Comparative cases: what participation systems have achieved and where they fell short

Plain-language summary

People often point to a few famous examples when they talk about better civic participation. Switzerland votes on many questions. Taiwan used an online process to help write rules. Reykjavik lets residents choose neighborhood projects. Porto Alegre, in Brazil, let residents help shape the city budget. Ireland asked randomly chosen citizens to study hard questions before national votes.

Each of these is real, and each did something useful. Each also had problems. Some had low turnout. Some made recommendations that the government did not follow. One lost most of its power after a change of city government. One had to throw out some answers because members were not chosen the way the rules required. A popular vote can also produce results that raise questions about the rights of minorities.

This page puts the strengths and the limits side by side, with sources. My Vote Forward did not invent any of these practices. It studies them. For the history of how these mechanisms developed, see participation between elections.

Defined terms

• Direct democracy: arrangements in which voters decide a question themselves, usually through a referendum or a citizens' initiative, rather than only electing representatives who decide.
• Popular initiative: a proposal that reaches the ballot because enough eligible voters sign a petition.
• Optional referendum: a vote on a law already passed by a legislature, triggered when enough voters sign a petition.
• Participatory budgeting (PB): a process in which residents propose and choose how part of a public budget is spent.
• Citizens' assembly (mini-public): a group of residents selected by lot, usually with quotas so it resembles the population, which hears evidence, deliberates and makes recommendations.
• Binding versus advisory: a binding result must be implemented under law; an advisory result informs a decision maker who keeps the authority to decide. See proposals and the glossary.
• Turnout: the share of eligible people who took part. A result always has a denominator; see measurement.

Evidence

Each case below separates what the source states (evidence) from what this page concludes (interpretation).

Case 1: Switzerland, national direct democracy

Evidence. The Swiss federal government's information service states that a federal popular initiative needs 100,000 signatures collected within 18 months, and that an optional referendum against new legislation needs 50,000 valid signatures within 100 days of publication¹. Constitutional amendments approved by Parliament go to a mandatory national vote¹. The same source says voters are called to the polls about four times a year on an average of fifteen issues, and that turnout in recent decades has averaged a little over 40%¹.

Limit: who counts as "the people." Switzerland did not give women the federal vote until a referendum held on February 7, 1971, which 65.7% of the then all-male electorate approved; an earlier 1959 referendum had been rejected by 66.9%². In the canton of Appenzell Inner-Rhoden, women gained the cantonal vote only after the Federal Supreme Court held in 1990 that excluding them violated the constitution's equality clause². In this case a court, not a popular vote, ended an exclusion that popular votes had kept in place.

Limit: majorities and minority rights. On November 29, 2009, a popular initiative against the construction of minarets passed with 57.50% yes, on turnout of 53.76%³. The Swiss constitution does contain a check: the Federal Assembly must declare an initiative invalid in whole or in part if it breaches unity of form or subject matter, or "mandatory provisions of international law"⁴. That check is narrow and is applied by Parliament⁴. The minaret initiative was put to the vote and adopted³.

Interpretation. Switzerland shows that frequent voting on specific questions can be institutionalized for generations. It also shows that a high-participation system can coexist with turnout under half, with long exclusions, and with outcomes that some people view as limiting a religious minority. Frequent voting does not by itself settle who is eligible or which rights are outside the reach of a majority. See representation.

Case 2: vTaiwan, online consultation on national rules

Evidence. vTaiwan describes itself as "an open consultation process for the entire society to engage in rational discussion on national issues," launched in 2014, with four stages: proposal, opinion, reflection and legislation⁵. It uses Pol.is, a platform for opinion collection, to look for "rough consensus"⁵. Its own account says more than 4,000 participants took part in the ride-hailing (Uber) case, and that across more than 28 discussed cases, 80% led to "decisive government action"⁵.

Limit: no obligation to act. A 2018 MIT Technology Review report states that "the government isn't required to heed the outcomes" of vTaiwan debates, quotes a legislator calling it "a tiger without teeth," and reports that an online alcohol sales bill drafted from vTaiwan suggestions "never saw the light of day"⁶. The same report quotes a participant warning that ignored recommendations risk being seen as "openwashing"⁶.

Design detail. In Pol.is, as described in that report, participants cannot reply to one another's comments; they vote on statements, and the system clusters people who voted similarly⁶.

Interpretation. vTaiwan shows that structured online input can map agreement across groups that disagree. The 80% figure is the project's own count, not an independent evaluation, and it does not say which issues were never brought to the process. Its weakest point was authority: an advisory process depends on officials choosing to use it.

Case 3: Reykjavik, neighborhood participatory budgeting

Evidence. The City of Reykjavik describes My Neighborhood as a collaboration between residents and the city administration in which, every other year, residents submit project ideas, the ideas are reviewed, selected ideas go to a vote, and winning ideas become projects the following year; the city states the funding is 850 million krónur each time, divided between districts⁷. For the 2018 round, the city reports that 13,003 of almost 106,000 eligible residents aged 15 and older voted, a turnout of 12.3%, which it described as the highest ever, and that 88 projects were chosen⁸.

Limit. By the city's own figures, roughly 88% of eligible residents did not vote in its best year to that point⁸. The city page does not describe the vote as legally binding; it says winning ideas become projects⁷.

Interpretation. Reykjavik shows a durable, repeated cycle with a clear path from idea to project, which is what many consultations lack. It also shows that a small, self-selected share of residents decides how the money is spent. Whether that share resembles the city depends on who participates, a question covered in measurement.

