Taking Voice Seriously

Abstract

Fifty years after the publication of Exit, Voice, and Loyalty, Albert Hirschman’s work is receiving renewed attention in leading journals, including efforts to develop formal models and new strands of democratic theory that capture his insights. These efforts have led some to conclude that power flows exclusively from an ability to exit at low cost, and that we can only empower the weaker elements in society by giving them more choices—more ability to switch from one public goods supplier to another. I examine work by leading proponents of these views and argue that their approaches ignore the role that constrained exit opportunities play in fomenting collective action and voice, not merely to signal dissatisfaction, but to provide information on where leaders have gone too far and what needs to be done. We should expect to see leaders respond to member dissatisfaction both when members can cheaply leave and when they have no way out, with the problematic zone lying in between—exactly where “voucher” systems for public goods propose to take us.

The year 2020 marked the fiftieth anniversary of the publication of Albert Hirschman’s Exit, Voice, and Loyalty (1970). The book has had an enduring impact on the fields of political science and political economy, cited over 30,000 times according to Google Scholar. About 10,000 of those citations have been recorded since 2015, suggesting that the impact of the book is only growing with time. One of Hirschman’s goals in writing the book was to convince economists that “voice” could be a helpful complement to the market-mechanisms they emphasized. Early in the book (16), the author quotes Milton Friedman to provide an example of his home discipline’s neglect of voice. Friedman bemoaned the “cumbrous political channels” parents needed to navigate to influence education policy through political action and suggested that a voucher system would allow parents to “express their views about schools directly by withdrawing their children from one school and sending them to another.” Hirschman points out that “a person less well trained in economics might naively suggest that the direct way of expressing views is to express them!” The author hoped his book would convince his economist colleagues to recognize that “voice” offered an alternative to “exit” that could under some conditions be more effective in stimulating firms, organizations, and states to be responsive to their consumer/member/citizens, and he called on readers to value “the digging, the use, and hopefully the slow improvement” of the cumbrous channels Friedman bemoaned.

Hirschman also wanted to convince political scientists to recognize how “exit” represented a second channel, outside of voting and other political processes, through which individuals could express their views. What is striking, fifty years after publication of the article, is how political scientists have so completely taken on board this suggestion that many have fallen into the same trap Hirschman saw bedeviling the work of Milton Friedman. We too have given markets primacy over politics in our models and have forgotten Hirschman’s call on us to value “voice” as the more direct mechanism through which citizens influence states.

In this reflection, I review recent applications of Hirschman’s ideas in political science to see how we reached the point where the most prominent efforts to model his approach conclude that exit is the ultimate source of influence in politics, even when citizens employ “voice.” Without the capacity to exit, citizens have no power. The more attractive their exit options, the greater their influence. Through a close reading of two prominent works, one offering a formal model and the other employing normative political theory, I argue that both ignore a separate process through which limitations on exit empower individuals by stimulating collective action. I agree that exit empowers citizens when it is cheap, but I propose that citizens are also empowered when exit is costly. Only when we view these two processes together can we see that the problematic zone lies in between.

The focus of my analysis is two recent articles published in leading journals: the game theoretic analysis of exit, voice, and loyalty (EVL) developed by William Roberts Clark, Matt Golder, and Sona N. Golder (2017) and the work titled “Voting with Your Feet: Exit-Based Empowerment in Democratic Theory,” by political theorist Mark E. Warren (2011). But before turning to these works, let us review how Hirschman himself laid out his ideas.

Exit, Voice, and Loyalty in Hirschman’s Original Text

Hirschman argued that economists, with their focus on markets, and political scientists, with their focus on politics, were each missing an important part of the story in their domains. Political scientists paid close attention to how individuals responded to discontent through campaigns, lobbying, and protests but saw the only alternative to voice as acquiescence or indifference. On the contrary, he pointed out, union members faced with poorly performing leaders had the option of leaving the union. Those unhappy with public schools could take their children out of the public schools and put them in private school. He referred to these opportunities for individuals to improve their situation through individual, uncoordinated action, as “exit.”

