voter_engagement v1.0.0

name: voter_engagement description: Teach and run nonpartisan voter engagement work in plain language: plan a registration or turnout effort backward from verified state deadlines, stay inside the nonpartisan and contact law lines, use door, phone, and text scripts that build a concrete plan to vote and respect opt outs at once, and measure contact quality (completed conversations, plans made, registrations confirmed) instead of raw attempt counts. Ships a contact quality checker, full scripts, a compliance checklist with a state fact sheet template, and a worked campaign plan. Use when planning a voter registration drive or get out the vote effort, training canvassers or phone and text bankers, explaining registration and turnout to a new team or class, writing voter contact scripts, or reporting what a voter contact programme actually achieved.

Voter Engagement

Purpose

Teach voter engagement as a learnable job with three parts: tell people true things about how to register and vote, help them make a concrete plan to do it, and count honestly what the contact produced.

This skill covers nonpartisan registration and turnout work in the United States: drives, canvassing, phone and text banking, campus and workplace tabling, and reminder programmes run by nonprofits, schools, unions, congregations, and neighbour groups. It teaches the work in plain language for people doing it for the first time. It does not run candidate or party campaigns, does not write persuasion copy for or against anyone, and is not legal advice. Where a step touches law, the skill points to the rule and its source and tells the user to get the plan reviewed when an organisation's tax status or a paid programme is involved.

Two findings shape the method. First, contact quality beats contact volume in the field evidence. In six local election experiments summarised by J-PAL, door to door canvassing raised turnout by 2.1 percentage points on average across everyone assigned a visit, and by 7.1 points among people actually reached, while contact rates ran only 14 to 45 percent. Second, a plan beats an intention. In the Nickerson and Rogers experiment, voters who stated a concrete plan (time, starting point, route) turned out at higher rates, while a standard encouragement script had no effect in that study. So this skill trains people to have real conversations that end in a plan, and measures those, not dials.

Workflow

1. Fix the kind of effort and who owns the legal risk

• Name the effort in one line: registration, turnout, or both, for which election, run by what kind of organisation. A 501(c)(3), a campaign, and a group of neighbours face different rules, and the scripts and targeting that are legal for one can be prohibited for another.
• If the organisation is a 501(c)(3), apply the IRS line from the start: registration and turnout work is allowed only in a neutral, nonpartisan manner, with no reference to any candidate or political party, and no campaign intervention of any kind. Read references/compliance-and-fact-sheet.md section A before any planning.
• Assign one named owner for compliance and one for data. Voter contact fails in the hand offs: forms nobody delivered, opt outs nobody suppressed, deadlines nobody owned.
• Write the neutral reason for every audience choice: geography, recent moves, registration status, turnout history, language need. If a targeting choice can only be explained by how people are expected to vote, a nonpartisan effort does not make that choice.

2. Build the state fact sheet before writing a single script

• Copy the fact sheet template in references/compliance-and-fact-sheet.md section E and fill every field for the state and election, from official sources in this order: state election office, local election office, vote.gov, the U.S. Election Assistance Commission.
• Treat deadlines as state facts, never national ones. Vote.gov states plainly that there is no national registration deadline: some states close registration 30 days before Election Day, others allow it on Election Day. The same spread applies to ID rules, mail ballot rules, and early voting. A script with another state's deadline in it is misinformation with good intentions.
• Use the National Mail Voter Registration Form only as its own state instructions allow. The EAC notes the federal form cannot be used the same way in every state, with North Dakota, New Hampshire, and Wyoming the named exceptions in its guidance. Read the form's state page, not a summary.
• Recheck the sheet at each phase change and after any election office announcement. Date and initial every check. No fact sheet, no contact: a campaign that cannot fill the sheet is not ready to talk to voters.

3. Plan backward from the deadlines

• Put the fixed dates on a wall calendar first: registration deadline by channel, mail ballot request and return deadlines, early voting window, Election Day and poll hours. Work backward from each one, as in references/worked-campaign-plan.md.
• Split the work into a registration phase that ends at the registration deadline and a turnout phase that ends when polls close. Mixing the two late produces the classic failure: beautifully planned turnout conversations with people who were never registered in time.
• Size the plan from real hours. State volunteer hours, shift length, and a realistic contact rate, then promise attempts and contacts from that math. Published canvassing contact rates of 14 to 45 percent are a sane planning range for doors. A plan that assumes everyone answers is fiction.
• Build the barrier list for the audience before contact starts: ID confusion, work hours, transport, childcare, language, disability access, polling place moves, mail delivery problems. Each barrier gets a named resource or referral, drawn from the fact sheet, or it stays on the list as unsolved and the plan says so.

4. Run registration work by the handling rules

• Train every collector on section B of the compliance reference and log the training. The core rules: the voter fills and signs their own form, help is given to anyone who asks regardless of any answer, every completed form goes to the election office by the fact sheet deadline and method including forms that look wrong, and no copies or photos of forms or ID numbers are kept.
• Never pay per completed form and check state rules before paying registration workers at all. Per form pay rewards volume, and volume pressure is how fake forms happen.
• Offer the official online route alongside paper. Vote.gov routes to the state system and does not itself register anyone or store their data. Say that plainly when a voter asks where their information goes.
• Follow up on confirmation, with permission. A helped form is a workflow count. A registration counts when the voter or the official lookup confirms it. Build the confirmation check into the timeline while there is still time to fix a rejected form.

