What Do Americans Want from (Private) Government? Experimental Evidence Demonstrates that Americans Want Workplace Democracy

Abstract

A majority of Americans spend a substantial amount of time at work where they have little to no say over many issues—a phenomenon that philosophers have likened to a “private government” that resembles a dictatorship. Is this because Americans are indifferent to or even prefer to work for firms that resemble dictatorships? To answer this question, we field a conjoint experiment on a nationally-representative sample of Americans to isolate public preferences over “corporate regime type.” We find that Americans prefer workplace democracy. In a second experiment, we find that most Americans support workplace democracy even after being exposed to framing emphasizing democratization’s costs. The results suggest that social scientists must look beyond public opinion to understand the lack of workplace democracy in the United States. This article forges new ground by applying a political science lens to corporate governance—a field ripe with politics but bereft of political science.

INTRODUCTION

At their best, markets are democratic spaces where consumers and producers express their preferences through exit, voice, or loyalty (Hirschman 1972). However, the reality of labor exposes a democratic dilemma. Production within the modern firm is organized via dictatorial control (Coase 1937, 387). Philosopher Elizabeth Anderson (2017, 37–9) notes, “[T]he boundary of the firm is defined as the point at which markets end and authoritarian centralized planning and direction begin.” Ironically, managers organize production more like a “communist dictatorship” than a democracy. However, theories of the firm often conceptualize the relationship between workers and employers as one where both sides freely associate (Alchian and Demsetz 1972; Rajan and Zingales 1998; 2002).¹ The economic reality of labor markets under American capitalism where most individuals must work in order to survive runs afoul of the assumption that workers freely associate with businesses. Given that individuals spend substantial amounts of their time working, this poses an important puzzle: do firms fail to resemble democracies because Americans simply don’t care for democracy at work?

Understanding the answer to this question is important because elected officials are considering many alternatives to the dictatorial firm. For example in 2019, Senator Tammy Baldwin (D-WI) sponsored the Reward Work Act, which proposed that large companies reserve a third of corporate board seats for elected employees (codetermination).² Major Democratic Party presidential primary candidates, such as Elizabeth Warren and Bernie Sanders, adopted similar ideas. Other forms of democratic firms, such as co-operatives, currently exist, but they are rare.³ Other types of firms that redistribute profits through Employee Stock Ownership Programs (ESOPs) are nowhere near the norm.⁴ While many more democratic alternatives, such as ESOPs and worker co-operatives, where decisions are made through voting, the reality is that they are rare in the modern American economy. Although existing research shows that firms that are organized more democratically are no less productive or successful than their less democratic counterparts (Blasi, Freeman, and Kruse 2014; Jager, Schoefer, and Heining 2019), workers may perceive them to be more costly to work at.

Understanding a firm’s political design is important not only for unpacking how firm governance affects economic outcomes, such as wages, working conditions, and firm survival but also for how work structures and socializes individuals into politics. Political theorists have posited that even banal work experiences teach citizens about the political process (Pateman 1970). Empirical research demonstrates that variation within the workplace can explain why individuals participate in politics (Brady, Verba, and Schlozman 1995; Hertel-Fernandez 2016) and their beliefs about politics, authority, and society (Elden 1981; Mondak and Mutz 2006; Weber, Unterrainer, and Hoge 2020). Recent field experimental evidence shows that brief exposure to more participatory practices in the workplace decreased authoritarian orientations (Wu and Paluck 2020). In short, democracy at work matters not just for work, but for the health of democracy writ large.

We study whether Americans prefer more democratic workplaces. We use a conjoint experiment to randomize a large array of firm attributes.⁵ We randomly vary the presence of workplace democracy policies, such as ESOPs, codetermination, and the direct election of management by employees, thereby capturing several varieties of workplace democracy from more common profit sharing to more radical forms of worker co-operativism. Other experimental designs might suffer from compound treatment effects. Because we randomize features such as salary and benefits, factors that respondents might change if they worked in more democratic workplaces, we decouple intrinsic preferences for workplace democracy from its material benefits. Additionally, we investigate why respondents prefer workplace democracy and how they evaluate trade-offs. While other designs, like field experiments, have more external validity, our design isolates the intrinsic benefits of workplace democracy, measures additional outcomes, and benchmarks our estimates against economic features, like salary and unionization.

We summarize our experiment’s results below:

1. 1. Respondents want to work at workplace democracies: We find that respondents prefer to work at firms with workers on their corporate board, employee ownership programs, and direct election of management by 7, 10, and 5 percentage points, respectively. Thus, we demonstrate that workers’ preferences cannot explain the lack of workplace democracy in the United States.

2. 2. Workplace democracy preferences are economically meaningful: We estimate that respondents value workplace democracy equivalently to a wage increase of approximately 70 cents to $20/hour.⁶

3. 3. Respondents prefer to work at democratic firms because they intrinsically value having more workplace power: Our experimental design holds many job search factors constant, like salary, benefits, and workplace culture—features that workplace democratization likely affects. Therefore, we show that individuals value workplace democracy because they believe they would have more power over workplace decision making.

4. 4. Unlike many aspects of American political life, we find little partisan polarization around workplace democracy: Despite the spillover of partisan polarization into many areas, like hiring (Iyengar and Westwood 2015), consumption (McConnell et al. 2018), dating (Huber and Malhotra 2016), and perceptions of people (Ahler and Sood 2018; Mason 2018), we find little evidence of partisan polarization around workplace democratization. We find, however, that Republicans slightly oppose codetermination despite recognizing that codetermination would provide more workplace power.