Similarly, he argued, economists paid attention to how economic actors motivated a response through their separate and cumulative decisions to stop buying from one firm and turn to new suppliers, but assumed the only alternative to abandoning a supplier was to be “dumbly faithful” (Hirschman 1970, 31). On the contrary, he argued, consumers concerned that automakers were selling dangerous vehicles had the option of joining together to press management to build safer cars.

Hirschman boldly offered to address both disciplines’ blind spots by offering his EVL framework. An individual faced with a decline in the quality of products offered by a firm, services offered by a membership organization, or policies offered by a government had two distinct options: she could exit, or she could remain in the organization and use voice to call for improvements. His third factor, “loyalty,” was a characteristic of some individuals that made them remain in the organization using voice to call for improvements even when they had the capacity to exit.

His starting point was to argue that one could get responsive organizations that were primarily attuned to exit or to voice. An organization operating in a competitive market could do well by focusing on exit. It could closely monitor sales, investigate any decline, and respond quickly by offering product improvements. Or it could be structured to respond to voice, especially if it was a “monopoly” that provided few exit options, since in this case “members” would be forced to develop channels for communicating with management when problems arose. Either one of these organizational structures, however, was likely to struggle if the market in which it operated was between these extremes. If markets were imperfect (three automakers all producing unsafe vehicles), exit would do little to force organizations set up to monitor sales to respond, even if individual consumers and Ralph Nader started making noise. Similarly, public schools designed to respond to the demands of citizens were unlikely to respond if parents started taking their children out of the public schools. Officials might even welcome the quieter life they could lead once complainers moved their kids into private schools. The intermediate zone between monopoly and free markets was particularly problematic, he argued, because of the tendency of exit and voice to be related to each other in “hydraulic” fashion. If pressure was released through exit, there would be less steam to generate voice (Hirschman 1993, 176).

Hirschman did not offer a formal model to explicate how exactly the costs of exit and voice and the level of member loyalty were related to organizational response, and for this reason his framework has long been criticized for being difficult to operationalize. The “loyalty” factor was subject to particular criticism, with Barry (1974, 95) calling it an “ad hoc equation-filler” that Hirschman was using to account for almost any outcome. He also objected to Hirschman’s implication that individuals faced with a declining organization had to choose between exit and voice. He pointed out that individuals were making two separate decisions: exit or don’t exit; and voice or don’t voice (91-92). This meant individuals had two choices besides silent exit or non-exiting voice. They could use voice while exiting, or they could remain silently inside the organization even as it declined. These challenges did not prevent EVL from influencing a variety of social sciences, including sociology and psychology, but they led to a divergence in the ways it was applied. ¹

Formal Models of EVL in Political Science

So how is it being applied in political science? The first political scientist to offer a formalization was Scott Gehlbach in the journal Rationality and Society (2006). He was followed by Clark, Golder, and Golder (henceforth CGG) who offered their own formalization in a British Journal of Political Science article in 2017. Both noted that they were the only social scientists to offer formal models of Hirschman’s framework, and I am aware of only one other (Slapin 2009) that has been published in the intervening years. Both Gehlbach and CGG develop models based on the assumption that politics is a zero-sum game in which the government wants as big a portion of the pie as it can get and citizens are trying to keep as much of the pie as they can, and both come to the same conclusion: without the ability to exit, voice has no power. In both models, it is as if Hirschman’s insights about how limited exit can motivate reform via the voice mechanism has disappeared. Exit and voice only drive reform when exit costs are low enough to allow citizens to escape from the effects of the government’s policy through their own actions. When exit costs are this low, citizens can leave or use voice to make demands backed by a threat to leave, in either case getting their way. When exit costs are high, the government has no incentive to listen.

