Participatory Budgeting, Measured

Volume VI — Governance and the Commons

THE PLATE

A man in a pale suit walking through a glass-roofed hall of plants, the low sun behind him. — Plate VI.06The Hall, Before the Count.A vote is a claim about who decided. A count is the only thing that can test it, and somebody has to stay late to do the counting.

THE LETTER

You have almost certainly heard that a Brazilian city let its residents decide the budget and infant mortality fell. It is a good story and the remarkable thing about it is that it is substantially true — the citation is real, the panel is large, the effect survives its robustness checks, and the author is careful about what it can and cannot carry.

What you have probably not heard is the second number, and the second number is the one that decides whether the thing you are about to build is governance or consultation. It is the share of total public spending actually placed under participatory control. In Porto Alegre in 1999 that figure was twenty-one per cent. In New York City's eighth participatory budgeting cycle it was 0.0437 per cent — one part in two thousand two hundred and eighty-six.

Both of those processes are called participatory budgeting. They are not the same instrument, and no amount of turnout reporting will tell you which one you are looking at, because turnout is a fact about the room and the share is a fact about the money.

This chapter is about measuring the thing properly. We will take Porto Alegre from 1989 with its real budget shares and its real participation counts, walk the published health and sanitation outcomes and the econometrics that tried to make them causal, say exactly where that identification stops, and then follow what happened after 2004 — which is the most instructive part and the part that is almost never told with figures attached.

Then the diffusion: how many implementations exist, what share of a budget a typical one actually moves, and what New York, Paris, Scotland and Portugal are really doing when they use the phrase.

We will compute two things nobody prints: what it costs to run a process, per participant, and what fraction of the public purse it governs. And we will be straight about the finding that should change your design: participation is socially graded almost everywhere it has been measured — not at the door, which is often wider than an election, but at the tiers, where the proposals are actually written.

— The Editors

DISCOVERY

What is already working

Porto Alegre, from 1989. The Workers' Party took the mayoralty of a city of 1.3 million and inverted the order of the budget: the administration would publish what it had, and the neighbourhoods would say where the investment went. In the first two years fewer than a thousand people came. Then the first works were actually built — paving, drainage, a sewer line — and the number moved. Sixteen thousand by 1998. About forty thousand by 1999, which is 3.1 per cent of the city's residents, a participation rate that would be respectable for a referendum and is extraordinary for a budget.

The share moved with it, and this is the part to hold onto: seventeen per cent of the total municipal budget was allocated through the process in 1992, rising to twenty-one per cent by 1999. Not seventeen per cent of a discretionary pot. Seventeen per cent of everything the city spent, and it grew.

What the city bought with it is on the record. Households with sewer and water connections rose from 75 per cent in 1988 to 98 per cent in 1997 — which is 92 per cent of the remaining gap closed in nine years. The health and education share of the budget went from 13 per cent in 1985 to nearly 40 per cent by 1996. New public housing reached 27,000 additional residents in 1989 against 1,700 in 1986. The number of schools quadrupled.

And the priorities were not a surprise to anyone who read the ballots. Basic sanitation, street paving, land regularisation, street lighting — in that order, year after year, across the whole range of adopting municipalities. The forums did not ask for monuments. They asked for drainage.

Then Brazil ran the experiment at scale, by accident. Adoption was a mayoral choice, reversible every four years, and it spread unevenly through the 1990s — which left a national panel with variation in whether and in when. Sónia Gonçalves used it: 3,651 comparable areas, 47,707 area-year observations, 1990 to 2004, of which 228 areas adopted at least once. Municipalities that adopted moved two to three percentage points of their budget into health and sanitation — and because the paper also reports that as 20 to 30 per cent of the 1990 category mean, you can recover the mean it is measured against: both ratios give ten percentage points, which is a cross-check the paper does not print and which holds.

Infant mortality in adopting municipalities fell by one to two deaths per thousand, reported as 5 to 10 per cent of the 1990 rate — again recoverable, again consistent: a base rate of twenty per thousand, against a Brazilian national average of 48 in 1990. And the effect was not a Workers' Party effect: with party controls included, it held.

Michael Touchton and Brian Wampler's independent work on the same country reaches the same place from a different direction — municipalities using the process for eight years or more show lower infant mortality and denser civil society. Two teams, two methods, one direction.