Case 4: Porto Alegre and participatory budgeting in Brazil

Evidence. A 2022 study in the Brazilian Political Science Review states that the name Participatory Budgeting and its assembly format were created in Porto Alegre, starting in 1990⁹. The study counts 267 Brazilian municipalities that adopted PB between 1989 and 2016, finds adoption peaked at 137 cases in 2004 and then fell continuously to 2016, and reports that adoption and continuity were strongly associated with per capita budget and investment levels⁹. Its authors note limits in data reliability and that only municipalities over 50,000 inhabitants were analyzed⁹.

On outcomes, a 2013 World Development study of Brazilian municipalities from 1990 to 2004 reports that municipalities using PB allocated spending that "closely matched popular preferences," shifted spending toward sanitation and health, and that this was "accompanied by a reduction in infant mortality rates"¹⁰. The author frames the result as suggesting an effect, not proving one¹⁰.

Limit: durability. A 2014 Journal of Latin American Studies article examines Porto Alegre after the party that created PB lost the mayoralty in 2004. It finds that later administrations kept PB in form while weakening it in substance, reports that from 2005 to 2010 an annual average of 43% of PB projects were completed, and argues that broad executive control over budget execution was the key mechanism¹¹.

Interpretation. Porto Alegre is both the strongest origin story for participatory budgeting and one of the clearest warnings about it. A process that depends on the goodwill of the executive that runs it can be hollowed out without being formally abolished. The national study adds a second explanation: when budgets tighten, there is less discretionary money to allocate, and PB becomes harder to sustain⁹. The two explanations are not mutually exclusive.

Case 5: Ireland, citizens' assemblies before referendums

Evidence. Ireland's 2016 to 2018 Citizens' Assembly had 99 citizen members plus a chairperson, selected by a polling firm using random sampling points and quotas for gender, age and social class, from people on the electoral register¹². Ireland has run several assemblies since a Convention on the Constitution, each established by resolution of both Houses of the Oireachtas, on topics including the Eighth Amendment, gender equality, biodiversity loss and drug use¹³. A peer-reviewed account describes the two early mini-publics as playing "a significant role in supporting key referendums," naming marriage equality in 2015 and abortion in 2018¹⁴. The 2018 referendum on the Thirty-sixth Amendment was approved with 66.40% yes¹⁵.

Limit: recruitment error. In February 2018 the assembly itself reported that seven replacement members had been recruited by phone through friends and family of a recruiter, rather than door to door as the agreed method required¹⁶. As a result, the majority view on four questions from its January 2018 meeting could not be determined, and the seven members took no further part¹⁶.

Limit: recommendations and voters can diverge. The Citizens' Assembly on Gender Equality recommended that Article 41.2 of the constitution be deleted and replaced¹⁷. On March 8, 2024, voters rejected the two related government proposals, with 67.7% voting no on the family amendment and 73.9% voting no on the care amendment, on turnout of about 44%¹⁸. The sources consulted for this page do not establish how closely the government's wording followed the assembly's recommendation, so that question is left open below.

Interpretation. Ireland shows that a well-run assembly can prepare a public for a decision, and that its recommendations still pass through government drafting and a public vote, either of which can change or reject them. It also shows that the legitimacy of sortition depends on a recruitment method that is followed and checked; the Irish error was found by the assembly's own randomized check¹⁶.

Competing interpretations

• "These cases prove participation works." The evidence supports narrower claims: Swiss votes are frequent and durable¹; Brazilian PB is associated with spending shifts and health improvements in one major study¹⁰; Irish assemblies preceded referendums they helped frame¹⁴. None of the sources shows that participation reliably produces better policy in every setting.
• "These cases prove participation fails." Also too broad. Low turnout in Reykjavik⁸ and nonbinding status in Taiwan⁶ are limits of specific designs, not proof that participation cannot work.
• Authority view. Several failures here are about who holds power after the input is collected: an executive in Porto Alegre¹¹, a government in Taiwan⁶. On this view, the critical design question is legal authority and follow-through, covered in proposals and governance.
• Rights view. The Swiss suffrage and minaret cases suggest that the most important design question is what majorities may not decide, and who enforces that limit²³⁴.
• Representativeness view. Turnout of 12%⁸ or 40%¹ raises the question of whose preferences are being counted. Sortition answers this one way¹²; open participation answers it another; each has costs.

Application to this framework

This is design proposal and interpretation, not a finding.

• Collective priorities are not mandates. The cases show advisory outputs being ignored⁶ and binding votes being bounded by law⁴. My Vote Forward treats published collective priorities as information, not instructions to any institution.
• Report the denominator every time. Reykjavik's turnout figure is useful precisely because the city published both numbers⁸. Any release should state who could have taken part and who did; see published releases.
• Plan for capture and drift. Porto Alegre's experience¹¹ argues for written rules about who controls the agenda and how a process can be changed; see governance.
• Check selection, publicly. The Irish recruitment error¹⁶ is an example of a check that worked. Any sampling or verification method should be auditable; see audits and verification.
• Rights sit outside the priority list. A shared priority does not license a majority to decide another group's rights; see representation.
• No case here is evidence that My Vote Forward works. The framework has not been evaluated; see evaluation.

Unresolved questions

• How closely did the 2024 Irish care and family wording follow the Citizens' Assembly recommendation, and did any difference affect the result? Not established by the sources consulted here.
• What independent evaluation exists of vTaiwan's 80% action figure, and which issues were never referred to it?
• Did Reykjavik turnout change after 2018? Later figures were not verified for this page.
• Can a participatory budget be made durable against a hostile executive without making it rigid?
• How should a system report outcomes of participation when the group that took part is small and self-selected?