Here is how CGG get to this conclusion. First the government attempts to take something from citizens (to offer policy that requires them to pay more taxes or receive fewer benefits). Then citizens have the option of exiting, using voice, or doing nothing—which they describe as “loyalty.” This is their nod to Barry’s critique, as they accept his argument that individuals have the option of choosing to remain silent if the costs of exiting and using voice are both higher than the cost of accepting the loss imposed by the government. ² Citizens who have cheap exit options will exit as soon as the government attempts to take something away since it is assumed there will always be some positive cost to using voice. But if exit too comes at a cost, citizens will have to weigh the cost of exit and the cost and expected efficacy of voice against the value taken away by the government, choosing exit or voice only when they anticipate the government will respond and return what it has taken. The government will only concede if its need for the citizen to stay inside the organization is greater than the cost of conceding, and the citizen’s cost of exiting is less than the benefit the citizen is seeking. In that narrow range, the citizen does not even need to use voice. The knowledge that he could leave forces the government to respond and give up its attempt to take something from this citizen.

The example CGG use to illustrate their logic is the story of how the Wallenberg family, which controlled 40% of the Swedish stock market, got its way in Swedish politics. The family has not needed to use voice to avoid high taxation on returns from invested capital. The government needed the Wallenbergs to keep their capital in Sweden and continue to invest in new projects. As a result, even when cabinets have been dominated by the Social Democrats, the government kept taxes on invested capital low because it knew that if it raised them, the family could take this capital abroad, leaving the government worse off than it was before it attempted to tax the family’s wealth (729).

The authors recognize that the point that capital has long had an advantage in contests over economic policy because it is mobile and can threaten to leave is one that has been made many times in studies of political economy (e.g. Bates and Lien 1985; Winter 1996), but they suggest that their formal model helps elucidate why exactly this is the case, highlighting how labor’s weakness stems directly from its inability to relocate when the government imposes taxes on work income.

While the article offers a contribution in explicating how exactly low exit costs contribute to the power of political actors with attractive exit options, it goes too far in claiming that their model “captures the key elements common to most political situations” (719) and “voice is only powerful when an individual has the power to exit” (741). These statements are true given the assumptions of their model, but to claim that these statements apply to “most political situations” is to overlook the capacity of voice, motivated by limited exit opportunities, to influence political outcomes.

To prepare us to focus on this missing element, I have shown in figure 1 the relationship between exit costs and the responsiveness of the government to citizens’ use of exit and voice, as described in the CGG model. Their model argues that there is only a narrow range of exit costs (close to zero) wherein some citizens can use exit or voice to force the government to address their concerns. The government will give citizens all of what they want only up to the point where exit costs are greater than the value of the concession sought from the government. The figure depicts the relationship between exit costs and responsiveness, assuming a modest and fixed cost of voice. It also assumes that the other condition required by the model (the government depends on the citizen and needs him to remain loyal) is also in effect. Even with these advantageous conditions, beyond the level shown the citizen cannot expect his exit or voice to have any effect, and so he will remain silent and loyal.

Figure 1 Likelihood of exit-driven reform as a function of exit costs

Note: Assumes modest, fixed cost of voice and fixed value of concession sought from the government.

Does this model truly capture “most political situations”? This statement may be true if we consider politics to be composed exclusively of situations in which contestants fight over their share of a fixed-size pie and only the power to impose costs on one’s counterpart can budge them. But there is a great deal of politics, especially in states have been organized so that the government needs to maintain the support of the governed, where voice can make a difference even when (especially when) exit opportunities are not available.

The CGG model attempts to capture some of the situations in which democratic governments respond to citizens by framing electoral and coalition politics as exercises in exit. Democratic governments seeking to stay in power need to worry about: 1) losing a bloc of voters that is big enough to cause it to lose election; 2) losing a party that is a pivotal member of its governing coalition; and 3) losing the support of key interest groups that fund and support the governing party. CGG argue that they capture these elements of electoral politics and government formation with their model of exit-driven politics. If groups are big enough that the government cannot afford to lose their loyalty and the groups have attractive exit options (other parties they can support), the government will do exactly what they say because their exit threats are credible.