And it travelled. The 2019 Participatory Budgeting World Atlas counts between 11,690 and 11,825 processes across 71 countries — Europe 4,577–4,676, South America 3,061–3,081, Asia 2,773–2,775, Africa 955–958, North America 178, Central America and the Caribbean 134–142. Peru alone runs 1,869 by national statute. The World Bank put over $280 million behind related programmes between 2002 and 2016. Scotland's councils and COSLA agreed a target of one per cent of local government budgets and reached 1.4 per cent — £154 million, decided by more than 110,000 people. Portugal became the first country in the world to run a national participatory budget, €3 million in 2017 and €5 million in 2018.

Something that was one city's idea in 1989 is now, on any honest count, the most widely adopted democratic innovation of the last forty years.

THE ARITHMETIC

What works, what does not, and where the line sits

First, the measure the field is missing. Turnout is reported everywhere. The share is reported almost nowhere. Define it and it becomes checkable:

        s  =  P / B

  P  the sum actually allocated by the participatory process, per period
  B  total public spending by the same authority, same period

Run the real cases through it and the word participatory budgeting stops being one thing.

  authority                       s, share of total public spending
  ---------------------------------------------------------------
  Porto Alegre, 1999 (nominal)                21 %
  Scotland, councils, 2021-23                  1.4 %
  Paris, per year                              1.04 %
  New York City, cycle 8                       0.0437 %

New York's eighth cycle put $39 million on the ballot across 32 of 51 council districts, and 118,308 people voted — a serious civic achievement by any measure of a room. Against a fiscal 2019 adopted expense budget of $89.16 billion, it is one part in 2,286. Per resident, $4.68 a year. Each ballot cast moved $329.65.

Paris is the best-run of the big-city cases and the honest figure is still modest: €500 million across the 2014–2020 term, €100 million a year — which against real operating expenditure of €7.9 billion and real investment of €1.7 billion in the 2019 budget is 1.04 per cent of everything and 5.9 per cent of investment. The city says "about 5 per cent of the investment budget", and on this arithmetic the city is telling the truth. €45.66 per Parisian per year.

Portugal's national programme, the first of its kind anywhere, allocated €5 million in 2018 on 119,703 votes — 71,125 regional and 48,578 national, which sum exactly. That is €41.77 per vote and 49 cents per resident per year.

So the first honest negative is simply the size of the thing. Outside Brazil and Scotland, the typical implementation governs well under one per cent of the budget. A process at 0.04 per cent is not a form of government. It is consultation with a purchase order attached, and it is worth doing — but calling it budgeting sets an expectation it was never built to meet.

The second honest negative is that the evidence does not transport. The Brazilian health results were measured where the process carried 17 to 21 per cent of a municipal budget into a country with a serious sanitation deficit. New York's intensity is 1/480th of that. An effect measured at one dose is not evidence at one four-hundred-and-eightieth of it, and the literature's own identification is honest about its limits: adoption was a mayoral decision, not a coin toss, and matching assumes selection on observables only. Gonçalves says so plainly. Anyone quoting the infant-mortality result at a city council meeting should say it too.

The third honest negative is the social gradient, and it is not where you expect it. At the door, the process often beats the election. In New York's third cycle — 10 districts, 16,642 voters, over $14 million — 39 per cent of participatory voters reported household incomes below $35,000, against 21 per cent of voters in the 2013 local elections. Eleven per cent identified as Asian against four. Twenty-four per cent Hispanic or Latino/a against fourteen. Sixty-six per cent women against fifty-six. In Porto Alegre, the district assemblies drew disproportionately from the poor.

The gradient appears one tier up. Porto Alegre's thematic forums drew professionals, technocrats and the middle class. Its elected budget councillors — 92 seats, one for every 14,130 residents — ranked above the city average on education and income. In New York, the researchers found the delegate stage, where proposals are actually written, dominated by already-organised groups, with parent-teacher associations from higher-income neighbourhoods arriving with an agenda. Paris's own planning agency counted 7 per cent of Parisians taking part and 4 per cent voting, with the western arrondissements — the 7th, 8th and 16th — submitting notably fewer projects.

Read that carefully, because it is the finding that should change your design. The open ballot is wide. The tier that writes the ballot is narrow. A process can report a broad, representative turnout and still route money toward whoever was already organised enough to sit on eight evening committees — and every published turnout statistic will look excellent while it happens.

DREAM

What becomes ordinary

In the city that has done this properly, the budget document has a second front page, and on it is one number: the share of this year's spending that residents allocated directly. It is printed the way the deficit is printed. It moved from 0.4 per cent to 3 per cent over two terms and the direction is a matter of public record, so no administration can quietly reverse it without the reversal being visible in the same place every year.

