How many votes does it take to win a local election?

Your win number is the number of votes that guarantees a win in a two-way race: expected ballots cast, times 50%, plus one. Expected ballots cast is registered voters in the district times the turnout rate you expect. Everything else in campaign planning hangs off this one number, so spend real time getting the turnout estimate right.

Then add a cushion. Turnout estimates are wrong by a few points in either direction, and a plan with no margin is a plan to lose by eleven votes. Most managers plan to 52% to 55% of expected ballots rather than 50% plus one, and they treat the extra as the price of not knowing the future. In a race with three or more candidates and no runoff, the winning share can fall below 50%, but you should still plan for a majority unless you have a strong reason to believe the field will stay split.

Comparable means the same kind of election on the same kind of date. A May city election in an odd year is not comparable to a November election in a presidential year, even in the same city. Pull ballots cast and registered voters for your exact jurisdiction from the county elections office for the last two or three elections of the same type, compute turnout for each, and average them. If the trend is clearly up or down, weight the most recent one.

Local turnout is lower than most people expect. Portland State University's Who Votes for Mayor project found that in the most recent mayoral elections in America's 30 largest cities, turnout of eligible citizens was below 20% in 15 of them. Council, school board, and special district races held off-cycle typically run lower still, while a local race consolidated onto a November even-year ballot inherits the turnout of the top of the ticket.

Cross-check the estimate against the voter file. Count the voters in your district who voted in three or four of the last four comparable elections; that group is your turnout floor, because those people show up regardless of the campaign. If your estimate is far below that count, your estimate is wrong. How to read a voter file explains how to build that count.

• Same election type, same month, same jurisdiction boundaries. If the district was redrawn, ask the county for turnout in the new lines.
• Adjust upward for a contested race at the top of the same ballot, a controversial bond, or a recall; adjust downward for an uncontested ballot.
• In vote-by-mail states, use the ballots returned figure, not the ballots mailed figure.

Suppose your council district has 12,000 registered voters and the last three council elections averaged 18% turnout. Expected ballots cast is 12,000 x 0.18 = 2,160. The bare win number is 2,160 x 0.5 + 1 = 1,081 votes. Planning to 55% gives 2,160 x 0.55 = 1,188 votes, and that is the number you build the field plan around.

The cushion matters because turnout moves. If a school bond on the same ballot lifts turnout to 21%, ballots cast become 2,520 and a majority becomes 1,261. Your 1,188 plan is now short. If turnout falls to 15%, ballots cast drop to 1,800 and 901 wins; your plan is comfortable. Plan for the higher case and treat the lower case as a gift.

School board races are usually larger, lower-turnout, and often at-large. Take a district with 40,000 registered voters electing trustees in an off-cycle election. Suppose the last two comparable elections drew 4,000 and 4,800 ballots (10% and 12%; illustrative figures, not a real district). Average 11% gives 4,400 expected ballots. A single-seat majority is 2,201; the 55% planning number is 2,420.

Two wrinkles change the math. First, if it is a vote-for-two at-large race, you do not need a majority; you need to finish in the top two. Look up how many votes the second-place finisher received last time and plan to beat that by 10%. Second, some voters who cast a ballot skip the school board contest at the bottom of the page, so the ballots cast in your contest will be lower than the total ballots cast. Use the contest total from the official results, not the top-line turnout.

In the council example, 12,000 people are registered and 2,160 will vote. A mailer to all 12,000 spends 82% of its budget on people who are not going to cast a ballot. That is the whole argument for targeting: your resources go to people who will vote, plus a carefully chosen slice of people who might.

The research says the same thing from the other direction. A meta-analysis of 49 field experiments by Kalla and Broockman found that the average persuasive effect of campaign contact in general elections is zero; the measurable gains from contact come from turning out people who already agree with you and from identifying who those people are. Field experiments in six cities' local elections by Green, Gerber and Nickerson found that a face-to-face canvass raised turnout among those actually contacted by about 7 percentage points, roughly one extra vote per 14 conversations. The lesson: find your supporters among likely voters, then make sure they vote.

This does not mean ignoring low-propensity voters entirely. A candidate with a natural base among voters who rarely turn out can win by mobilizing them, and Nickerson's household experiment found that about 60% of a canvass's turnout effect passes to the other voter in the household. But that is a deliberate strategy with a higher cost per vote, not a default.

Score every voter by how many of the last four comparable elections they participated in. The tiers that fall out are the working vocabulary of field operations, and they tell you which message each voter should get.

4-of-4 and 3-of-4 voters are going to vote. Your job with them is persuasion and identification, not turnout. 2-of-4 voters are the swing turnout group: they vote sometimes, and a contact from a campaign is the kind of thing that tips them. Once you know they support you, they are your highest-value get-out-the-vote target. 1-of-4 and 0-of-4 voters rarely show up; spend on them only when you have identified them as supporters or when your strategy is built on expanding the electorate.

Work backward from the planning number. Say you need 1,188 votes and you want to identify at least that many supporters before election day. Assume, for planning, that 45% of the voters you actually talk to will say they support you (a figure you will replace with real data after the first two weekends). You need 1,188 / 0.45 = 2,640 conversations. Green, Gerber and Nickerson's canvassers reached about 30% of the voters on their lists, so 2,640 conversations means about 8,800 door attempts. At a practitioner's rule of thumb of 30 doors per hour in a suburban district, that is roughly 290 volunteer hours, or about 100 three-hour shifts spread over the campaign.

Phones and texts fill in where doors cannot reach. Gerber and Green's New Haven experiment found that commercial phone-bank calls had no measurable effect on turnout while personal canvassing had a substantial one, so use phones for identification and reminders rather than as your main persuasion channel, and use volunteers reading a conversational script rather than a vendor reading a monotone one. Dale and Strauss found that a text-message reminder raised turnout by 3.0 percentage points in a 2006 nationwide experiment; texts are cheap and effective as reminders to identified supporters, but they require consent and carrier registration, covered in political texting rules.

Runoffs reset the math. If your state or city sends the top two to a second round when nobody reaches 50%, plan the first round to finish first or second and the second round as a fresh two-way race with its own, usually lower, turnout estimate. Ask the county what turnout was in the last runoff of the same type; it is often half of the first round.

Multi-seat at-large races reward a different strategy. When each voter has three votes for three seats, a single-shot supporter who votes only for you is worth more than a supporter who fills all three slots, and your win number is the third-place vote total, not a majority.

Universal vote-by-mail changes the calendar more than the number. Ballots go out two to four weeks before election day and a large share are returned early, so the win number has to be reached across a window, and the daily returned-ballot file tells you exactly how many of your supporters have already voted. Ballot chase in vote-by-mail states covers that loop.

1. 1.Portland State University, Who Votes for Mayor? (turnout below 20% in 15 of the 30 largest cities)
2. 2.Green, Gerber and Nickerson (2003), Getting Out the Vote in Local Elections: Six Door-to-Door Canvassing Experiments, Journal of Politics
3. 3.Gerber and Green (2000), The Effects of Canvassing, Telephone Calls, and Direct Mail on Voter Turnout, American Political Science Review
4. 4.Kalla and Broockman (2018), The Minimal Persuasive Effects of Campaign Contact in General Elections, American Political Science Review
5. 5.Nickerson (2008), Is Voting Contagious? Evidence from Two Field Experiments, American Political Science Review
6. 6.Dale and Strauss (2009), Don't Forget to Vote: Text Message Reminders as a Mobilization Tool, American Journal of Political Science
7. 7.Green and Gerber, Get Out the Vote: How to Increase Voter Turnout (Brookings, 2019 edition)