Contribution
It studies holdup and capture together. The usual argument for an independent regulator is that it protects investors from electorally driven price cuts. The paper shows, in a model and in the lab, that the same commitment locks in whatever the designing policymaker wanted, so independence helps when the designer serves consumers and harms them when the designer serves the regulated firm. It names the second case “entrenched institutional capture, the institutionalization of regulatory capture in an independent regulator.”
Key insight
“Independence” here is independence from future elections, not from the industry. Whoever sets up the independent regime still sets it. If that person is captured, independence removes the one force that was pushing prices down, electoral pressure, and keeps the captured price in place. “The intuition underlying this effect is that flexible regulation allows electoral pressure to reduce the price, but independence does not. Instead, it offers the policymakers an opportunity to ensure lax regulation in the long run.”
Assumptions
- The game. Per period: the operator (IO) decides whether to invest (irreversible, costly, relieves a capacity constraint as demand grows); the IO and the two policymakers (PMs) announce prices (cheap talk); three consumer-voters elect one PM by majority; the elected PM sets the price in {0, …, 5}; the IO charges it. Nine periods, each group played once.
- Capture is exogenous and reduced-form. “In the case of capture, [PMs’] interests coincide perfectly with those of the industry”; without capture they are aligned with consumers. Alignment is common knowledge. “A detailed modeling of the nature of regulatory capture is beyond the scope of this paper.” No influence is exercised in the game. Capture is a treatment.
- Regulatory independence is commitment. In [RegInd] “only the PME elected in period 1 can choose the price which then holds for all periods.” In [RegFlex] the elected PM sets the price each period.
- Information. Consumers know their own payoffs and that higher prices are better for the investor, not the IO’s payoff table.
- Design. 2 × 2 between subjects (capture × independence), 351 subjects at CREED, Amsterdam, 59 groups of six; 41 groups without capture, 18 with. Fisher–Pitman permutation tests, two-sided.
- Outcomes. P6, the mean price over periods 1–6; I5, investment by period 5; ΔEP, the mean absolute deviation from the efficient no-loss price (1 after investment).
Key results
- Equilibria (Table 2, two boundary cases of voting). With no electoral pressure, capture gives high prices (4–5) and early enough investment; without capture, flexible regulation gives price 0 and no investment, the holdup. Independence without capture fixes price 1 and investment in period 4. With strong electoral pressure, capture does not matter in theory.
- Period-1 investment. 45.8% of IOs invested in period 1, a dominated choice, independent of treatment (p = 0.869, p > 0.999). Kept in the main analysis; Appendix F excludes them.
- Capture (§4.2). P6 significantly higher with capture (p < 0.001). All IOs invest by period 5 under capture; without capture 50% do, and 32% of groups never invest (p = 0.025). ΔEP 1.93 with capture, 0.77 without (p < 0.001).
- Table 4. Without capture: flexible P6 = 0.16, I5 = 63%, ΔEP = 0.98; independent P6 = 0.47, I5 = 84%, ΔEP = 0.53. With capture: flexible P6 = 1.78, I5 = 100%, ΔEP = 1.56; independent P6 = 3.19, I5 = 100%, ΔEP = 2.31.
- Independence without capture (§4.4). Prices rise toward 1 (p = 0.015), closer to efficient (p < 0.001). The investment effect is not significant (p = 0.255).
- Independence with capture (§4.5). Prices rise further from 1 (p = 0.001; p < 0.10 in the restricted sample). ΔEP higher, marginally by permutation test (p = 0.066), significantly by the Tobit regression of Table 5 (F = 8.80, significant at 1%; p = 0.07 in the restricted sample).
- Electoral pressure. Observed capture prices lie between the two equilibria: “Electoral pressure is stronger than assumed in the no-pressure equilibrium but weaker than in the pressure equilibrium.”
- Voting (Tables 6–7). Voters punish higher past prices (−23 percentage points for a higher last price) and weigh promises (−34 points per unit in period 1, −12 later). Promises count for more when PMs are not captured.
- Tacit collusion. Under capture with flexible regulation, only 36% of newly elected PMs lowered the predecessor’s price and 24% raised it.
