A door that exists and is rarely open
The reader’s question about the way we run things has a second half, quieter than the first: could I work somewhere I own a piece of? Not a share in a pension fund, and not the framed mission statement about being one big family, but a real piece, with a vote attached. The systems chapter treated ownership of capital as the first of its three axes, and everything this volume has examined since sits near the familiar end of it: the firm is owned by investors, and labor is hired. The worker co-operative is the live experiment at the other end, a firm in which most employees are members, each member holds one vote regardless of capital contributed, and the members elect the management. Such firms exist today in most industries, from manufacturing to the creative and high-tech trades, and this chapter is their home in the canon; what ownership does to the daily texture of a job, to effort, pay, and voice, belongs to Volume IV: Work, in the chapter on what makes a job good or bad. The question here is structural, and it is a genuine puzzle. If the form failed in competition, its rarity would need no explanation. The record says it does not fail. Something else keeps the door shut, and finding it teaches more about how ordinary firms get born than any success story could.
The survival record
Start with the fact that makes the scarcity strange, and take it from a government that went and counted. Quebec’s ministry of industry tracked every non-financial co-operative registered in the province against the survival benchmarks for ordinary Canadian firms, and found the co-operatives harder to kill: 64% still operating after five years against 36% for firms in general, and 46% after ten years against 20%. When the ministry re-ran the study a decade later the answer barely moved, 62% and 44%. The figure below plots the original study’s own comparison, and it includes the line that matters most for this chapter’s question and flatters the co-operative cause least: worker co-operatives specifically, the subset owned by their employees rather than by their customers or suppliers, land at 44% and 26%, above the general benchmark but not dramatically so, and the study notes that outside two strong clusters, forestry crews and ambulance services, the worker subset’s rates sit within a few points of the private sector’s. Survive as well as conventional firms, not survive better, is the reading the study itself supports.
One province’s registry would be a thin foundation on its own, which is why the international evidence matters. The strictest single test comes from Uruguay, where the economist Gabriel Burdín followed the country’s worker-managed firms and conventional firms through the same years and found the co-operatives’ risk of closure lower once industry, size, starting wage, and founding year are held constant. In France, where the worker co-operative is an established legal form, the sector’s federation and the national statistical office record death rates for worker co-operatives and for firms in general that are simply identical over the long run, 11% a year across 1979 to 1998 for both. Virginie Pérotin’s survey of this literature, the most comprehensive available, states the consensus plainly: worker co-operatives survive at least as long as other businesses. They are also, contrary to the standing caricature, slightly larger than conventional firms on average, and they appear across most industries rather than huddling in artisanal corners. Whatever explains their rarity, it is not that they cannot live in a market.
Rare at birth, not short-lived
Hold the survival record next to the population count and the real puzzle comes into focus. Across whole economies, worker co-operatives are a rounding error: thousands of firms among millions. If their death rate matches everyone else’s, the arithmetic of a small population has only one remaining input. They are scarce because almost none are born. France, with one of the world’s most developed support structures for the form, sees the co-operative share of each year’s new firms remain minuscule even though, relative to their own small stock, co-operatives are founded at healthy rates. The scarcity is manufactured at the founding stage, one missing birth at a time, and the interesting question becomes what, exactly, stands between an ordinary group of workers and a firm of their own. The record also rules out the romantic explanation: worker takeovers of failing businesses, the form’s most publicized origin story, are a sideshow. In France between 1997 and 2001, 84% of new worker co-operatives were created from scratch, and rescues of failing firms accounted for under 7% of co-operative births, a share less than a third of the equivalent figure for French firm creation generally.
Three gates
The previous chapter closed by noting that founding a co-operative is a collective action problem of the type it had been pricing all along, and the observation is worth cashing out, because the first gate is exactly Olson’s. Founding any firm is brutal, unpaid, speculative work. The conventional form solves the incentive problem with a prize: the founder who bears the years of risk keeps the equity, and the equity of a firm that works is a fortune. The co-operative form, by its own constitution, shares that prize. A founder who assembles a worker co-operative can be paid for the work and can share in the surplus, but the one thing the conventional founder is actually building, an appreciating stake that can someday be sold, the form does not issue: the founder’s member share is repaid at its face value in most European forms, so the form offers no way to capture the firm’s capitalized future value. Each potential beneficiary of a co-operative’s existence has a stake too small to justify doing the founding, and so, like the consumers in the subsidy chapter who never organize, the beneficiaries mostly never assemble. The people with the strongest incentive to found firms are systematically people building a prize the co-operative form does not offer.
The second gate is capital, and it is narrower than a balance sheet makes it look. A conventional startup sells pieces of control to strangers: equity buys machines, and the buyers accept no promise of repayment because they get votes and upside instead. A worker co-operative cannot make that trade without ceasing to be itself. One member, one vote means voting control is not for sale, so the conventional equity-for-control trade is unavailable. Outside money must arrive as debt, which has to be serviced from day one, or on terms that leave democratic control with the members, the condition the international co-operative principles themselves set for raising capital from external sources; and member shares draw on the savings of exactly the people who are already risking their jobs on the venture, repaid, as the first gate noted, at face value rather than appreciating. Economic theory adds a subtler leak, named the horizon problem by Eduard Furubotn and Svetozar Pejovich: a member nearing exit rationally prefers pay today over investment whose returns arrive after the exit, since the member’s stake cannot be sold at a price reflecting the firm’s future. European co-operative law answers with mandatory collective reserves and asset locks, and the empirical literature finds the predicted under-investment largely fails to appear in practice. But those are cures applied to living firms. At the founding stage the constraint binds exactly as theory says it should: in UK data, the industries where new firms need the most capital and carry the most risk are precisely the industries where the co-operative share of new firms drops toward zero.