But how do these groups form and become large enough to have that leverage? CGG recognize that many of these groups have to overcome collective action problems to become large enough to influence the government via the exit mechanism, but they are comfortable with putting these concerns aside because “our primary focus here is on understanding the power relationship between citizens and governments when collective action problems either do not exist or when they have already been solved” (CGG 2017, 724). Gehlbach (2006, 400), in his similar formal model of exit and voice deals with collective action problems in the same way, treating the question of whether they are solved as an exogenous process that takes place at the first step in his game. If voice is “established” because it makes sense to the relevant group of citizens to invest what it takes to overcome collective action problems, he argues, the key determinant of power is the group’s ability to threaten exit.

I propose that exit, voice, loyalty models that leave out or bracket the process through which collective action problems are solved and voice is “established” are missing the heart of what politics is all about—the part Hirschman was describing when he wrote that voice was an art that could atrophy if exit was too easy. Ever since Olson (1965) identified the “free rider problem” and the role it played in undermining collective action, we have been aware that the ability of a group to organize to pursue common interests is undermined when individuals who choose not to join in providing collective goods are nevertheless able to enjoy the benefits provided by the collective efforts of other players (public goods). If everyone acts on this logic, public goods will not be provided at all.

Of course, in the real world, collective action does take place and public goods are provided. Protests bring down governments; voters vote; and the National Rifle Association fights for gun rights that are enjoyed even by Americans who are not members. Because of the centrality of this issue to politics, political scientists have produced many formal models and diverse strands of literature explaining when and how it emerges. Early work, including by Olson himself, focused on how groups, once formed, could sustain themselves by providing selective incentives to those who contributed (e.g., discounts for NRA members); and on the ability of small groups of individuals, interacting repeatedly in tit-for-tat games, to settle into cooperative patterns. These may be the situations CGG and Gehlbach were thinking of when writing about how things work once voice is “established” and collective action problems “solved.”

But how do groups composed of larger numbers of individuals organize in the first place, especially when the costs of joining in a collective endeavor are very high, or when the nature of the game makes it difficult to provide selective incentives or compel participation? Early in the evolution of this literature, cooperation in such cases was considered a “paradox” since it was impossible to get to collective action under these conditions in a single equilibrium game. But scholars have since that time identified several distinct paths leading to collective action in situations where players face the following incentive structure: if they can achieve high levels of participation, they are sure to deliver the collective good at little to no cost to individuals; but if they fail to achieve this threshold of participation they will not deliver the collective good and will instead pay the cost of trying but failing. A game with this structure has multiple equilibria, including the extreme cases of no participation and high participation, depending on how things “tip” (Schelling 1978) based on the expectations the participants hold about how others in the game will behave.

Two of the mechanisms that can drive the outcome toward high participation are identified in Luis Medina’s review of recent literature on collective action problems. First, he points to the ability of “shocks” to trigger all members of the group to suddenly shift their expectations about how others will act. If this shock worsens many individuals’ positions and creates a situation where the benefits of cooperation are suddenly much higher, it will “[push] the system into the region of increasing returns [that] will result in an upward spiral toward a higher level of participation” (Medina 2013, 166). Second, social networks can nudge individuals toward participation by embedding the question of whether to participate in a discrete new collective endeavor in a broader social context of diffuse ties among individuals who regularly and routinely interact. In this context, individuals at key nodes can make participation a “focal point” (Schelling 1978) and tip the entire group toward joining together (Medina 2013, 180).

Medina does not invoke Hirschman or “exit costs” as he lays out these two mechanisms in his discussion of the tipping model of collective action, but if we look closely at these arguments it becomes clear that both operate best when individuals are at the end of the exit cost spectrum where they have limited alternative options or “no way out.” ³ Let’s look first at his concept of “shocks” that tip things suddenly toward higher levels of collective action. The example Medina uses to motivate his analysis of these dynamics is the story of a dictator wondering if it is safe to raise water fees by a factor of five. The dictator has managed to keep the opposition under control, but unrest is simmering, and he seeks advice on whether a move of this type can be carried out without provoking massive protests that would bring down his government (Medina 2007, 3). Medina agrees with the “common sense theory” that such a move would surely increase the risk of such an uprising, and devotes much of the remainder of the book to developing the formal theory that explains why actions such as this one serve to tip society from quiescence to unrest. It is not a coincidence that he chose as his example an increase in the price of water, a necessity of life. Citizens in this story do not have any way to escape, through uncoordinated individual action, a sudden deterioration in their quality of life. They have to consume water, and they can’t move from one place to another to find something similar that is cheaper. So a big part of what is driving the “tipping” here is the inability to exit that causes first a few and then a torrent of individuals to join in collective action to use voice to overthrow the regime and survive. Recent work showing that protests have often followed when regimes raise gasoline prices, via tax increases or subsidy reductions, fits this model of collective action driven by no way out, since gasoline is a commodity, much like water, which individuals in modern societies cannot live without. ⁴