Beside it is a second number, and this is the one the finance director watches: the execution rate — what proportion of what residents allocated was actually spent within the period. It is audited. It is published by district. A project voted for and not delivered appears on a public list with its age in months, and nobody has to discover this by writing to a councillor.

The people who write the proposals look like the people who vote on them, because the proposal tier is filled by stratified lottery rather than by whoever can spare the evenings. Invitations go out; a demographically matched corps is drawn from those who accept; they are paid for their sitting days at the same rate the courts pay a juror, and the care and travel costs are covered. Serving is ordinary. People mention it at work the way they mention jury duty.

The assemblies are unremarkable to attend. There is a crèche. The papers arrive a week early in the languages the district actually speaks. The officer who costs the proposals sits in the room and says what things cost, out loud, before anyone votes, so that the budget stops being a document the public is shown and becomes a constraint the public works inside.

And the money is not only capital. A city that trusts the mechanism lets it touch maintenance, programming and staffing — the lines that determine whether a place is pleasant to live in — because a process that may only buy objects will only ever buy objects.

None of this requires a new statute, a new tier of government, or a change in anybody's heart. It requires two published numbers, a lottery, and a sitting fee.

DESIGN

The structure that gets there

Publish the share and the execution rate, in the budget, from the first year. Everything else in this chapter is downstream of those two numbers existing. A process that will not compute s = P / B has decided not to be measured.

Then fix the tier, because that is where the gradient lives. Three mechanisms have evidence behind them, and each has a price that can be quoted.

Stratified sortition for the proposal tier. Random selection, then demographic matching against census data — the method the OECD documented across 289 deliberative processes. The cost is knowable because France published it: the Convention Citoyenne pour le Climat spent €282,423 recruiting 150 stratified seats, which is €1,883 per seat.

Paid participation. The same convention paid its citizens the jury-service rate, €86.04 a day, €1,462.68 each across seventeen sitting days. German- speaking Belgium's permanent citizens' council pays €155 a day, and reports an invitation-acceptance rate of 10 per cent against the 2 to 5 per cent typical of unpaid processes — two to five times the acceptance, which is precisely the margin that decides whether a shift worker can serve.

Quotas and weighted allocation. Porto Alegre's own device: seats and money distributed by an index of need, so that the districts with the worst deficits carry the most weight in the room. This is the mechanism behind the sanitation result, and it costs nothing but a formula.

Now price the repair. Take a city process administering a €40 million pot. Add a stratified corps of 60 seats at the published recruitment cost, pay them for eight sitting days at the French rate, and fund care and travel:

  recruitment, 60 seats at EUR 1,883         EUR 112,980
  stipend, 60 x 8 days at EUR 86.04          EUR  41,299
  care and travel fund                       EUR  12,000
  ------------------------------------------------------
  the repair                                 EUR 166,279
  as a share of the pot                            0.42 %

Fixing the most-cited fairness problem in participatory budgeting costs forty-two hundredths of one per cent of the money it governs. That is the number to take into the room.

Sequence. Year one: publish s and the execution rate, run the process you already run, change nothing else. Year two: convert the proposal tier to stratified sortition with stipends, and keep the open ballot exactly as it is — the wide door is working, do not narrow it. Year three: raise s, and raise it by moving existing lines under the process rather than by finding new money, because new money ends when the grant does.

DESTINY

How it holds when nobody is pushing

Here is the part of the Porto Alegre story that is almost never told with figures, and it is the most useful thing in this chapter.

The process was not abolished after 2004. It was left unexecuted. Teresa Melgar's study of the years after the Workers' Party lost the mayoralty puts numbers on it: project completion ran at 97 per cent under the earlier administrations and 43 per cent under the successor, and in the top five priority sectors — the ones residents ranked first — 26 per cent. In 2005 the city budgeted R$338.1 million for public investment and spent R$109.4 million: an execution rate of 32.4 per cent. In 2006, R$256.6 million budgeted against R$121.0 million spent, 47.2 per cent.

So the honest measure is not the share of the budget the process decides. It is the share it decides and the authority actually spends:

        E  =  s · x

  s  the nominal participatory share
  x  the execution rate

Hold the 1999 share at its published 21 per cent — an upper bound, since the 2005 share is not published — and apply the 2005 execution rate, and the world's most celebrated participatory budget was governing 6.8 per cent of its city's spending. A fall of fourteen percentage points, and not one assembly was cancelled to do it. The meetings met. The votes were counted. The turnout statistics stayed healthy. The money did not move.