One potential issue with our conjoint is that respondents may not internalize workplace democracy’s potential costs. Specifically, they may not realize that they need to work more. To address this, we fielded a follow-up experiment that randomly assigned cost/benefit frames in describing each workplace democracy proposal (codetermination, ESOPs, and direct election of managers). We find that a majority of respondents still support workplace democracy, want to work at workplace democracies, and believe they would have more power at workplace democracies while being told about workers’ potentially increased workload. Even when we only prime the costs, respondents overwhelmingly support workplace democracy. Thus, we conclude that support for workplace democracy is robust to purely negative information.

This project encourages scholars of politics to investigate the workplace as a site of politics. While other social sciences, like sociology and economics, study power in the workplace, mainstream political science has been mostly silent.⁷ While political scientists have worked at the intersection of business and politics, like the lobbying (Hall and Deardorff 2006) and regulation literatures (Carpenter 2010; Malhotra, Monin, and Tomz 2018), there is far less work investigating politics within the firm (Li 2018; Min and You 2019; Mondak and Mutz 2006) or preferences over types of firms (Hertel-Fernandez, Kimball, and Kochan 2019). Normative theorists and activists argue that workers should want workplace democracy because it checks the power of employers by employees (Anderson 2017; Ferreras 2017; Gourevitch 2014). By designing the experiment to separate intrinsic preferences from instrumental preferences, we show workers desire a more equitable distribution of power among employees and managers within the firm. While existing conceptualizations of corporate governance focus on the role of contracting (Alchian and Demsetz 1972) and shareholder democracy (Bebchuk 2005), both conceptualizations neglect the role of everyday workers in firm governance. Existing conceptualizations’ failure to recognize empirical facts, like labor market monopsony (Ashenfelter, Farber, and Ransom 2010; Azar, Marinescu, and Steinbaum 2022) and structural domination in the relationship between bosses and workers (Gourevitch 2018), may explain this neglect. Considering this, theorists across many traditions argue for more democratic corporate governance (Anderson 2017; Dahl 1986; Ferreras 2017; Wright 2010). We find that workers agree.

This project also contributes to and extends several literatures in the political economy of the workplace. Broadly, our project extends existing work in economics and management that focus on preferences for types of work along economic dimensions such as compensation, flexibility, and benefits (Beglo and Gorges 2018; Eriksson and Kristensen 2014; Flory, Leibbrandt, and List 2015; Freeman and Rogers 2006; Frymer 2005; Kostiuk 1990; Mas and Pallais 2017; Wiswall and Zafar 2018). Building on Freeman and Rogers (2006) and Kochan et al. (2019), we argue that workers have preferences over the workplace’s economic and political fundamentals (Anderson 2017; Dahl 1986; Eidlin and Uetricht 2018; Gourevitch 2014).⁸ This implies that studies of labor supply and job search overlook an important dimension of work. We improve over related work by Freeman and Rogers (2006), by using a conjoint experimental design where we can elicit preferences over the political arrangements of work and compare them to preferences over economic arrangements holding all else equal. While our study most closely relates to Hertel-Fernandez, Kimball, and Kochan (2019; 2022), they focus on preferences over labor union arrangements and collective bargaining. We extend this work by investigating how workplace democratization affects workers’ work preferences. Our work also complements Collom (2003) by investigating whether Americans support specific workplace democracy proposals and how they consider potential costs.

While existing political economy research focuses on the relationship between labor and business as they relate to the state (Fligstein 1993; Hall and Soskice 2001; Hertel-Fernandez 2019; Przeworski and Wallerstein 1988; Vogel 1989; Wright 2000), we join empirical (Hertel-Fernandez 2018; Li 2018) and normative scholars (Anderson 2017; Gourevitch 2014; 2018) in urging political scientists to broaden their notion of politics in the spirit of seminal political scientists like Robert Dahl and refocus political scientists’ lens to better understand the politics of ostensibly nonpolitical spaces. At its core, the study of politics is the study of power (Bachrach and Baratz 1962; Dahl 2005; Lukes 1974); as such, we show that Americans want workplace power shared more equitably.

Finally, our project establishes a stylized fact that the vast majority of Americans want democratized workplaces. There are many structural reasons that might explain why authoritarian workplaces persist, however, to the best of our knowledge, no one has explored whether the public wants workplace democracy.⁹ We have long known that the public wants political democracy, yet the place in which most of the public spends the majority of their waking hours remains authoritarian. Americans may not want workplace democracy for many reasons. First, the public may believe democratic firms would be more inefficient. Second, they may believe they would have more work. These concerns may be sufficiently large to reduce support for workplace democracy, even if workers realize they would have more power. We ask do Americans want workplace democracy despite these potential costs, and we find they do.

Workplace Democratization and its Premises: The View from the Labor Republican Tradition

In this section, we provide a brief overview of existing understandings of the relationship between owners and workers in firms. We then engage with political theorists’ justifications for workplace democracy. Finally, we distill our hypotheses around preferences for workplace democracy.

