Priorities

A priority says what matters most to you right now. "Schools matter more to me than roads this year." It does not say what you want done about schools. It does not say how much money you would spend. It is a statement about attention.

Priorities are the part of My Vote Forward that can be shown in public, but only as a group total. That makes them useful: a city can see which subjects many people want addressed. It also makes them easy to bend. The list of subjects, the order they appear in, and any pre-filled answer can all change the result. This chapter explains how, and what the framework does about it.

Defined terms

Priority. A judgment of relative importance across a set of subjects at a given time.

Subject list. The set of subjects a person is asked to compare. In the demo this is a starting list of ten subjects. That number is a demo choice, not a rule of the framework.

Ranking. An ordering of subjects from most to least important. Rankings are ordinal: they show order, not distance.

Resource allocation. A division of a fixed amount (tokens, points, or money) across subjects. Allocation expresses relative size and forces tradeoffs. It is a different question from ranking. See Tradeoffs.

Agenda effect. Any influence on answers that comes from what is included, excluded, emphasized, or ordered, rather than from the respondent's own views.

Default. A value that is recorded if the person does nothing.

Confirmed selection. A value the person actively chose or actively confirmed.

Ranking versus allocation

Ranking and allocation are often treated as two ways of asking the same thing. They are not.

A ranking tells you that a person puts housing above transit. It does not tell you whether housing matters slightly more or overwhelmingly more. Two people with identical rankings may feel very differently about the gap. Averaging ranks across people therefore needs care, because a rank of 1 versus 2 may mean a small or a large difference. See Measurement.

An allocation asks a person to divide a fixed amount. If you have 100 points and give 60 to housing and 5 to transit, the gap is visible. Allocation also brings in scarcity: giving more to one subject means giving less to another.

Researchers working on participatory budgeting have studied this difference in a closely related setting. Goel, Krishnaswamy, Sakshuwong, and Aitamurto describe the method commonly used in practice, which they call k-approval, as approval voting "with a cap on the number of alternatives a voter can choose." They propose instead "Knapsack Voting," in which each voter "has to adhere to the budget constraint, thus keeping costs in mind while giving her vote" ¹. Their work concerns choosing funded projects, not ranking subjects, but it shows the same structural point: a method that ignores constraints asks a different question from one that imposes them.

The framework keeps ranking and allocation separate. A priority ranking answers "what should get attention?" An allocation of a defined pool answers "how would you divide this specific resource?" Neither answers "does this program work?", which is a third question. See Tradeoffs.

Hypothetical example. A resident ranks public safety first and parks fifth. Asked to divide a fictional pool of new funding, she gives parks the largest share, because she believes safety is already well funded and parks are not. Both answers are coherent. Reading her ranking as a budget request would misstate her view.

The subject list shapes the answer

Whoever writes the list of subjects has power over what is counted.

Survey research shows how strong this effect can be. In a 2008 Pew Research Center experiment on the most important issue in the presidential vote, 58% chose "the economy" when it was offered as an option in a closed list, while only 35% named it when asked an open question. In the closed version, fewer than 8% gave an answer outside the five options offered; in the open version, 43% did ². The list did not just record views. It shaped them.

There is a related, older finding about public attention more broadly. In 1972 McCombs and Shaw argued that news media, through what they cover and how prominently, may set "the 'agenda'" of a campaign: audiences learn how much importance to attach to an issue from the amount of coverage it receives ³. Whatever its size in a given setting, this agenda-setting effect means that people arrive at a priority list already influenced by what they have been shown.

The framework's response is not to pretend the list is neutral. It is to make the list accountable:

• The subject list is versioned and published, with the reasons each subject is included.
• People can see what is not on the list and can suggest additions through a structured process. See Governance.
• Subject order is randomized or rotated where that is practical, and the method is stated.
• Results are always reported with the list version that produced them.

Open questions remain about who decides the list and how. These are treated in Who chooses the subjects? and Governance.

