California has the most underfunded public pension system in the country. CalSTRS, the fund that pays the retirements of the state’s teachers, carries an unfunded liability of more than $100 billion—larger than California’s entire annual K–12 education budget. CalPERS, the parallel system for non-teaching public employees, brings the combined liability to roughly $136 billion. Between 2013 and 2021, the share of a California teacher’s salary that school districts had to send to CalSTRS more than doubled, from 8.3 percent to 19.1 percent, and roughly forty percent of recent increases in school budgets have been absorbed by pension costs rather than reaching classrooms. Three California cities—Vallejo, Stockton, and San Bernardino—have declared bankruptcy in the last two decades, with police and firefighter pension obligations as one of the largest drivers in each case. Cities that have not gone bankrupt have responded by cutting parks, libraries, road maintenance, arts programs, and, in some cases, the very police staffing levels their pension obligations were originally meant to support.
These are the visible fiscal effects of California’s public-sector unionization. By that term I mean the organized collective bargaining of government employees: teachers, police officers, firefighters, and the rest of the state and municipal workforce. This institutional arrangement is distinct from private-sector unionism in ways that turn out to matter a great deal.
Get Liberalism.org in your inbox.
The fiscal effects of public-sector unionism have produced a small library of policy reports and a steady stream of newspaper investigations. But they are not the deepest version of the problem, and they are not the version I want to focus on here. The deeper problem—and the one that should concern classical and modern liberals alike—is a problem of power.
***
The first essay in this series argued, following the pioneering University of Chicago economist Henry Simons, that concentrations of power are the principal threat liberals of all stripes should be working to prevent. The second argued that this concern reaches into wealth itself: when private fortunes grow large enough to capture the institutions that are supposed to constrain them, the classical liberal commitment to limited government may find itself pointed at the wrong targets.
This essay extends the same diagnosis to another concentration of power, one that classical liberals have long named as such, but that modern liberals have tended to read as a victory for democratic accountability. I want to argue that modern liberals should read the situation more carefully, and that public-sector unions—particularly teachers’ unions and police unions—exhibit exactly the kind of power-problem Simons taught us to look for.
The case for public-sector unions, as usually made, runs through their private-sector cousins. Workers facing concentrated capital need organization. A single worker bargaining against a large firm is at a structural disadvantage that the law of contract cannot easily remedy. Collective bargaining restores something like parity. This is the countervailing-power argument: Concentrating capital can unlock economic efficiencies, making it socially useful, but the results can also bias market outcomes in a corporation’s favor, above all when the counterparty is a single individual. In times like those, it can help to have an organization on the individual’s side. In private-sector contexts, that argument has real force. Samuel Bagg, in his recent book The Dispersion of Power, has recently developed this argument with particular clarity: Liberal democracy works only when concentrated power on one side of a social relation faces sufficient organized power on the other, and the labor union is among the few institutional vehicles we have for the latter.
But the structural setup of public-sector unionism is different in two specific and consequential ways, and the difference complicates the countervailing-power argument rather than extending it.
The first difference is on the union side. A private-sector union bargains with a counterparty, the corporation’s management, whose election it does not finance and whose tenure it cannot directly affect. A public-sector union bargains with a counterparty—a state legislator, a city councilmember, a school board member—whose campaign it has often funded. If today’s negotiation does not go well, it can fund the opposition tomorrow. A public official’s continued tenure depends on the kind of organized political support that public-sector unions specialize in providing. This is not an arm’s-length transaction. The employer is an institution that the employee has had a hand in shaping, and the negotiator across the table is, in significant part, the union’s own political product. A counterparty in that position has weaker reasons than a normal employer to drive a hard bargain, and stronger reasons to please the people sitting opposite.
The second difference is on the government side. A private employer who agrees to a generous contract feels the consequences in firm revenue, share price, and maybe in personal compensation. The negotiator either has equity at stake or reports to people who do. A public-sector negotiator faces none of these incentives. The money committed to wages, benefits, and pensions is not the negotiator’s money; it is the taxpayer’s. The political costs of granting a generous contract are diffuse and deferred; the political costs of resisting are immediate and concentrated, since the union is in the room and the taxpayer is not. This is the public-choice problem that James Buchanan, Gordon Tullock, and Mancur Olson described. It has particular force here, where a party with coordinated interests, the union, faces a counterparty, the state, that is inherently conflicted and lacks meaningful incentives to strike a good bargain. The state’s incentive structure tilts systematically toward giving in.
These two structural facts—a politically advantaged union and fiscally insulated state negotiators—are not separate problems. They are the same problem, viewed from different sides of the table. Bagg’s countervailing-power argument, applied without modification, sees only the union side and reads it as labor exercising legitimate organized power against an employer. The framework misses what it is actually best equipped to see: the asymmetry of the underlying relation. In the public-sector case, the counterparty isn’t organized in the pursuit of the relevant interests. Instead, the union has helped to fund its election, while its financial discipline depends on a constituency, the taxpayers, that has no presence in the room. The union, having organized to face a concentrated employer that does not exist, now faces a counterparty that cannot resist it.