There are multiple ways to think about “the firm.” Coase (1937) conceptualizes the firm as a relationship between the owner and workers, in which, the owner directs the allocation of resources. Later theorists conceptualized the firm as a “team” of relationships for the coordination of inputs into outputs, in which, workers enter into contractual relationships with employers (Alchian and Demsetz 1972; Rajan and Zingales 1998; 2002). In short, the firm is an organization in which the owner allocates labor and resources to produce goods and services. See Appendix A of the Supplementary Material for a fuller discussion of conceptualizations of the firm.

Several empirical features of the workplace should dispel the notion that modern or historical employment relationships are or have been necessarily free. First, economists have found the concentration of employers limits worker bargaining power (Ashenfelter, Farber, and Ransom 2010; Autor et al. 2019; Azar, Marinescu, and Steinbaum 2022; Benmelech, Bergman, and Kim 2018; Dube et al. 2020; Manning 2003). Second, labor coercion, where workers are forced to provide their labor, has historically been violent and racialized (Acemoglu and Wolitzky 2011; Chwe 1990; Naidu 2010). Third, because social policy, most notably health insurance, is privatized and tied to employment (Hacker 2004), workers with employer-provided health insurance are forced to choose between their health and their freedom to leave.

These forces, consisting of rising firm power in dictating labor conditions and the subtle or pervasive nature of labor coercion in capitalist economies, suggest that most workers work in unfree conditions. In other words, as Anderson (2017) argues, most Americans spend most of their lives in dictatorships. While these ideas are often associated with Marxists theorists, they also appear in less “radical” political traditions. This critique goes far back in American intellectual history to the Jeffersonian vision of the ideal “yeoman” farmer and artisan and into the nineteenth and early twentieth century with the union and co-operative movements (Blasi, Freeman, and Kruse 2014; Gourevitch 2014).

Republican theorists conceptualize one group’s unchecked domination over another as an unfree relationship. Anderson (2017), building on labor republican theorists (Gourevitch 2014), extends this logic to employment relationships. Essentially, workers who cannot check managers and employers exist in an unfree relationship of domination. To eliminate unfreedom, republican theorists argue that the dominated must be able to constrain the dominant through electoral mechanisms, veto power, and so forth. Anderson (2017) concludes that free employment relationships require workers to have a check on managerial power.¹⁰ We discuss the conditions that have led to unfree conditions and their links to labor republican theory in more detail in Appendix A of the Supplementary Material.

We conceptualize workplace democracy as follows. Like many conceptualizations of political democracy, a workplace is “democratic” if and only if workers have a direct and meaningful say over firm decisions that affect them in a substantive way. We define “workers” as individuals who primarily supply labor to a firm. We define “substantive” as decisions that affect economic (e.g., wages, hours, etc.) and physical/mental security (e.g., not being physically harmed at work, safety from identity-based harassment, etc.). We conceptualize “a direct and meaningful say” as the ability to influence substantive decisions through a commonly understood and accepted mechanism, such as voting. There could be a variety of workplace democracies. Most obviously, a workplace in which all employees vote directly on all decisions satisfies our condition for workplace democracy, but this is likely rare because it may be too inefficient to operate. This suggests some room for delegation by having workers directly elect their managers, sit on corporate boards, or own company shares.¹¹ While we focus on direct manager elections, codetermination, and ESOPs in this article because they vary in how they delegate worker voice, they are not an exhaustive list and we encourage other scholars to study other forms in theory and practice (Ferreras 2017).

Workers could “vote with their feet” a la Hirschman (1972), but it is an insufficient means for having a “direct and meaningful say” for two reasons. First, worker exit is an indirect means insomuch as the worker who exits receives none of the direct benefits of any policy changes due to the exit. Second, exiting, as discussed above, is not uniformly applicable across firms for a variety of reasons. For instance, a salesperson may only have one major employer within commuting distance, thus limiting their ability to “vote with their feet” unless they pay substantial costs by moving to a different location. Finally, “exit” may be unavailable to many workers because they cannot afford to forgo income as they look for jobs. Thus, while worker exit may give workers some freedoms, it does not meet our standard of a “direct and meaningful say.”

A natural question is what is the relationship between labor unions and workplace democracy? Like existing political science scholarship (Greenstone 1969), we view a union as an interest group that is embedded within a workplace whose purpose is to raise the bargaining power of its members in two ways: (1) through coordinating workers’ information and interests and (2) through exercising the right to strike. While (1) may be useful in the negotiating process between workers and management, it may not be the primary means through which unions exercise power. Rather, unions bargain with management with the strike threat, which is coercive because its purpose is to change behavior by imposing costs on management.¹² By relying on coercion to exercise power, unions, we argue, are insufficient in and of themselves to be considered workplace democracies. Despite this, unions foster democracy within workplaces by raising the bargaining power of workers vis-à-vis management by putting coercive pressure on management via strikes (Gourevitch 2016), disseminating information to its members (Kim and Margalit 2017), and allying with other unions (Ahlquist and Levi 2013).¹³

Figures 1 and 2 visualize the power logic within firms in positions of domination (autocracy) or non-domination (democracy). Figure 1 represents how managers have power over workers in the modern corporation. This schematic captures the power imbalance where employers and managers have power over employees with employees having little power to respond. Alternatively, labor republicans argue that workers should be able to check the power of managers as shown in Figure 2. Many proposals for workplace democracy fall neatly within this framework. For instance, codetermination allows workers to check the domination of employers by providing workers seats on the corporate board. We also examine ESOPs and direct managerial elections. While this is not an exhaustive list of ways to democratize workplaces, they tap into basic principles of workplace democracy.

