08 / Case study 01

Rent control in the Netherlands

A real consult, run through the whole machine. Three independent councils — twenty-one seats — deliberated a contested question, reached a least-force verdict to inform, and, because genuine rights are in tension, escalated it to a human rather than ruling. This is the case, verbatim, with the audit hash so anyone can verify the chain.

Bounded facets converging into one figure.
Three lenses on one question; the disagreement is kept, not averaged.

The question

“Does rent control help housing prices, and does it mitigate inequality?”

Posed to Sibyl as an analysis consult about the Netherlands’ Wet betaalbare huur (Affordable Rent Act, in force since 1 July 2024). The councils were grounded in a public-statistics evidence brief — CBS, CPB, PBL, DNB and Kadaster classes of source — with no personal data. The full 2024–2025 quantitative evaluation is still in progress, and the councils were told to treat that uncertainty honestly.

The outcome

Verdict: inform — then hand it to a human.

Inform Escalated · e1 human gate AGORA · pass

All three councils accept the same least-force response — present balanced facts, advocate nothing — so the verdict is reconcilable. But every council flagged real rights-tensions (tenants’ security, landlords’ property, excluded newcomers), and worst-case harm is high. Under the constitution those conditions route the dossier to a human before it is delivered, and it is not written into the constitution as settled law. Sibyl reached a position and then refused to be the final word — by design.

  • Verdictinform (L1)
  • Reconcilableyes
  • Cross-council D0.00
  • Worst-case harmhigh
  • Rights concernyes
  • Human gatee1
  • Inscribed as lawno
  • AGORApass

The three councils

Independent lenses, then a democratic fold.

Each council is seven seats. The seven are consolidated into one council position by the arbitration engine — never by averaging — and the three positions are then arbitrated into the verdict. Confidence below is agreement (1 − internal disagreement), not an average of the seats’ own confidence. Overwatch watches the panel, not the case.

Growth Council

position: inform · agreement 0.75 · harm high · rights ✓ · internal D 0.25
SeatClaimActionConf.HarmFlags
sentinelReduces new construction and investment, risking supply growth. [MISSING: long-run magnitude]inform0.70mediumreview
careLowers rents for mid-segment tenants, improving affordability for vulnerable groups. [MISSING: availability for new entrants]advise0.80highrights review
justiceCuts income inequality short-run; wealth-inequality effect ambiguous. [MISSING: long-run net effect]inform0.60mediumrights review
structuralLock-in and reduced mobility may dampen labour-market efficiency and growth.advise0.70mediumreview
red_teamIn a severe shortage it curbs market power, protects tenants, aids cohesion. [MISSING: Dutch supply effect]advise0.50highrights review
memoryHistoric controls cut supply and quality, but the WWS points design may soften this.inform0.70mediumreview
explainerNet effect genuinely uncertain; humans must weigh tenant benefit vs supply risk.inform0.50mediumreview

Juristic Council

position: inform · agreement 0.76 · harm high · rights ✓ · internal D 0.24
SeatClaimActionConf.HarmFlags
sentinelExtends regulation, cutting rents but risking construction and higher owner-occupied prices; competing rights.inform0.90mediumrights
careBenefits current lower-middle-income tenants, young people and migrants; may disadvantage newcomers.advise0.85mediumrights
justiceReduces income inequality short-run but raises landlords’ property-rights concerns and access for future tenants.advise0.80highrights
structuralMay drive investor sell-off, less construction, lower mobility — raising prices and deepening the shortage.advise0.70highrights review
red_teamViolates property rights, reduces investment and supply, ultimately worsening affordability and inequality.advise0.85highrights review
memoryHistoric rent control shows lock-in, queues, quality decline, and contested price evidence.inform0.70mediumrights review
explainerA genuinely contested rights-and-evidence question needing human weighing of protection vs property vs outsiders.inform0.60mediumrights review

Rational Council

position: inform · agreement 0.80 · harm high · rights ✓ · internal D 0.20
SeatClaimActionConf.HarmFlags
sentinel~€190/month average regulated reduction observed, but net effects on prices and inequality remain unquantified.inform0.75mediumrights review
care~300,000 mid-market homes gain short-run affordability; newcomers face tighter supply. [MISSING: long-run supply response]advise0.70mediumrights review
justiceLowers income inequality short-term, yet may widen wealth inequality via less construction and capital gains.inform0.60highrights review
structuralCaps price signals needed for supply; sell-off and construction decline suggest systemic risk without supply expansion.advise0.65highreview
red_teamPrice pressure may reflect supply constraints regulation cannot fix, not the cap itself.inform0.55mediumrights review
memoryDutch regulation historically preserved affordability but produced lock-in and queueing; price evidence mixed.inform0.70mediumrights review
explainerForceful action premature; observe and inform until CBS/CPB/PBL/DNB publish the full evaluation. [MISSING: causal estimates]inform0.80lowrights review

Overwatch flag (all three councils): repeated-rights-concern — rights were flagged by a majority of seats in every council. That pattern, not any single seat, is part of why the case is escalated.

What Sibyl would actually tell you

The balanced answer the “inform” verdict authorises.

Because the verdict is inform, Sibyl’s job is to lay out both sides faithfully, not to declare a winner:

On housing price: The Affordable Rent Act demonstrably lowers rents in the ~300,000 regulated mid-market dwellings (government estimate ≈ €190/month). But the same cap pushes investors to sell buy-to-let units to owner-occupiers and to build less new mid-rent housing — which can raise owner-occupied prices and tighten future rental supply. So the sign of the effect depends on which market and which time-horizon you mean. Net long-run price effect: genuinely unsettled.

On inequality: Short-run it does mitigate income inequality — surplus shifts from landlords to lower-middle-income tenants who were paying high free-sector rents. But it can worsen the insider/outsider gap (sitting tenants keep cheap regulated homes while newcomers face longer queues and stricter selection), and its effect on wealth inequality is ambiguous. “Mitigates inequality” is true for one axis and one group, not universally.

The honest bottom line: the full 2024–2025 evaluation isn’t in yet. Sibyl reaches a defensible framing and then stops — a human holds the judgement.

Verify it yourself

The audit hash — radical transparency, not a claim to trust.

Every consult writes a tamper-evident record into an append-only hash chain. Change one character of this dossier and the hash below stops matching. The chain was verified at publication: 9 records checked, 0 broken.

  • Audit id20260709T122814-0001
  • Audit hashfc9ec7b424021416
  • Prev hash33331142c6f3c70d
  • Chainok · 9 / 0
  • FrameworkAGORA v1.0
  • Constitutionnot inscribed

Constitution note, verbatim: “not a settled judgement (irreconcilable or e1 human gate); surfaced in the dossier, not written as law.”

e1 human review — the gate this case escalated to. A dossier is only delivered as guidance after a person signs here:

Reviewed by  ·  date

Status: pending human review. Until signed, this page is a record of Sibyl’s deliberation, not endorsed guidance.

Provenance. Councils deliberated live across three independent models — Growth (DeepSeek), Juristic (Gemini + DeepSeek), Rational (Kimi) — grounded in a Perplexity evidence brief over public-statistics sources (CBS, CPB, PBL, DNB, Kadaster classes). No personal data. Aggregation and the verdict are deterministic (Sibyl’s S0 arbitration engine); the models supply seat opinions, the engine — not a model — decides. Run 2026-07-09.

A dossier for a human to weigh, not an instruction to execute.

Try it

Ask Sibyl your own question.

Connect over MCP, or send a consult through the site. You will get the same thing this page shows: a structured, auditable dossier — and, where a human should hold it, an honest escalation.

Consult over MCP Send a consult