Daniel DiSalvo’s Government against Itself remains the most thorough scholarly treatment of these dynamics, and DiSalvo names the structural fact more compactly than I have: “in collective bargaining negotiations with the government, the people’s elected representatives partially cede control of the bureaucracy to unelected labor leaders. Many policy choices are then the outcome of negotiations between officeholders and unions, rather than the expression of the people’s will channeled through their representatives.” The negotiation, on this view, is not between government and a private counterparty. It is between two factions of government. One has reasons to enlarge itself, and the other has no real reason to resist.
***
The teachers union is the cleanest case, partly because the modern liberal stake in the question is easier to see. The basic story is familiar to anyone who has followed California education politics over the last two decades. The California Teachers Association and its local affiliates—United Teachers Los Angeles is the largest—are by some metrics the most powerful political organizations in the state. The CTA’s political action arm has consistently been among the largest contributors to state legislative campaigns. School board elections in major districts are typically lopsided. The union-endorsed candidate enters with a substantial advantage in funding and organized support that opposing candidates rarely match. The contracts that result govern not only wages but tenure protections, dismissal procedures, layoff order, evaluation criteria, classroom assignment rules, and a great deal of what looks at first like operational discretion but turns out to be significant public policy. Terry Moe’s Special Interest: Teachers Unions and American Schools provides the canonical scholarly account of these dynamics.
This is the half of the story classical liberals have always emphasized: an organized interest captures the public institutions that are supposed to constrain it, and the institutions then function in significant part for that interest’s benefit. The pension obligations sketched at the beginning of this essay are one downstream effect. The work-rule provisions that make it nearly impossible to remove an ineffective teacher are another. Decades of California school reform efforts—from charter expansion to performance pay to tenure reform to merit-based layoffs—have foundered in part because the teachers union has the political muscle to defeat them, and the politicians who would face them are the politicians the union helped to elect.
But there is another half to the story, one I would urge modern liberals to take more seriously than they have. The students most damaged by these arrangements are not the children of suburban affluence. The teachers who are nearly impossible to dismiss are concentrated, by the standard operation of seniority and transfer rules, in schools that serve poor and minority children. Layoff procedures that protect senior teachers regardless of effectiveness fall hardest on the schools where high-quality new teachers are most needed. Curricular and pedagogical decisions that emerge from union-shaped bargaining are decisions the families of the affected children have very limited ability to influence. The pattern is familiar. An organized political constituency captures the institutions intended to serve a more diffuse one, and the unorganized constituency pays the cost.
The clearest recent illustration is the COVID school closures. The opening months of the pandemic, before vaccines and before reliable data on in-school transmission, presented a period of genuine uncertainty in which workplace safety concerns were not unreasonable. What followed those months was different. Well into 2021 and 2022, after vaccines were widely available and most of the rest of American life had returned to something like normal, major urban teachers’ unions—the Chicago Teachers Union, United Teachers Los Angeles, the Boston Teachers Union, the United Federation of Teachers in New York—continued to resist a return to in-person instruction. The American Federation of Teachers, the national umbrella, was caught lobbying the CDC to make its reopening guidance more restrictive than CDC scientists had originally proposed. Schools in jurisdictions where unions had less political leverage reopened months—sometimes more than a year—earlier than schools where the unions held effective veto power over the decision. The costs of those delays fell, predictably and measurably, on the children least able to bear them. Wealthier families put their children into private schools that had reopened, into learning pods with hired tutors, or they simply kept a parent home to supervise the laptop. Poor and minority families had none of those exits. The NAEP results that began arriving in 2022 confirmed what had been visible in real time: substantial learning losses across the board, the deepest losses concentrated among low-income and minority students, and a sharp widening in the achievement gaps that American education had been trying to narrow for decades.
This is not a critique modern liberals should find difficult to recognize. It is the same critique they have made, correctly, of pharmaceutical companies’ influence over the FDA, of fossil-fuel companies’ influence over environmental regulation, of the financial sector’s influence over its supposed regulators. The structural pattern—organized interest captures the institution that is meant to constrain it, diffuse counterparty bears the cost—is the same. The difference is that in those cases modern liberals see the capture; in the public-sector union case they have tended to see solidarity. Both readings can be partly true. But even when solidarity is a real value, it is not a costless one—and the cost in this case has been tremendous.
***
If the teachers union is the case where the costs of capture are easiest to see, the police union is the case where the capture runs deepest—reaching beyond the bargaining table into law itself. Police unions have used their political power in two registers at once: at the bargaining table, where they negotiate contracts that constrain how officers may be investigated and disciplined; and in state legislatures, where they have lobbied for laws that codify those constraints into statute.
The collective bargaining agreements these unions negotiate are not merely about wages. They typically include provisions that delay misconduct interviews for forty-eight hours or more, that expunge disciplinary records after a few years, that exclude civilians from internal review processes, that prevent the public release of personnel files, and that limit the disciplinary authority of police chiefs over their own departments. The legal scholar Stephen Rushin documented the pattern empirically in a 2017 Duke Law Journal study that collected and analyzed 178 union contracts from major American police departments and found a substantial number to include precisely these provisions.