Figure 1. Autocratic Workplace

Note: Solid lines represent one person having great say over another person’s job, while dashed lines represent one person having less say over another person’s job.

Figure 2. Democratic Workplace

Note: Solid lines represent one person having greater say over another person’s job, while dashed lines represent one person having less say over another person’s job.

An open question is whether workers want workplace democracy. Workers may value having more power over decision making at work because it might translate into better salaries, working conditions, and benefits. Assuming that workers focus on these features while searching for jobs, then workers might prefer workplace democracy for material reasons. Beyond this, a core feature of labor republicanism is that power matters for checking managerial power. In short, power has intrinsic value. In addition to checking managers, workers may value having more power for other reasons, such as on procedural grounds or by increasing their own sense of agency. Thus, we argue that power itself—the ability to meaningfully alter workplace decision making whether they be material or not—is the key causal mechanism.

While labor republicans argue that workers should want workplace democracy, do workers want workplace democracy when considering the potential costs? Perhaps workers do not advocate for workplace democracy because they believe they will have more work. Importantly, we are not arguing that workers actively oppose democratization; rather, they may not prefer it given the costs. Or, as labor republicans argue, workers may value workplace decision making power enough to want workplace democracy despite the costs. We test the labor republican assumption directly.¹⁴

Thus, we distill the labor republican hypothesis as follows: if workers intrinsically value having power in the workplace (and could choose between democratic and authoritarian workplaces), then they should want to work at workplace democracies compared to authoritarian firms, after considering the costs.

Study 1

Experimental Design

In this section, we use a conjoint survey to attain a comprehensive sense for Americans’ workplace preferences. Although the “ideal” experiment would randomize real-world job advertisements, the field experimental approach cannot disentangle intrinsic preferences for workplace democracy as theorized by labor republicans from instrumental ones because we would need to randomize a large number of potentially confounding factors to differentiate intrinsic from instrumental preferences.¹⁵ Particularly, we design our experiment to test the labor republican proposition that workers want workplace democracy. Conjoint designs are a popular tool in political science (Bansak et al. 2018; Hainmueller and Hopkins 2014; Hainmueller, Hopkins, and Yamamoto 2014), and they approximate real-world behavior (Hainmueller, Hangartner, and Yamamoto 2015). This design allows us to estimate the Average Marginal Component Effect (AMCE) of an attribute relative to the baseline. In other words, we can estimate the individual contribution of a feature relative to some baseline feature averaging over all other feature combinations. As pointed out in recent work (Abramson, Koçak, and Magazinnik 2022; Abramson et al. 2020), in conjoint designs, a few respondents who strongly prefer one component can drive the AMCE. We address this by using machine-learning techniques that uncover underlying effect heterogeneity (Athey, Tibshirani, and Wager 2020). To preview, we find minimal evidence that preference intensity explains our findings.

Our experiment places respondents in the position of a job seeker weighing job offers.¹⁶ After introducing respondents to the task, we show respondents two firm profiles as displayed in Figure S8 in the Supplementary Material. After reading the firm profiles, respondents answer four outcome measures. The first question forces respondents to indicate at which firm they would rather work. We code our key dependent variable, Firm Preferred, as 1 if the respondent prefers the firm, and 0 otherwise. It approximates the real labor market in that when someone accepts a job offer, they must reject the other offer. We then use three questions to measure whether respondents believe that workplace democracies improve their day-to-day work experiences. We ask respondents which firm is “best at resolving any problems or complaints that arise at work,” at which firm “workers would have more power,” and at which firm they “would have more responsibilities.” Importantly, by measuring whether respondents believe that they have more power at one firm, we can understand whether labor republican theories of workplace freedom are driven by a desire for more equitable power distributions. These additional measures capture how respondents understand the firm is run. All measures are forced choice.

Each respondent evaluates four pairs of firms on separate pages. We set the industry of the firms to the industry in which the respondent works. We randomly vary the two firms’ profiles of 18 attributes that one would consider when choosing between job offers. The attributes include each firm’s corporate governance, firm size, sick leave policy, health insurance, work hours, race of owners, gender of owners, corporate social responsibility commitment, parental leave policy, characterization of their work, retirement fund, relocation requirements, unionization, working conditions, specialized job training programs, work from home policy, political donations, and income offer. Each of these attributes could take on multiple values. For example, corporate governance, our main variable of interest, can take on the values: “Workers sit on the corporate board,” “Workers elect their managers,” “Workers are shareholders,” “Privately owned by non-worker shareholders,” “Publicly owned by shareholders,” and “Privately owned by one individual.” The other 17 variables reduce the chance that respondents infer extraneous information about the firm that could lead to compound treatment (Dafoe, Zhang, and Caughey 2018). Additionally, we included other attributes because people on the job market might see similar information when applying for jobs. We randomly assigned the values for each attribute for each profile so that the two firms vary within and across comparisons. We also randomize the order in which attributes are listed to avoid order effects. For all analyses, we set the baseline category to private ownership by non-worker shareholders because private corporations form the vast majority of American firms and employment.¹⁷ Tables S10 and S11 in the Supplementary Material contain the full list of attribute values.