The other mechanism highlighted by Medina in his review of the literature on collective action is social networks. Recent work on turnout by voters has placed a heavy emphasis on the role of mobilization efforts, relying on personal connections, in nudging individuals toward taking the time to vote even when there is little chance their single vote will swing the result (Rosenstone and Hansen 1993; Huckfeldt and Sprague 1995; Fowler 2005; Abrams, Iversen, and Soskice 2011). A rich, related literature emphasizes the role social networks and social capital play in facilitating collective action (Granovetter 1973; Coleman 1988; Putnam 2000; Aldrich 2008; Siegel 2009; Campbell 2013). Again, while these literatures rarely cite Hirschman or discuss “exit costs,” their models rely on personal connections built up over time in societies where people interact repeatedly in neighborhoods and social organizations (such as bowling leagues). Such networks only arise when people are slow to exit from neighborhoods and groups. These networks deteriorate when a society sees individuals routinely and frequently move in and out of such organizations. Social structures that make exiting from neighborhoods, workplaces, and groups costly through cultural norms or laws and regulations that make exit costly will be more likely to produce an environment of dense social ties that facilitates collective action.

Both of these mechanisms that tip groups of citizens toward collective action, then, work to foment voice in circumstances where exit is costly or impossible. Contrary to the claim by Barry (1974, 88), who accused Hirschman of engaging in “teleological functionalism” for assuming that voice would emerge merely because it was needed, we now have models that explain why rational self-interested individuals will be more likely to engage in collective action precisely in those situations where they have no or few exit opportunities.

Skeptics may still be wondering why the self-interested government with a monopoly would respond in this situation. If we are operating under the CGG assumptions of a zero sum game with perfect information, we do need to go back to the model’s emphasis on exit opportunities to understand how, once this collective action has emerged, this group can threaten the regime with withholding its loyalty (via rebellion, desertion at election time, etc.) in order to force it to respond.

If we relax some of the assumptions of the CGG model, however, we have additional channels through which voice motivated by exit-constraints can alter policy—without necessarily requiring the group to fully mobilize to the point where it can topple the government. First, let’s relax the perfect information assumption. Indeed, it is highly unlikely that governments know better than their citizens how they are affected by new policies the government is putting into place, as illustrated by Medina’s story about a dictator who sought advice on whether raising water prices by five times would trigger a rebellion. A rational regime trying to learn how high it can raise water prices does not need to wait until the peasantry has mobilized and is ready to topple the government to learn how far it can go. In a world of imperfect information, it will test the waters by first doubling and then tripling prices, watching to see at what point the peasantry begins to organize, stopping or pulling back when it sees the earliest signs that it is pushing too far. What is critical here is the ability of nascent collective action, triggered as discussed earlier by citizens who have no way out, to provide the regime with information that compels it to pull back and reform the policy.

Governments are also likely to lack information about Medina’s second trigger for collective action: social capital. If a group of citizens has never mobilized before against a government policy, the regime may not realize that this set of individuals has dense social ties that can be leveraged to organize a rebellion until it begins to mobilize. Peasants have rarely been powerful enough to overthrow regimes, but by engaging in brief rebellions they have provided regimes with information about the group’s ability to organize that motivated caution and even caused rulers to reduce the tax burden on such communities (Steele, Paik, and Tanaka 2017). If voice can alter outcomes by providing information in autocratic regimes like these, how much more might it affect outcomes in societies that are set up to make it possible for citizens to signal their preferences without resorting to rebellion?