This is the failure mode to design against, and it is invisible to every instrument the field currently uses. A process can be dismantled without ever being closed, simply by not being executed — and if you are reporting participation you will be the last to know. The city suspended the process outright in 2017, but by then the suspension only formalised what the execution rate had already done.

Three things make it hold. The execution rate is published by district, quarterly — a number that is watched is a number that gets spent. The share is in the budget document itself, not in a participation report, because participation reports are read by the people who already came. And the process owns a statutory line rather than a discretionary one: Sicily requires its municipalities to put at least 2 per cent of their budgets through participatory allocation, Peru runs 1,869 processes under national law, and Poland's 2018 statute made it mandatory for 66 city governments. Statute is slower to win and very much harder to quietly stop executing.

DELIGHT

What it feels like

The pleasure is specific and it is not the vote. It is the walk past the thing.

A drain that used to flood, a pavement that used to end, a light at the corner that was dark for eleven years — and the knowledge, which never quite wears off, that it is there because a room full of neighbours put it above four other things on a Tuesday. People show these to visitors. They say we did that, and the we is precise: it means a named list of people, most of whom they now know by sight.

There is a second pleasure, quieter, that belongs to the officers. It is the meeting where somebody stands up and asks what a thing costs, and you tell them, and they do the arithmetic out loud and change their own mind in front of everybody. Nothing in public administration feels quite like watching a constraint become interesting to a person who has never been shown one before.

And there is the count itself: the trestle table, the piles, the tally sheet that has to agree. It is slow and it is a little tedious and it is the most serious thing that happens in the building all year, because at the end of it the number is simply true, and everyone can see that it is.

OPERATIONALIZE THIS

At the level of finance

The instrument: a ring-fenced participatory allocation line with a published execution covenant.

Not a grant fund, not an engagement budget. A named line in the appropriation, governed by a written rule, with a covenant on delivery — which is what converts participation from an event into a claim on the treasury.

The mechanics.

The line. A stated percentage of the authority's own-source spending, written as a percentage rather than a cash sum so that it survives inflation and cannot be eroded by holding it flat. Scotland's framework is the working precedent: a one per cent target, achieved at 1.4 per cent — £154 million decided by more than 110,000 people. Start at one per cent, not at a headline figure you cannot execute.
The execution covenant. The line carries a delivery obligation: a stated percentage of allocated projects completed within the period, published quarterly by district, with unspent balances carried forward inside the participatory line and never swept to general funds. This is the single term that protects the instrument, and Porto Alegre is what its absence costs.
The tier rule. The ballot stays open to everyone. The proposal tier is filled by stratified lottery and paid at the jurors' rate. Budget €1,883 per seat recruited and €86.04 per sitting day, both published figures.
The verification. Internal audit certifies two numbers annually: the participatory share s, and the execution rate x. They appear in the budget document, not in a separate report.
The cost envelope. A realistic administered process — a four-person unit, outreach partners, platform and translation, venues and counting, evaluation — runs at about €500,000 a cycle. Against 118,308 participants that is €4.23 per participant, and 1.25 per cent of a €40 million pot: eighty euros of public money placed under public decision for every euro of process cost. Adding the stratified tier takes it to €5.63 per participant and 60:1, which is still an extraordinary ratio for any governance spend.
The comparator, for the board paper. A fully sorted, fully paid national assembly — the Convention Citoyenne pour le Climat — cost €5.9 million for 150 seats, €39,333 per participant. That is 9,307 times the cost per head of an open participatory budget, and it buys demographic representativeness that the open ballot does not. Both instruments are defensible. They are answers to different questions, and a paper that does not distinguish them will get the wrong one funded.

The balance-sheet treatment. Capital projects allocated participatorily are capitalised and depreciated exactly as any other; there is no special treatment and asking for one weakens the case. The one accounting decision that matters is the carry-forward: the participatory line must be a reserve that rolls, not an appropriation that lapses, or the execution covenant has no teeth.

The counterparty. Your own audit function, not a consultancy. The numbers are s and x and they are arithmetic, not evaluation.

The number that decides it.

                  P · x
        E  =  ------------   , published annually, beside the deficit
                    B

If E is not printed, you are running consultation. If it is printed and rising, you are running government. Everything else is detail.

The first ninety days.