Conjoint designs sometimes require a potential trade-off between reducing informational equivalence bias (Dafoe, Zhang, and Caughey 2018) and risking survey satisficing (Krosnick 1991) from including many features. Given that we want to isolate preferences for workplace democracy beyond purely instrumental gains (e.g., salary, benefits, etc.), we included many features to isolate the mechanism of interest: whether Americans intrinsically prefer to work in workplace democracies. We believe this trade-off is reasonable. First, other research using conjoint techniques has found that respondents impute other features from observed features. For instance, Sen (2017) shows that respondents impute characteristics, such as partisanship, when only exposed to a limited number of demographic features. Thus, we wanted to reduce this risk by including other features that a respondent might impute from corporate governance. Second, a number of studies demonstrate that the risk of survey satisficing in conjoint designs is relatively low for both the number of features (Bansak et al. 2019; Jenke et al. 2021) and the number of conjoint tasks (Bansak et al. 2018). If anything, by including more features, we risk attenuating the impact of workplace democracy on job preferences toward zero, meaning our estimates are conservative.

A conjoint survey can address several weaknesses that observational and other experimental designs face. First, because every attribute value is independently randomized, we can decouple attribute values that may be highly correlated in reality. For example, worker-friendly corporate governance may be correlated with better pay, benefits, and working conditions. This is critically important for understanding whether respondents intrinsically prefer workplace democracy in addition to instrumentally preferring it. Second, we can compare the relative importance of each attribute. For example, we can compare respondents’ preference for workplace democracy against their preferences for higher income. Finally, we can investigate whether respondents prefer certain types of workplaces by interacting respondent and firm characteristics. Given partisanship affects many features of contemporary American politics, we explore whether support for workplace democracy varies by partisanship.

Sample

We fielded a nationally representative sample with YouGov Blue the week of June 29, 2020. YouGov is a high-quality sample used by academics and organizations that maintains a consistent panel of survey respondents on a wide variety of demographics. Based on power calculations and our budget, we collected a sample of 1,002 respondents weighted to be nationally representative.¹⁸ Our total sample size is 8,016 respondent-firms. We estimate all models using ordinary least squares (OLS) with standard errors clustered at the respondent level to account for within-respondent correlation in errors. For simplicity and clarity, we focus all estimates on the corporate governance conditions and report estimates with 90 and 95% confidence intervals.

Results

We argue that workers should prefer workplace democracies compared to authoritarian workplaces. In this section, we show that relative to private ownership by nonworker shareholders (private corporations), respondents prefer to work at democratic workplaces. They also recognize that democratic workplaces empower workers and better handle workplace complaints but that workers will have more responsibilities. As a manipulation check, we also find that variables, like salary, operate in the intended direction (i.e., firm preference increases with salary) thus assuaging inattentiveness and satisficing concerns (see Figure S3 in the Supplementary Material). Because these results also randomize other firm characteristics, our results provide a lower bound for democratic workplace preferences because more democratic firms might also provide better wages and benefits.

Beyond our main effects, we use machine learning methods to understand heterogeneous effects and show that while both Democrats and Republicans tend to recognize that workplace democracies would give them more power, Republicans are less likely than Democrats to prefer to work at firms with codetermination, but not ESOPs or manager-elected firms. Thus, while work preferences are politically polarized around proposals like codetermination that have received more elite attention, other democratizing proposals do not exhibit partisan polarization.

We first examine the main effects of workplace democracy treatments on respondents’ work preferences and perceptions of workplace power. Figure 3 plots the AMCE for each corporate governance proposal relative to the baseline of private ownership by nonworker shareholders for each outcome.¹⁹. We cluster all standard errors at the respondent level and report 90 and 95% confidence intervals (dark- and light-shaded regions, respectively). In the left panel, respondents prefer to work in workplace democracies compared to private corporations. Relative to the baseline, co-respondents prefer codetermination by 7 percentage points ( $ \mathrm{p}<0.001 $ ), ESOPs by 10 percentage points ( $ \mathrm{p}<0.001 $ ), and management elections by 5 percentage points ( $ \mathrm{p}=0.02 $ ). While respondents were also more likely to prefer public ownership by shareholders, the coefficients for ESOPs and codetermination are statistically distinguishable ( $ {p}_{one-sided}=0.03 $ and $ {p}_{two-sided}=0.06 $ ) from public ownership by shareholders.²⁰ While we present unweighted estimates following our pre-analysis plan, we also report substantively-similar weighted estimates in Figure S1 in the Supplementary Material. Our results suggest that workers prefer democratic workplaces, despite understanding that they may have more responsibilities. ²¹

Figure 3. Effect of Different Workplace Regimes

Notes: Estimates generated via Ordinary Least Squares with standard errors clustered at the respondent level. The dark shaded region represents the 90% confidence interval while the light shaded region represents the 95% confidence interval. An F-test of the equality of codetermination and employee ownership with public ownership for work preference yields a p-value of 0.058.