Most contemporary regimes are not zero-sum games between an autocrat and the citizenry. So let’s see what happens when we relax that assumption and consider how voice works in regimes where the citizenry and rulers have some common interest. A useful place to start is with the model developed by Ahmed and Stasavage (2020), who recognize that even exploitative regimes that simply want to extract the maximum sum of revenue from the citizenry have an interest in not raising taxes to the point where peasants are eating the seed corn and large numbers of firms are going bankrupt. Such regimes do not play a one-shot game but are instead seeking to extract revenue in the future as well as today. In a world of imperfect information and a common interest in sustaining the economy so that it can be milked in the future, regimes have an incentive to create institutions that will provide it with information on how the citizenry are faring—to lower the cost of voice. Ahmed and Stasavage find in their 2020 study of agricultural societies that where it was particularly difficult for regimes to determine the maximum tax rate the peasantry could bear (for example because the productivity of the soil varied by terrain), rulers were more likely to establish local and national councils to provide them with advice to guide their rule. The aim here was to create an environment where the regime did not need to wait until crops failed for lack of seed or the peasantry had formed an insurgency to realize it had raised taxes too high. Council representatives closer to the people would see when peasants in some areas were being pressed to the point where they had no way out and would communicate this information to the ruler.

In contemporary democratic regimes where the government relies on voter support in elections, institutions are designed to make it even easier for the government to know how the citizenry are faring. And in these regimes, there is even more of a common interest between the ruler and the governed. Rulers are not simply trying to milk the economy but want to help the economy grow so that citizens will reelect the government. But even with the costs of voice lowered to facilitate the flow of information and the game structured to be positive sum instead of zero sum, the relationship between exit costs and the effectiveness of voice retains the same basic structure, shown in figure 2. Note that I have placed the voice-driven reform function on the same “exit cost continuum” as in figure 1, to enable us to think about how the two functions fit together in further discussion below.

Figure 2 Likelihood of voice-driven reform as a function of exit costs

Note: Assumes modest, fixed cost of voice and fixed value of concession sought from the government)

Where citizens have exit options at a modest cost (for example, they are able to send their children to private school if public schools are failing), they are less likely to organize or make the effort to collectively communicate their concerns to the local government. In the absence of exit opportunities, in contrast, we expect in the model of collective action presented here that citizens will be more likely to start organizing and communicating as exit costs rise, with their propensity to do so rising sharply at the tipping point when they look around and see others expressing concerns and realize that if everyone who shares their concerns joins together, they have the potential to change policy. Under more democratic institutions, the government will be watching for evidence that citizens are upset, so even nascent movements toward collective action and signals that a group has enough social capital to organize will give the government the information that compels it to change course. Citizens do not need to coalesce into a formally organized political party or interest group that is large enough to bring down the government.

Clark, Golder, and Golder miss all of this because the assumptions in their model (perfect information, zero sum game) lead them to conclude that voice can do no more than demonstrate the ability of a group to threaten the ruler’s hold on power. Voice in their model loses an essential feature of this mode of influence, which is its ability to communicate information, updating the government’s understanding of how policy is affecting various constituencies, showing that a group has the social capital to organize if pressed, and even reframing the way the other side thinks about the game they are playing to emphasize how the government would benefit by listening to the concerns they are expressing. Because the CGG model treats collective action as exogenous, there is nothing within their game that citizens can do to affect the government’s response by using voice to signal their needs and desires or ability to organize.

Exit, Voice, and Democratic Theory

Before putting the two functions together, however, I wanted to turn to the second prominent recent political science article that has applied the EVL framework—this time to democratic theory. Mark Warren (2011) is motivated to think about the role of exit in democratic theory because he feels that most political theorists spend all of their time on how to improve the functioning of the voice mechanism. They operate in the world in which the government is the monopoly provider of a public good (public education or policing, for example) and they think about how to organize the decision-making process about what kind of policy to provide so that the voice mechanism functions as democratically as possible. Warren references in particular the burgeoning literature on deliberative democracy, which has focused on how to make sure everyone is heard.

