2005-age-sex-composition.html (November 30, 2019).