To provide a sense for the economic magnitude, we benchmark our workplace democracy results with respect to salaries. Because we also randomized salary offers in $10,000 increments, we can generate a labor supply curve based on salary. Using these estimates (see Figure S3 and Table S1 in the Supplementary Material for details), we estimate that codetermination, ESOPs, and management elections are approximately equivalent to a $70,000, $100,000, and $40,000 increases in salaries holding constant the implied utility that a respondent would get from working at a given hypothetical firm. This translates to approximately a $20/hour wage increase. This is likely an upper bound on the wage growth that individuals might be willing to forego so we caution against overinterpreting these magnitudes.²² Future work, using field experimental methods, is better suited to test for the wages that individuals might forgo for workplace democracy.²³ Nevertheless, our results suggest that workplace democracy preferences are economically meaningful. In short, respondents prefer to work in workplace democracies relative to private corporations, and the preference is substantively meaningful, suggesting that workers’ preferences do not explain the absence of workplace democracy in the United States.

Causal Mechanism: Americans Prefer to Work in Democratic Workplaces Because They Intrinsically Value Having Power over Decision Making

As we suggested before, our experimental design allows us to randomly vary many potential features that job seekers might focus on when applying for jobs. Because we randomize many features including salary, benefits, and the workplace culture of the firm, we hold constant many features that workplace democratization might affect. This generates (1) a lower bound for the preference for workplace democracy and (2) allows us to isolate why individuals support workplace democracy. Using an additional item on whether respondents believe a firm would provide them with more power relative to another one, we can investigate whether respondents prefer workplace democracy because it gives them more power over their economic lives. The top right panel of Figure 3 demonstrates respondents believe workplace democracies will give them more workplace power by about 10 percentage points across all proposals. Consistent with our proposed causal mechanism, we find that workers believe that workplace democracy gives them more workplace power. Our results suggest that respondents prefer workplace democracy beyond the potential economic benefits. They want workplace democracy to have more power over their economic lives.

We use mediation analysis and present distributions of bootstrapped ANIE and ACDE estimates in Figure 4.²⁴ Overall, we find that perceptions of increased power mediate workers’ preference for workplace democracies. We find that perceptions of increased workplace power nearly entirely mediate preferences for codetermination and management elections. We find that perceptions of power partially mediate the preference for ESOPs. While we did not hypothesize about the difference in mediation effects, we speculate that profit-sharing schemes are less directly connected to workers having power themselves. Future research should disentangle how different workplace democracies activate different motivations. In short, our results demonstrate that Americans prefer workplace democracy because they increase their workplace power.²⁵

Figure 4. Mediation Effect Decomposition via the Nonparametric Bootstrap

Notes: This figure plots the distribution of estimated AMCE, ACDE, and ANIEs via the non-parametric bootstrap with 1000 draws. We estimate the AMCE via Ordinary Lease Squares compared to the baseline condition of private ownership by non-worker shareholders. We estimate the ACDE via sequential-g estimation via the DirectEffects R package. The estimate of the ANIE comes from subtracting the bootstrapped ACDE from the bootstrapped AMCE.

Americans Understand That Workplace Democracy Requires More Responsibilities

Workplace democracy likely has costs. Workers will spend more time deliberating over how the firm ought to be managed in addition to their normal tasks. To see if workers anticipated this cost, we asked respondents which firm would give them more responsibilities. Figure 3 suggests that workers believe that, relative to the baseline condition, they would have more responsibilities under codetermination by 4 percentage points ( $ \mathrm{p}<0.001 $ ), ESOPs by 3 percentage points ( $ \mathrm{p}<0.001 $ ), and management elections by 2 percentage points ( $ \mathrm{p}<0.001 $ ). Our results suggest that workers recognize that workplace democracy can be costly.

Americans Believe That Democratized Workplaces Will Handle Complaints Better

Democratized workplaces might also handle workplace complaints better than authoritarian firms. Workers will deliberate to address their grievances, and they will be empowered to enact preferred changes. To see if workers perceived benefits associated with adjudicating complaints, we asked respondents which firm they believed would best handle complaints. Figure 3 suggests that, relative to the baseline condition, workers believed that they would better handle complaints under codetermination by 7 percentage points ( $ \mathrm{p}<0.001 $ ), ESOPs by 9 percentage points ( $ \mathrm{p}<0.001 $ ), and management elections by 3 percentage points ( $ \mathrm{p}=009 $ ). Our results suggest that workers believe that workplace democracies will solve workplace complaints better than authoritarian firms.

Partisan Polarization Does Not Govern Preferences for Workplace Democracy Yet, But It Has the Potential to Become Polarized

Partisans are polarized in almost all spaces today. Does this polarization spill over into workplace preferences? We use recently developed machine-learning methods to uncover underlying estimate heterogeneity. Particularly, we use generalized random forests (GRF) developed by Athey, Tibshirani, and Wager (2020), which is an algorithm that finds subgroups with the maximal difference between treated and control units. This algorithm iteratively partitions the covariate space to find these subgroups and uses multiple splits to generate “honest” confidence intervals that accurately reflect uncertainty from randomization and prediction. Once trained, the model can predict individual-level causal effects. In our context, we train the algorithm on all pre-treatment covariates, including age, race, education, union status, partisanship, gender, and employment status. This exercise can also uncover whether a few respondents drive our results (Abramson et al. 2020; Abramson, Koçak, and Magazinnik 2022).

Figure 5 shows our results. Each column presents a different comparison of a workplace democracy proposal against the baseline condition of private corporations, and each row presents a given outcome’s estimate. We plot the estimate of the causal effect for a given trial and rank the trials by the estimate. Each shaded bar represents the 95% confidence interval generated from the GRF algorithm. For reference, we also plot the estimated AMCE and its associated 95% confidence interval represented by the solid black vertical line and shaded region. We shade each bar by whether the respondent identifies as a Democrat (blue), Independent (white), and Republican (red). We visually identify regions of polarization when the colors separate and regions of similarity when the shades intermix.²⁶ In other words, if predicted effects for Democrats and Republicans are concentrated in different areas, then there is partisan polarization.