5. Run turnout contact as a plan making conversation

• Use the scripts in references/contact-scripts.md. The turnout script is built around four plan questions: what time will you vote, where will you be coming from, how will you get there, and what will you be doing right before. Reflect the answers back as one sentence, then name the backup route (early voting or mail) from the fact sheet.
• Offer a pledge only as the voter's own words, never as a pressure device. In 2016 randomized trials, pledging raised turnout by 3.7 points overall and 5.6 points among first time voters compared with standard materials, which is a reason to ask for a clear yes, not a reason to push a reluctant person into one.
• Keep relational contact relational: volunteers contact people they actually know, with the short script in the reference. Do not rebrand a cold list as a friend list.
• Close every conversation with the official source for changes: where to check a polling place, the election office phone number, and what to do if a name is not on the list, which is to ask for a provisional ballot and the written follow up information, per HAVA Section 302.
• Include access in every phase, not as an afterthought: materials in the languages the area needs (check Voting Rights Act Section 203 coverage), a reminder that a voter who needs help reading or marking a ballot may bring a person of their choice under Section 208, with the employer and union exceptions stated, and venues and shifts planned around disability access.

6. Stay inside the contact law lines for calls and texts

• Work section C of the compliance reference before any phone or text programme starts. The short version: automated calls and texts to cell phones need prior express consent with a recorded source, every message identifies the organisation, and campaign contact is exempt from the National Do Not Call Registry but not from consent and opt out rules.
• Honour every opt out at once and in every channel. A revocation can come in any reasonable form, including replying stop, quit, end, revoke, opt out, cancel, or unsubscribe, or saying do not contact me to a canvasser. Suppress the same day across calls and texts. The FCC outer limit of ten business days is a legal ceiling, not a service standard.
• Set and respect quiet hours in the recipient's local time, and never send a reminder the person did not ask for.
• Log complaints and opt out spikes as quality signals, not as noise. A spike after a send means the list, the consent basis, or the wording is wrong, and the programme pauses that channel until the cause is found.

7. Measure contact quality, not raw counts

• Log every attempt in the CSV shape the checker expects: one row per attempt, with channel, attempted, contacted, completed, registration helped and confirmed, plan made, opt out, wrong number, and barrier reported as 0 or 1 flags.
• Run the checker after each phase and before any report leaves the building:

python3 scripts/contact_quality.py contacts.csv
python3 scripts/contact_quality.py contacts.csv --check-only

• Report in this order: completed conversations, plans made, registrations confirmed, then attempts. Attempts are effort and belong in a footnote. A report that leads with attempts is hiding the contact rate.
• Read the rates as diagnosis. Low contact rate points at list quality, hours, or turf. Low completion rate points at the open and the script. Low plan rate points at conversations ending at a vague yes. High wrong number share points at a stale list. Fix the cause the rate names before buying more volume.
• Separate what the programme produced from what it caused. Outputs (plans, confirmations) are facts. A turnout effect needs a comparison: a holdout group, a matched untargeted area, or a published study cited as context with its setting named. Without one, say the effect is unknown. Never divide votes by attempts and call the result your impact.
• Keep a barrier log as a deliverable in its own right. Counts of ID problems, moved polling places, and access failures, sent to the election office after the election, are often the most useful thing a small programme produces.

8. Debrief and hand the next cycle a clean start

• Run the debrief questions in the worked plan within two weeks, while the team still remembers the shifts.
• Archive the suppression list, the fact sheet with its check dates, the training log, and the barrier log together. The next cycle inherits the opt outs. Losing them means contacting people who already said no, which is both a compliance failure and a trust failure.
• Report suspected intimidation or false election information to the election office and the Department of Justice Civil Rights Division, and record that the report was made. Volunteers report and step back. They never confront.

Output Contract

For a voter engagement effort, produce:

• The effort statement: kind of organisation, election, phases, the named compliance and data owners, and the neutral reason for each audience choice.
• The completed state fact sheet, every field sourced from an official election source with a checked date, following the template in references/compliance-and-fact-sheet.md.
• The backward timeline sized from real volunteer hours and a stated contact rate range, with the registration and turnout phases separated at the registration deadline.
• The training record and the handling rules used for registration forms, including the delivery log format.
• The scripts in use, with every fact in them traceable to the fact sheet, following references/contact-scripts.md.
• The contact log CSV and the checker output for each phase, plus the report written in the order the workflow sets: completed conversations, plans, confirmed registrations, attempts last.
• The barrier log and the debrief note, including any comparison used for a turnout effect claim or an explicit statement that no comparison exists.

Operating Rules

• Official sources only for election facts. If a deadline, hour, place, ID rule, or eligibility claim is not on the checked fact sheet, it does not go in a script, a post, or a conversation.
• Nonpartisan means the whole design, not just the script: targeting, materials, venues, and partners are held to the same line as the words at the door.
• The voter decides. Help is offered, never imposed. A clear no ends the contact, gets logged, and is honoured in every channel.
• A form helped is not a voter registered. Only a confirmation counts, and only election officials decide eligibility.
• Never state a turnout effect without a comparison. Outputs are reportable facts, impact is a claim that needs evidence.
• Ballot choices are secret. Never ask how a person voted in a way that pressures an answer, never record a choice, and correct anyone who suggests a vote can be checked.
• This skill teaches United States federal baselines plus the habit of checking state law. It does not replace a lawyer, an election official, or the state guide for the state in question.
• No em dashes in any output this skill produces. Use commas, periods, or colons instead.