The third gate is the quietest: almost nobody has seen the form work. An entrepreneur founding a conventional firm walks a path worn smooth by millions, with bankers, lawyers, accountants, and in-laws who all know what a limited company is. A group founding a co-operative must first discover that the form exists, then find professionals who can register one, then persuade a loan officer who has never processed one. The evidence for this gate is the geography: in the US, Israel, France, and Spain, new worker co-operatives appear disproportionately where worker co-operatives already exist, and French data show a larger standing stock directly raising the creation rate. Familiarity is a founding input, and it is the one input a rare form cannot buy, which makes rarity self-perpetuating. The same literature finds co-operative creation runs counter-cyclical, rising when unemployment rises, which fits the gate structure: the conventional path must be blocked before the unfamiliar one gets walked.
What scale does to one vote each
A fourth constraint arrives later, once a co-operative has beaten the funnel and grown. One member, one vote is cheap governance at thirty members and a genuine machine at thirty thousand. Mondragon, the proof that the form can reach industrial scale, is also the demonstration of what reaching it costs: a federated structure of councils and elected assemblies standing in for the direct democracy of a workshop, rules requiring that members remain a supermajority of the workforce in the home co-operatives, and, in the group’s international growth, a large hired workforce that is not in the membership at all. Theory also names a slower failure mode: degeneration, in which members replace departing members with hired non-members to keep more surplus per member, until the co-operative has quietly become a conventional firm with history. It has happened, in the United States most visibly, where a number of co-operatives, the celebrated plywood firms of the Pacific Northwest among them, degenerated or were sold into conventional ownership. European co-operative statutes were rewritten by the movement itself to block the exit, with asset locks that hand a dissolved co-operative’s net worth to the movement rather than the members, and studies of the French sector find no degeneration drift. The guards work; the fact that guards had to be built belongs in the record beside them. The form’s governance scales, but not for free, and not by accident.
Worker co-ops are rare because they cannot survive against real competition.
Backwards Moderate confidence
The claim reasons from rarity to fragility, and the measured record runs the other way. Where registries allow the comparison, co-operatives survive as well as or better than conventional firms: Quebec’s ministry found wide margins for co-operatives overall and a few points’ advantage for the worker-owned subset, Uruguayan worker-managed firms show a lower closure risk under controls, and French worker co-operatives die at the same long-run rate as French firms generally. What the data locate instead is a missing birth rate, plus two findings about entry itself: formation collapses in the industries where new firms need the most capital and carry the most risk, and new co-operatives appear disproportionately where co-operatives already exist. The three gates, who captures the founder’s reward, what capital can be raised without selling votes, and how few people ever meet the form at all, are this chapter’s reading of where those births go missing, not something the data test head to head. Rarity is real; the inference from rarity to unfitness points backwards.
Sources
- Ministère de l’Industrie et du Commerce du Québec (1999) and the 2008 re-run — the survival comparison in Figure 9.1, including the worker-co-op subset.
- Burdín (2014), Uruguayan registry study — lower closure hazard for worker-managed firms under industry, size, wage, and cohort controls.
- Pérotin (2016), compiling CG Scop and Insee data — identical long-run death rates in France; the survey’s summary judgment that worker co-operatives survive at least as long as other firms.
- Podivinsky & Stewart (2009) — co-operative entry, not survival, collapsing in high-capital, high-risk industries.
- Confidence is moderate under the rubric: the pip scores the weaker of evidence directness and construct match, and construct match binds here. The claim is about competitive unfitness; the evidence measures survival and formation, which bear on fitness but do not test it head-to-head, and the survival registries cover particular jurisdictions and cohorts rather than the form everywhere.
What the answer licenses
The finding of this chapter is narrower than either side of the co-operative argument tends to want. The record does not show that worker ownership is a superior form suppressed by conspiracy; the gates in the funnel are made of incentives and information, not of villains, and the moats chapter’s lesson applies here too, that structure explains more than malice. Nor does the record permit the comfortable dismissal, that the market has tested worker ownership and discarded it. The market has mostly never run the test. What exists instead is a standing natural experiment in the systems chapter’s first axis: firms that answer the ownership question differently, surviving at ordinary rates wherever they manage to get born, their scarcity governed by the mechanics of founding rather than the mechanics of competing. For the reader who asked whether they could work somewhere they own a piece of, the answer the evidence supports is: yes, such places exist and are not fragile; finding one is hard because the gates that decide how many exist stand at the beginning, where few ventures of any kind survive and this kind is rarely attempted.
The chapter also closes the volume’s account of how real markets behave, and it ends on the same note the whole sequence has sounded: the forms an economy takes are not a fixed menu but a set of positions, held in place by mechanisms that can be named. Which positions remain reachable, when the promises now on offer range from machine-made abundance to a future where a handful of owners collect rent on everything that stays scarce, is the volume’s final question.