Figure 5. Effect of Different Workplace Regimes by Partisanship

Notes: Estimates for the treatment effect for each trial where the comparison is against private ownership by non-worker shareholders are ranked by magnitude and generated via causal forests. The thin vertical line represent the estimated causal effect of each treatment with the vertical shaded regions representing the 95% confidence interval. The horizontal shaded regions represent 95% confidence intervals of individual level treatment effect estimates.

The top row presents the estimated effects on firm preference. Figure 5 shows that individuals are not very polarized by partisanship on work preferences and workplace democracy. For ESOPs and management elected firms, Republicans are clustered around the overall AMCE whereas they are clustered at zero and below for codetermination. Manager elections seem to induce some asymmetric polarization where Democrats are clustered at the top of the ranked causal effects. We believe that elite scrutiny explains why there is more partisan polarization over codetermination compared to ESOPs and management elections. While our study is unable to answer why certain policies became polarized, future studies could use alternative designs to understand why these issues polarize.

Do Republicans not prefer codetermination because they believe that they wouldn’t have more workplace power? We find no evidence for this. The bottom set of panels in Figure 5 demonstrates that partisans across the spectrum tend to agree that workplace democracy proposals give them more power. Instead, Republicans recognize that they would have more power under codetermination, but that they might not intrinsically value having more workplace power.

Study 2

Experimental Design

In Study 2, we investigate whether making costs salient reduces Americans’ support for workplace democracy. Specifically, emphasizing that workers would have to work more under workplace democracy might reduce support for workplace democracy. We also investigate whether emphasizing that workplace democracy could empower workers increases support for workplace democracy. Additionally, we examine how covariates, like support for political democracy, class identity strength, position in the workplace hierarchy, and social dominance orientation, moderate treatment effects, in the Appendix E.6 of the Supplementary Material. This survey experiment’s goal is to investigate whether priming how workers’ workplace duties change for the better (e.g., having a voice) or the worse (e.g., having to work more) changes workplace democracy preferences.

We block randomized respondents to read one of four policy descriptions for three workplace democracies (e.g., codetermination, manager elections, and ESOPs) using respondents’ support for political democracy and social dominance orientation. First, the control condition describes the policy. Second, the cost framing condition emphasizes that workers would have more responsibilities and work. We chose to emphasize more work in the cost condition because it is the most immediate cost in a worker’s everyday life.²⁷ Third, the benefit framing condition emphasizes that workers might have more workplace voice. Finally, the cost and benefit framing condition combines both cost and benefit frames. We include the wording in Appendix E.3 of the Supplementary Material. Importantly, we randomize the experimental condition for each policy for each respondent. For example, a respondent might be in the control condition for codetermination but in the cost condition for manager elections. Additionally, we randomize the order in which respondents see policies to limit acquiescence bias.²⁸

Respondents answer five questions after reading each workplace democracy policy description. We ask respondents whether they (1) support the policy, (2) would work at the firm, (3) believe workers have more workplace power, and (4) believe workers have more responsibilities. Finally, we ask a factual policy recall question to see if respondents understood the policy. The first four questions are on seven-point scales, and we rescale them from 0 to 1 in the main text. We standardize outcome measures by subtracting by their means and dividing by their standard deviations when investigating heterogeneous treatment effects. This helps comparability across different policies.

Sample

We fielded a nationally-representative sample with Lucid on January 19, 2022. Recent work shows that Lucid respondents can be inattentive (Aronow et al. 2020), so we employed a more disguised attention check to aggressively filter out inattentive respondents. Based on our budget and power calculations, we recruited 2,105 respondents. In the main text, we visualize the outcome average for each policy by experimental condition. We use OLS, regressing the outcome on the experimental condition dummy variables with robust standard errors clustered at the respondent level. Additionally, when using multiple policies in our models, we include policy-level fixed effects.

Results

Figure 6 shows that policy frames failed to substantively erode support for workplace democracy.²⁹ Each row represents an outcome and each column represents a workplace democracy. First, we found that regardless of the treatment condition, a majority supported codetermination, manager elections, and ESOPs. Codetermination and ESOPs were the most strongly supported workplace democracies. Consistent with our theoretical expectations, the cost frame slightly reduced workplace democracy support by 3pp ( $ \mathrm{p}=0.006 $ ) while the benefit frame slightly increased workplace democracy support by 1.9pp ( $ \mathrm{p}=0.08 $ ) in our pooled estimates.³⁰

Figure 6. Framing Effects on Different Workplace Democracy Outcomes

Note: There are 2,105 Lucid respondents. We fielded the survey from January 19th to February 11th, 2022. We bootstrapped our means with 5,000 draws and show 95% confidence intervals.

We found similar patterns in respondents’ work preferences. Respondents preferred to work in workplace democracies regardless of the frame. Workers liked firms with codetermination and ESOPs about equally and liked firms with manager elections the least among the three corporate governance structures. We also found that the cost frame reduced respondents’ preference to work for workplace democracies by 3.1pp ( $ \mathrm{p}=0.002 $ ) and the benefit frame increased respondents’ preference to work for workplace democracies by 1.5pp ( $ \mathrm{p}=0.12 $ ).