In his review of the way political theory treats “exit,” he finds that the concept gets little positive attention. When theorists address it, they write about how it “damages collective loyalties, undermines provision of common goods, encourages excessive individualism, drains collectivities of their best and brightest, and substitutes silence for voice” (684). He argues, against this conventional wisdom, that the ability to exit is actually critical to empowering individuals. “Exit can often induce representatives or other elites to engage deliberatively those who can exit, and exit induces elites to cultivate reputations for trustworthiness” (684).

As this summary statement suggests, Warren’s argument echoes CGG in valuing exit for the role it plays in empowering citizens. He grants that at a smaller scale (the town assembly, where everyone is given a chance to speak) the “voice-monopoly” model can come close to democratic ideals, but argues that “as scale and size increase, so too do the thresholds for effective voice. The effects are not equally distributed: those with greater political resources—education, organization, attentiveness, and time—are more likely to surpass these thresholds than those without” (688). Mining a deep vein in the public choice literature (e.g., Chubb and Moe 1990), Warren argues that it is even more difficult for ordinary citizens to be heard when the implementation of policy is delegated to bureaucrats, who respond most attentively to the well-organized groups they deal with (professional associations and unions). In this world, citizens find that “the alternative to voice for those with few resources is not exit, but silence” (688).

Warren thus follows Barry 1974 and CGG in recognizing that individuals are faced with two distinct choices (exit or don’t exit; and voice or don’t voice) and have the option of accepting the status quo and remaining silent. He thus disagrees with Hirschman’s optimism that constraining exit will generate voice, worrying instead that it will simply force many ordinary citizens to quietly make do with whatever public services and public goods are provided by the government.

Instead of constraining exit, as Hirschman recommended, Warren calls on government to expand exit opportunities in order to bring us closer to democratic ideals. He agrees with CGG that a citizen’s power depends on his or her exit options, but he diverges in imagining how these exit opportunities might be structured to empower ordinary citizens. Recall that in CGG’s model, it is the rich—especially citizens who hold mobile assets—who are the most powerful because they have wealth they can move out of the control of the government. In the contemporary United States, the rich almost always have opportunities to opt out of public goods to get better service: they can choose private education when the local public schools are failing; private vehicles and Uber when public transit is slow; private pools when the public pools are run down; private police in their subdivisions when the local police force is stretched thin; private housing when public housing is run down.

Warren asks us to reimagine how exit might work if every citizen, and not just the elite, was given the ability to choose an alternative public service provider when faced with inadequate public goods and services. For this empowerment to serve truly democratic ends, he notes, the choice must be “enabled and institutionalized.” His model is the voucher system for education. Give every parent a voucher that they can use to send their children to a private school if the local public school is not good enough. Instead of organizing health care so that there is a single provider, give citizens a voucher they can use to obtain health services or insurance coverage from a provider of their choice. The public choice literature has made the case for expanded use of vouchers for housing and social welfare services on a similar rationale (Dowding and John 2012). Warren does not specify how we would do this with policing, parks, or public transportation, but in similar arguments in favor of greater choice, Robert Taylor (2017) suggests that this could be accomplished by ensuring that citizens have a choice of local governments that compete to offer better services at lower tax rates. Citizens could be “resourced” and provided with information so that the ability to choose one’s municipality would be available not only to the rich but also to poorer residents. Warren (and Taylor) thus advocate for a utopia of choice in which citizens do not need to lobby their government for better services. They simply need to “signal” the government by abandoning less preferred providers for better providers and tap market forces to drive improvements across the board. Once citizens are empowered by their exit opportunities, public service providers will have an incentive to reach out to them to learn what they want—creating the deliberative democracy that democratic theorists seek.

Warren’s emphasis on the need for exit to be enabled and institutionalized is on target. In the absence of policies designed to enable poor and needy citizens to have the same choices that are available to rich individuals, he recognizes, markets serve only those with means. My first problem with his analysis, however, is that he is too quick to assume that “enabled” exit options will force providers of social services to improve their performance. If you accept the logic CGG offer on how exit works, we should not be confident that an individual decision to opt out of one government-funded healthcare provider and choose another is going to force that provider to improve. Remember that CGG assume that the government will only give up on efforts to seize a larger share of the pie if it depends on the citizen’s staying inside the organization. But no healthcare service provider depends that much on a single subscriber, especially if that person is replaced by another subscriber. ⁵ Warren is appropriately skeptical of how attentive bureaucracies are to voice, but he does not apply the same level of skepticism to thinking about how market forces will operate to motivate government service providers to improve.