Defaults versus confirmed selections

A default is powerful because doing nothing is easy. Research on organ donation is the best-known example. Johnson and Goldstein, writing in Science in 2003, argued that defaults impose costs on anyone who wants to change them and estimated that changing U.S. organ-donor defaults could yield "additional thousands of donors a year" ⁴. The setting is very different from civic priorities, but the mechanism is general: whatever is pre-filled tends to stick.

For priorities, this creates a specific risk. If a priority screen starts with subjects in some order, and a person leaves without changing it, that order is not the person's view. If it is counted as one, the published result partly reflects the designer's starting order.

The framework therefore separates three states:

1. Untouched. The person saw the screen and changed nothing. Not counted as a priority.
2. Edited but unconfirmed. The person moved things but did not confirm. Not counted.
3. Confirmed. The person actively confirmed the ranking or allocation. Counted, with the date and list version.

In an allocation, a zero is a confirmed choice, not missing data, but only once the allocation is confirmed. See Measurement for how missing and incomplete responses are reported.

Group priorities and their limits

Group priorities are the one part of the framework that can be shown publicly, because they describe subjects, not people. Even so, several limits apply.

Aggregation is a choice. There is no neutral way to turn many rankings into one. The standard illustration is the Condorcet paradox: with three voters ranking A over B over C, B over C over A, and C over A over B, a majority prefers A to B, B to C, and C to A, so pairwise majority produces a cycle, not an order ⁵. Arrow's theorem generalizes the problem: with more than two alternatives, no aggregation rule meets a short list of reasonable conditions at once ⁵. Every published priority list must name its method. See Collective priorities.

Participants are not everyone. Voluntary participants are more engaged than the general public; Pew Research Center found that all nine online opt-in samples it tested included more politically and civically engaged people than benchmarks indicated ⁶. A group priority describes the people who took part. See Representation.

A priority is not a mandate. Many people ranking a subject first is a signal for attention. It does not commit any body to a policy, a budget, or a timeline. See Proposals.

Evidence

• Research result: Closed and open question formats produce very different issue priorities ².
• Research result (1972, election campaign coverage): media coverage influences the importance people attach to issues ³.
• Research result in a different domain: defaults strongly influence outcomes when changing them takes effort ⁴.
• Formal result: group aggregation of rankings involves unavoidable tradeoffs ⁵.
• Research result: opt-in participants overrepresent the civically engaged ⁶.
• Existing practice: cities already run structured priority and budgeting processes. Reykjavík's "My Neighbourhood" program divides 850 million krónur between districts every other year, with residents submitting ideas and then voting on a shortlisted ballot ⁷. The framework draws on such practice; it did not invent it. See Participation between elections.

Competing interpretations

Priorities as information. One view treats group priorities as useful information for officials, comparable to a structured version of the "most important problem" question that pollsters already ask ⁸. On this view, the main task is good measurement.

Priorities as power. Another view holds that any public priority list will be used as leverage, and that organized groups will mobilize to move it. On this view, the main task is resisting capture: limits on coordinated participation, transparency about who took part, and refusal to treat the list as a mandate. See Collective priorities.

Priorities as constructed. A third view, drawing on agenda-setting research, holds that priorities are partly produced by the process that asks for them. On this view, the framework should emphasize that results are conditional on the list and context, and should test how much the list moves answers.

These views are not exclusive. The framework adopts the safeguards each one implies.

Application to this framework

• Priorities are recorded only when confirmed. Untouched defaults are never counted.
• Ranking and allocation are separate questions, labeled separately, and reported separately.
• The subject list, its version, its order method, and the aggregation method are published with every result.
• Group priorities may be public. A person's own ranking is private unless they choose to share it.
• Results state the denominator and that they describe participants.
• No group priority is presented as a policy preference or a mandate.

Unresolved questions

• Who should write and revise the subject list, and what review should that process face? See Governance.
• How large are list and order effects in this specific format? This has not been measured. See Evaluation.
• Should people be able to add their own subjects, and if so, how are write-in subjects counted alongside listed ones?
• How should coordinated campaigns to move group priorities be detected without monitoring individuals? See Collective priorities.
• How often should priorities be refreshed, and how should old confirmations age?