A majority believed that workplace democracies would empower workers regardless of the frame. Respondents believed that codetermination gave workers the most power, followed by elections and ESOPs. Similar to before, we found that the cost frame reduced respondents’ perception that workplace democracies empowered workers by 0.6pp ( $ \mathrm{p}=0.41 $ ), and the benefit frame increased respondents’ perception that workplace democracies empowered workers by 2.2pp ( $ \mathrm{p}=0.002 $ ).

Finally, a majority understood that workplace democracy would entail more work. Respondents believed that codetermination and manager elections would place similar burdens on workers whereas ESOPs would be relatively less burdensome. The pattern was consistent with our theoretical expectations. The cost frame increased respondents’ perception that workplace democracies would require more work by 1.7pp ( $ \mathrm{p}=0.02 $ ) and the benefit frame decreased respondents’ perception that workplace democracies would require more work by 0.9pp ( $ \mathrm{p}=0.22 $ ). In sum, framing affected workplace democracy outcomes, but respondents overwhelmingly wanted workplace democracy, while recognizing the potential burdens.

Conclusion

If Americans overwhelmingly value political democracy, do they value workplace democracy? Using a conjoint experiment, we show that Americans want workplace democracy because they intrinsically value having power over their economic lives. In economic terms, workplace democracy proposals such as codetermination, ESOPs, and management elections are economically valuable. Although we find little partisan polarization in ESOPs and management elections, Republicans do not want to work at firms with codetermination (despite recognizing that they empower workers) suggesting partisan polarization may creep into workplace democracy preferences.

We recognize some important limitations of our study. Do our results map onto “real world” behavior? While we show Americans prefer workplace democracy, features of the job search might make democratized firms scarce. Although conjoint studies replicate “real world” behavior (Hainmueller, Hangartner, and Yamamoto 2015), we do not observe job search behavior as it relates to workplace democracy in equilibrium. Put simply, we observe workers’ preferences without employer constraints. Having established that workers value workplace democracy beyond economic benefits, we recommend field experiments to detect whether such preferences exist “in the wild.”

We also note that this manuscript cannot understand Americans’ full preferences over all workplace democracies. Rather, this study was designed to understand whether Americans prefer workplace democracy relative to other corporate forms. And although we tap into a few forms of workplace democracies, there is variation within those forms and interactions between those forms. For instance, certain forms of workplace democracy may only be democracies in name if they are co-opted by nonworkers. This raises a broader question of what is the “optimal design” of a democratic workplace? Thus, we believe that political scientists—as specialists in the study of power—are uniquely positioned to study workplace democracy’s “optimal design” and that this area deserves further research.

Future work should also understand why people want workplace democracy. We provide evidence that the public’s desire for more workplace power might explain workplace democracy support, but our conjoint study constructs an artificial environment. Additionally, due to the study’s nature, we could not easily ask why respondents preferred each firm or why they felt a lack of power. We recommend that scholars employ qualitative methods to understand how workers’ experiences explain workplace preferences.

We also encourage others to examine what predicts support for workplace democracy. We show that Republicans are less likely to support codetermination but that there is little partisan polarization over other workplace democracies. Given the lack of polarization, why do people support these governance structures? Having shown that desire for more power explains firm choice, future research should investigate whether psychological constructs related to hierarchy, like authoritarianism, explain support for workplace democracy. For instance, we found support for unionization is strongly polarized by partisanship between Democrats and Republicans likely because labor unions are already a politicized issue. Future work might better understand whether other workplace democracies are susceptible to partisan polarization.

Our results provide progress on an important puzzle in American political economy: why don’t we observe more democratic firms? We show Americans’ work preferences cannot explain the prevalence of authoritarian firms. In fact, we show their preferences are economically meaningful, even when accounting for workers’ costs. Thus, our results imply that other features of the political economy must be at play. These explanations could be discrimination in financial markets, owners’ desires to maintain power and the historical political repression of groups advocating for workplace democracy (Gourevitch 2014). Future work should explore what impedes workplace democratization despite worker support.

This article also raises a potential tension between preferences over labor unions and workplace democracy.³¹ While Hertel-Fernandez, Kimball, and Kochan (2019) find that workers prefer unions with collective bargaining strength, they also show that workers prefer unions that are unavailable under U.S. labor law. We also find that after accounting for many union benefits, such as better salaries, benefits, and working conditions, Americans do not prefer labor unions. Simultaneously, we show that material benefits alone cannot explain why American prefer workplace democracy. In an era of historically low private-sector union density, we show that unions might benefit from advocating for workplace democracy.³² As recently as 2014, the United Auto Workers campaigned on German-style workplace democracy in a Volkswagen plant in Tennessee.³³ Although they lost, our results suggest that advocating for workplace democracy and material benefits might reinvigorate the union movement. Workplace democracy could allow workers to prevent their bosses’ abuse (Wright 2010). Union support for economic democratization is essential.

Finally, this project sets up future work that examines the politics of corporate governance. For instance, where do these preferences come from? Why do some people prefer workplace democracy? Are workplace democracy reforms compatible with labor union organizing and systemic change? Once instituted, how do workplace democracies participate in politics relative to less democratic firms? These are just a few of the many questions that we hope our article will inspire.