Cases: what real participation systems achieved, and where they failed

The Cases section of My Vote Forward is a library of in-depth studies of real systems that let the public set agendas, deliberate, allocate budgets or vote outside the usual rhythm of elections. Each case gives a dated account of what happened, how the mechanism worked, measured results with denominators, the strongest criticisms, what transferred elsewhere, and what the case means for this educational framework. The cases are evidence about other people’s institutions; none of them is evidence that My Vote Forward works, because the framework has not been evaluated.

Why study cases instead of models

Participation cases matter because arguments about democratic reform are usually made with a few famous examples, told selectively. Supporters cite Swiss frequency, Irish assemblies or Porto Alegre’s budget; critics cite low turnout, ignored recommendations or abandoned experiments. Both are partly right. A case study that records only success is advertising, and one that records only failure is a polemic. The pages in this section try to hold both at once, with sources a skeptical reader can check.

The shorter comparative overview introduces several of these examples side by side. The pages here go much deeper, and each one ends with a section on what My Vote Forward can and cannot do in that setting, because the honest answer is usually “less than a reader might hope.”

Five kinds of case

The cases in this first set fall into five families, which differ in who takes part and what power they hold.

Direct votes on questions. In Switzerland, 100,000 voters can force a national vote on a constitutional amendment and 50,000 can force a vote on a law Parliament has passed, and voters go to the polls about four times a year.⁠¹ Results are binding, and constitutional changes need a majority of voters and of cantons. The Swiss record shows durability over generations alongside long exclusions, such as the denial of the federal vote to women until 1971,⁠² and outcomes some view as limiting minorities.

Randomly selected deliberative bodies. Ireland convened six assemblies of randomly selected citizens between 2012 and 2023. Two preceded referendums that passed, on marriage equality in 2015 and abortion in 2018; a later one preceded two 2024 referendums that failed after the government changed the recommended wording.⁠^3,18 Madrid created a permanent lottery-selected City Observatory of 49 residents in 2019; the next city government repealed it in February 2020.⁠⁴

Participatory budgets. Porto Alegre created the model in 1989 and drew tens of thousands of mostly lower-income residents into assemblies, before a backlog of unbuilt projects, budgets of zero and a pause from 2019 to 2022.⁠^5,6 Reykjavik runs a smaller, binding neighborhood budget with online voting, where turnout has ranged from about 11% to 16% of eligible residents.⁠^7,8 New York City and Paris adapted participatory budgeting to large cities with capped funds.

Online agenda-setting and consultation. Better Reykjavik commits city committees to consider the most-supported resident ideas every month, without a minimum vote threshold.⁠⁹ vTaiwan used opinion-mapping to find “rough consensus” on national digital regulation, most famously ride-hailing, but had no power to make a ministry act.⁠¹⁶ Decide Madrid set a 1% signature threshold that, in practice, only two proposals cleared before its first citywide vote in 2017.⁠¹⁰ Iceland posted constitutional drafts online for public comment; a study found that almost 10% of public proposals changed the text, but parliament never adopted it.⁠¹¹

Remote voting in official elections. Oregon has run every election by mailed paper ballot since 1999, with signature checks and audits, measured fraud convictions in the dozens across tens of millions of ballots, and a turnout effect that careful studies find to be small.⁠^13,14 Estonia has offered binding internet voting since 2005; in 2023 more than half of participating voters voted online, while security researchers and observers continue to debate verification, insider risk and coercion.⁠¹⁵

What the cases have in common

The cases differ in country, scale and tool, but five patterns recur. They are interpretations drawn from the case evidence, and each case page gives the sources.

1. Authority over the output decides almost everything. Advisory processes rise or fall on whether someone with power chooses to act. vTaiwan’s outputs depended on ministries; Ireland’s assemblies depended on governments calling referendums and keeping recommended wording; Iceland’s draft depended on two successive parliaments. Binding processes, such as Swiss votes, Reykjavik’s neighborhood budget or Oregon’s ballots, have a different problem: they need firm rules about what a majority may not decide and who checks the count.

2. Delivery sustains participation. Porto Alegre’s own survey data show that when approved projects went unbuilt, attendance fell, fewer new people came, and participants grew more pessimistic about whether they really decided anything.⁠⁵ Reykjavik officials made the same point in reverse: early vague answers to resident ideas undermined trust, and clearer answers helped.⁠⁹ A process that collects input and returns nothing teaches people not to take part.

3. Open participation is rarely representative. Oregon’s mail ballots made voting easier mostly for people who already voted.⁠¹⁴ Iceland’s online drafting input came largely from a small number of highly active men.⁠¹¹ Reykjavik’s budget is decided by roughly one in eight eligible residents. Lottery-selected bodies in Ireland and Madrid were designed to answer this problem, and they introduce their own questions about recruitment and legitimacy. Neither approach is free.

4. The count and the selection must be checkable. Iceland’s 2010 constitutional assembly election was annulled for defects such as traceable ballots and unlocked ballot boxes.⁠¹² Ireland’s 2018 assembly caught its own recruitment error through a randomized check.⁠¹⁷ Reykjavik’s internal auditors confirm that published neighborhood-vote results match the count.⁠⁸ Estonia’s online system has drawn the most sustained security debate of any case here. Verification is what turns a result into something people on the losing side can accept.

5. Institutions created by one government can be undone by the next. Madrid’s Observatory lasted about a year. Porto Alegre’s budget was weakened by later administrations without being formally abolished. Better Reykjavik survived several governments partly because city staff institutionalized it early. Durability seems to come from legal standing, a constituency beyond one party, and visible results.

How to read a case page

Every case page in this section follows the same structure so that cases can be compared:

• What happened gives a dated timeline from official and research sources.
• How it worked explains the mechanics: legal basis, thresholds, eligibility, selection, technology, and whether outputs are binding or advisory.
• Measured results gives numbers with their denominators. Where a source is the program’s own report, the page says so.
• Failures and criticisms presents the strongest objections from more than one direction, attributed to the people who make them.
• What transferred and what did not looks at adoption elsewhere.
• Lessons for My Vote Forward is labeled interpretation and design proposal, not finding.
• What My Vote Forward can and cannot do here states the framework’s limits in that setting.
• Unresolved questions lists what the sources could not answer.

Every material factual claim is cited inline and recorded in the site’s public sources ledger. When a figure could not be verified, it was removed or moved to the unresolved questions rather than estimated.

What these cases mean for My Vote Forward

My Vote Forward is an educational framework that connects shared priorities, optional public expression of opinions, and secure private voting, with a demonstration that uses simulated data. The cases shape it in specific ways, all of which are design choices still open to revision:

• Priorities are information, not mandates. Because authority decides almost everything, the framework never presents collective priorities as instructions to any institution.
• Status after publication matters. Because delivery sustains participation, the framework’s proposal lifecycle and participation over time track what happened after people spoke.
• Denominators are mandatory. Because open participation is rarely representative, every published release must state who could have taken part and who did; see measurement.
• Checks come before claims. Because results must be checkable, the framework treats audits and verification and security as requirements, not features, and makes no claim of official voting capability.
• Governance is written down. Because institutions can be undone, the governance chapter sets out who controls the agenda and how rules change.

Unresolved questions

• Which design features, if any, make advisory participation reliably influential without making it binding?
• Can participation be made broadly representative without lottery selection, and at what cost?
• How should a process report outcomes when the people who took part are a small, self-selected share of those affected?
• What legal or institutional forms best protect participatory processes from being hollowed out by later governments?
• How should remote participation balance convenience against the loss of the private booth that protects voters from pressure?