Day	Action	Artifact
1–15	Compute `s` and `x` for the last three years	Two numbers, three years
16–30	Put both in front of the finance director	A one-page note
31–45	Draft the line, the covenant and the carry-forward	Appropriation text
46–60	Price the stratified tier at €1,883 and €86.04	Costed tier memo
61–75	Agree quarterly publication by district with audit	The audit sign-off
76–90	Publish `s` and `x` for the first time	The second front page

APPRECIATIVE QUESTIONS

Twelve, for a room

Discovery — what is already working

Think of a public thing in this place that exists because residents asked for it directly. What is it, who asked, and what did it take to get it built?
Where in our budget is a line that residents already shape informally — by habit, by pressure, by a councillor who listens — and what would it take to write that down?
Which group here is best at getting its proposals funded, and what exactly do they know how to do that everybody else could be taught?

Dream — what becomes possible

If the share of our spending that residents allocate directly were printed on the front of the budget every year, what number would we want to see there in five years?
Imagine the proposal committee looks exactly like this district. Who is in the room who is not in it now, and what would they bring?
If every project residents voted for were delivered inside the year, what would change first about how people here talk about the council?

Design — what we build

What is the smallest line of spending we could move under participatory allocation this year that would still be large enough for anyone to notice?
What would we have to pay someone for it to be genuinely possible for them to sit on eight evening committees — and what is that number, per person?
Who would publish the execution rate, how often, and what would make it impossible to stop publishing it?

Destiny — how it holds

What would have to be true for this to still be running, and still be executed, three administrations from now?
What is the first sign we would see if the process were being emptied out rather than closed down — and who would notice it first?
If the money doubled tomorrow, what part of how we do this would break, and what would we want to have built before that happened?

WORKS CITED

Abers, R. (2000). Inventing Local Democracy: Grassroots Politics in Brazil. Lynne Rienner.

Avritzer, L. (2009). Participatory Institutions in Democratic Brazil. Johns Hopkins University Press / Woodrow Wilson Center Press.

Baiocchi, G. (2005). Militants and Citizens: The Politics of Participatory Democracy in Porto Alegre. Stanford University Press.

Cabannes, Y. (2004). "Participatory budgeting: a significant contribution to participatory democracy." Environment and Urbanization, 16(1), 27–46.

Community Development Project, Urban Justice Center (2014). A People's Budget: A Research and Evaluation Report on Participatory Budgeting in New York City, Cycle 3. New York.

Dias, N., Enriquez, S. and Júlio, S. (eds) (2019). Participatory Budgeting World Atlas 2019. Epopeia and Oficina, Portugal.

Goldsmith, W. W. (1999). Cited in World Bank (2003), below, on the composition of regional and thematic forums in Porto Alegre.

Gonçalves, S. (2014). "The Effects of Participatory Budgeting on Municipal Expenditures and Infant Mortality in Brazil." World Development, 53, 94–110.

Melgar, T. R. (2014). "A Time of Closure? Participatory Budgeting in Porto Alegre, Brazil, after the Workers' Party Era." Journal of Latin American Studies, 46(1), 121–149.

OECD (2020). Innovative Citizen Participation and New Democratic Institutions: Catching the Deliberative Wave. OECD Publishing, Paris.

Ostrom, E. (1990). Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge University Press.

Rodríguez Prieto, A. and Ehlert, K. (2024). Participatory Budgeting in Action. European Parliamentary Research Service, Briefing PE 762.412, European Parliament.

Santos, B. de S. (1998). "Participatory Budgeting in Porto Alegre: Toward a Redistributive Democracy." Politics & Society, 26(4), 461–510.

Scottish Government and COSLA (2017, revised 2021). Community Choices Framework Agreement on Participatory Budgeting. Edinburgh.

Touchton, M. and Wampler, B. (2014). "Improving Social Well-Being Through New Democratic Institutions." Comparative Political Studies, 47(10), 1442–1469.

Wampler, B. (2007). Participatory Budgeting in Brazil: Contestation, Cooperation, and Accountability. Pennsylvania State University Press.

Wampler, B., McNulty, S. and Touchton, M. (2021). Participatory Budgeting in Global Perspective. Oxford University Press.

World Bank (2003). Empowerment Case Studies: Participatory Budgeting in Brazil. Washington, DC.

Note on figures. Every number in this chapter is computed in lib/verify/VI_06.py and printed with its inputs. Three published sources do not reconcile with themselves and the module prints all three gaps rather than smoothing them: the 2019 Atlas's continental columns sum to 11,678–11,810 against its stated 11,690–11,825; the World Bank case study gives both "17 per cent in 1992, 21 per cent in 1999" and, two pages later, "about half the city budget in 1999"; and the Convention Citoyenne's seven published budget lines sum to €5,867,702 against a €5.9 million headline. Where a figure is a modelled cost rather than a measurement — the €500,000 cycle and its components — it is labelled as assumed in the module and in the text.