The legislative complement is the set of state statutes formally called Law Enforcement Officers’ Bill of Rights laws. First passed in Maryland in 1972 and now in force in fifteen states, these laws codify many of the same protections that appear in union contracts: cooling-off periods before officers under investigation may be questioned, the right of the officer to see complainants’ names and statements before answering them, restrictions confining investigators to sworn officers (which blocks independent civilian review), and time limits on internal investigations. Every state adoption has followed lobbying by police unions and associations. Where the contract cannot deliver the protection, the statute does.
What these provisions do in practice is documented most fully in the reporting of Radley Balko, a journalist and Liberalism.org Fellow who has spent decades researching the subject. His eighteen-month investigation of the Little Rock Police Department, published in The Intercept in 2021, traced the campaign by the local Fraternal Order of Police to oust the city’s first Black reformist police chief, Keith Humphrey. Mayor Frank Scott had appointed Humphrey in 2019 to implement accountability reforms. Within two years, Humphrey faced lawsuits from twelve of his own officers, including the two assistant chiefs who had applied for his job; a sustained pattern of human-resources complaints; and a coordinated effort by the FOP, the largely white old-guard officers, and sympathetic city officials to make his job impossible. The union did not need to win every collective bargaining dispute to defeat reform. It needed only to make the political and contractual cost of reform high enough that the city would back away. It largely succeeded.
These contracts, statutes, and campaigns are not byproducts of legitimate labor negotiation. They are the political accomplishments of an organized interest that has captured the institutions meant to constrain it.
Here again, modern liberals already have the conceptual framework they need. They have spent the last decade arguing—correctly—that policing in America imposes its heaviest costs on poor and minority communities, that accountability mechanisms are too weak, that the institutions of state violence have evaded democratic control. Each of those claims is consistent with, and arguably best explained by, the structural fact that police unions have organized to make accountability politically expensive for the officials who would impose it. The labor framing—workers exercising organized power—captures something real about how the unions feel from the inside. It misses what the framework is best equipped to see when applied to anyone else: that the organized power in question is being used to evade democratic control of state violence.
***
The point of these two case studies is not that public-sector unions are uniquely malign, or that the people who staff them are acting in bad faith. Most teachers and police officers do their jobs well, often under conditions classical liberal commentators have not always taken seriously. The point is structural. Public-sector unions exhibit, with unusual clarity, the kind of power-problem the first essay in this series identified as the core liberal concern: a concentrated organized interest that has captured the public institutions meant to constrain it, in a way that imposes costs on a diffuse counterparty without meaningful political resistance.
This is a Simons-type problem. It’s what Simons meant when he wrote that no one—among others, no leader, no faction, no class, no church, no corporation, no trade association, and no labor union—could be trusted with much power. The reason classical liberals should care about this is a familiar one: when private organizations grow strong enough to capture the institutions of public life, the rule of law and the conditions of fair political competition both erode. The reason modern liberals should care is different but no less serious: the costs of capture in these particular cases fall on exactly the constituencies modern liberals are most committed to defending—poor and minority students, the victims of police violence, and the diffuse public that depends on functioning schools and accountable policing.
The question I have been pressing across this series is whether classical and modern liberals can recognize one another as allies in resisting concentrated power. The public-sector union case is, I think, one of the cleanest test cases. Classical liberals have been visible critics of these unions for decades. Modern liberals have been the unions’ most reliable political defenders. Both stances have been partly correct. The classical critique misses something real about why public-sector workers organized in the first place, and about the legitimate function of organized labor in domains where the countervailing-power story applies. The modern defense misses the structural inversion that occurs when the “employer” is the state that the union has helped to elect, and it misses the costs the resulting arrangements impose on the modern liberal’s own constituencies. A liberalism worth the name should be able to see both.
***
The reforms this diagnosis suggests are not new. Both DiSalvo’s Government Against Itself and Moe’s Special Interest develop detailed reform agendas along familiar lines—narrowing the scope of collective bargaining out of policy domains (curriculum, accountability, pensions, healthcare) and back into traditional labor domains (wages, hours, working conditions), restricting the political activities of organizations that bargain with the state, and reforming the accounting and approval procedures around long-term liabilities so that pension and benefit commitments are visible at the time they are made, rather than decades later. Rushin’s work on police contracts develops the parallel agenda for law-enforcement labor relations. All of these reforms are responsive to the specific power problem that the structural inversion produces.
I want to close, however, by saying what the diagnosis is not. It is not a call for abolition. Public-sector workers face real workplace issues—safety, training, retaliation against whistleblowers—that organized representation can legitimately address. The teacher who fears retaliation for reporting administrative incompetence and the officer who fears retaliation for refusing to cover for a colleague’s misconduct both have an interest in collective protection, and the union is not the wrong vehicle for it. The point is that the institutional form public-sector unions have taken in California and elsewhere has, over decades, accumulated political and contractual power that goes well beyond protecting workers from arbitrary employer action. It has captured the public institutions meant to constrain it. That is the power problem. The reforms that follow are not anti-union; they are anti-capture.