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | In this game, policymakers aligned with the operator set higher prices and the operator invests earlier | strong | §4.2, p < 0.001 and p = 0.025, 59 groups |
| C2 | Without such alignment, fixing the price at period 1 moves prices toward the efficient level and reduces deviation from it | strong for prices, not significant for investment | §4.4, p = 0.015, p < 0.001; investment p = 0.255 |
| C3 | With such alignment, fixing the price raises prices further above the efficient level (“capture entrenchment”) | moderate | §4.5, p = 0.001 for P6 on 18 groups; ΔEP marginal by permutation, significant by regression |
| C4 | Voters’ electoral pressure is too weak to remove the effect of aligned policymakers, partly through tacit collusion between them | moderate | prices between the two equilibria; 36% / 24% of new PMs; three voters per group |
| C5 | Regulatory independence benefits consumers only when policymakers designing it serve consumers or electoral pressure is strong | moderate as a reading of C2–C3; a generalization beyond the lab | §5 |
| C6 | The electoral pressure in the lab is probably an upper bound for real elections | weak | asserted: three voters, one issue |
Concepts
- Holdup: delayed investment because electorally driven future price cuts may expropriate its return.
- Regulatory capture: here, policymakers whose payoffs equal the operator’s, “a reduced form approach … sufficiently general to allow for various underlying models”.
- Regulatory independence: a price fixed by the first elected policymaker that later ones cannot change.
- Capture entrenchment: an inefficient price set by captured policymakers that “cannot be corrected at later points in time”.
Connections
It cites Stigler (1971) and Peltzman (2022) for capture, Posner (1974) for dispersed voters and colluding politicians, and Martimort (1999) for agencies that drift toward industry as public attention declines, which its conclusion says it “complements”. Its experimental antecedents include Cox, Offerman, Olson and Schram (2002) on railways and Engel and Heine (2017) on price caps.
Bearing on the record
- THEORY-856 (Active) states C1–C4 as measured. THEORY-858 (Proposed) carries the authors’ general conclusion, C5.
- CLAIM-404. A25 cited the paper for “independence from one source of authority does not imply independence from every other source.” The paper supports a sharper form. Its independence is from voters, and capture enters through the designer. Making the regime independent of elections removed the only corrective channel left, the voters’ pressure, and so held the captured price in place. That is the claim’s last clause, that formal independence of a procedure does not make it independent, by a route the claim does not name: independence installed by a captured designer insulates the capture from correction. In CLAIM-342’s terms the voters were the evidence-bearing check outside the loop, and independence cut it. The paper’s capture is not informational or evaluative. It is an alignment of the decision maker’s own payoffs, so it does not bear on the claim’s two routes, which the LIT’s “complicates” overstated.
- TERM-066. The paper’s capture is the conventional interest-alignment sense, not the TERM’s fidelity-relative one, and it has no loop. It is a clean instance of what the TERM sets aside as the institutional sense, without the OECD’s intent condition: alignment is simply given.
- CLAIM-410. The two treatments are two governance profiles with the same price-setting authority at period 1 and different holders of revision after it: the electorate in [RegFlex], nobody in [RegInd]. That difference, not the number of rules, decides the outcome. It is an experimental instance of the claim’s point that authority over revision is a separate component.
- CASE-123 is stipulated. This paper is a laboratory measurement, not a documented episode of real regulation, so it does not supply the documented variant the case waits for.
No instruction for machine-learning practice; nothing for the anthology.
Limitations
- Capture is assigned, not produced. The study measures what independence does given captured policymakers. It says nothing about how capture arises, or whether independence makes it more likely, which the paper’s introduction cites as a further risk (Mitchener, 2007; “Rajam & Ramcharan, 2016”, as the paper spells it).
- Independence is commitment by the first policymaker. Real independent regulators have mandates, appointments and budgets that this does not model.
- Small groups. Three voters, two candidates, one issue. The capture cells have 18 groups. Some effects weaken in the restricted sample.
- 45.8% dominated period-1 investment. Treated as error and kept. The authors report that conclusions are mostly unaffected.
- An internal inconsistency. §4.2 says that without capture “only 50% of the IOs invest by period 5”, but Table 4 gives I5 = 63% and 84% for the two no-capture cells, so the pooled figure should lie between them. The text does not explain the difference.
- Appendices not read. The equilibrium derivations and the treatment pooling (Appendix D) were taken as stated.
Open questions
- Would independence also entrench capture that operates through the evidence a regulator receives rather than through its payoffs? The paper does not model that route.
- Does an intermediate design, such as fixed terms with periodic review, keep the holdup benefit without the entrenchment? The paper studies only the two boundary cases.