Warren (also Taylor 2017) assumes that social service providers operating in their enabled and institutionalized structure will not operate the way CGG assume but will instead act more like businesses seeking to grow markets by offering better services. Economists have certainly shown us the logic of how competitive markets will motivate firms to compete in these ways. It is not clear to me, however, if we can assume that markets for healthcare services and education, much less police, parks, and public transportation, can be set up to operate with perfect competition. Most schools and hospitals operate in markets that are geographically contained. Someone who opts out of using one hospital’s services does not generally have very many alternatives in the area—especially in an era of regional hospital chains dominating many geographies. Many families also hesitate to send small children over great distances to attend a different school or childcare center. So what we may end up with in many voucher systems is not perfect competition but oligopoly—somewhere in the middle of the range between perfect competition and monopoly.

This brings me to my second problem with Warren’s confidence in enabled exit. By describing the utopia that might arise if we can create competitive markets for most public goods and services and convince public service providers to compete in the way firms do in competitive markets, he implies that any movement toward expanded choice is a step toward a better democracy. He focuses his discussion on two discrete options: the “voice monopoly” world (where citizens have no choice); and perfect competition where exit options are bountiful and enabled. But there is a vast space between these two end points on the exit cost continuum.

But can we be certain that any movement toward greater choice will yield a more responsive government? In figure 3 I have combined the two separate functions presented in figures 1 and 2, one for exit-driven reform and one for voice-driven and shown how we can expect the overall level of government responsiveness to vary across the full continuum of exit costs. It shows a large section in the mid-range of exit costs where we should expect neither exit-driven nor voice-driven responses to declining government performance. If citizens were forced to choose between exit and voice in response to discontent, in the way Hirschman proposed in 1970, the dip would not be there. Any point where citizens responded less with voice would see them responding more with exit, so that the combination would produce pressure for reform at every level of exit costs. But once you allow citizens to choose “silent non-exit,” the logics of collective action and of exit combine to create a wide zone in which citizens find that it makes sense, given the costs and predicted efficacy of exit and voice, to respond with neither one of these strategies—and so you get no reform.

Figure 3 Combined likelihood of reform as a function of exit costs

Note: Assumes modest, fixed cost of voice and fixed value of concession sought from the government.

What this figure shows, reflecting CGG’s logic, is that we should expect a strong response to exit only for a narrow segment where exit costs are zero or near zero. For the remaining segment of exit costs, from near zero to no-way-out, responsiveness improves with increased exit costs, rising to the highest levels of responsiveness as exit costs approach “no way out” because the inability to escape at this end of the continuum triggers collective action and the communication of information about what citizens need by large numbers of individuals. Over this span, one would be mistaken to expect—as Warren implies—that expanding exit options will improve the government’s response. Oligopolistic competition for public services, in the middle of the exit cost continuum, will tend to produce neither exit- nor voice-driven reform.

Conclusion

Fifty years after Hirschman presented the EVL framework, it is striking that it can be operationalized in such different ways. CGG, Warren, and others working in the public choice tradition insist that only those with attractive exit options have power. If we want better democracy, the only way to get there is to offer individuals more choices in markets for public goods. My operationalization of EVL insists that citizens are empowered, not only when exit is cheap, but also when the absence of exit opportunities motivates collective action. This framing suggests that policies that offer limited choice will actually disempower individuals by sapping collective mobilization and the communication of information about how public goods and services could be improved. My alternative framing, more faithful to Hirschman’s original ideas, does not settle the debate, however. If we are to advance a research agenda informed by his work, what we need next are careful experiments and observational studies that examine how individuals choose between exit, voice, and silent non-exit and how organizations respond over the full range of exit costs.