Brief us

WP-32: Delivery

Half the money arrives after the minds are made up

In 2024, 51 percent of a record $11.1 billion in political advertising ran in the final eight weeks. By then, about four voters in five had decided. A campaign’s calendar is usually upside down.

By Brushworks research deskNovember 2026Working paper

Abstract. Campaign spending is built around the last eight weeks because that is when attention peaks and when the money arrives. Voter decisions are not. Exit polls in each of the last three presidential elections put the share deciding in the final week between 5 and 13 percent, and in 2024 about 80 percent had decided before September. The field-experimental literature adds that persuasion near Election Day is small and fades within days. We set out the mismatch, what it costs, and how a calendar built on when people decide reallocates the same budget.

In brief

The finding
Most of the money arrives after most of the decisions. In 2024, 51 percent of ad spending ran in the final eight weeks, when about one voter in six was still making up their mind.
For a campaign
Move persuasion forward, to the months when the undecided are still most of the people you could reach, and keep the closing weeks for the voters whose decision is turnout.
For a donor or a committee
Ask what share of the budget lands before Labor Day, and what the plan does for a voter who decided in June.

Every campaign knows the shape of its own spending. It rises through the summer, steepens after Labor Day and peaks in the last fortnight. The 2024 cycle took that shape to a record: of $11.1 billion spent on political advertising, 51 percent ran in the election’s final eight weeks.1 The shape of voter decision runs the other way. In the national exit poll that year, 80 percent of voters said they had decided before September.2 This paper sets out the mismatch, why it persists, and what a calendar built on when people decide does with the same money.

When voters decide

The national exit poll asks voters when they finally decided how to vote for president. In 2024, 3 percent said in the last few days, 3 percent in the last week, 5 percent in October, 6 percent in September, and 80 percent before that. Asked the question as a single choice, 7 percent put their decision in the final week.2

Most of the decisions were made before the campaign’s loudest months. The final eight weeks span roughly September through Election Day.

The share deciding in the final week has fallen across three elections. It was 13 percent in 2016, the year late deciders were blamed for the polling miss, 5 percent in 2020, and 7 percent in 2024.3, 4, 2 The national number hides some variation (in Wisconsin in 2024, 13 percent decided in the last month and 6 percent in the final week) but no battleground in 2024 came close to the late-decider share of 2016.2

The late decider is a smaller group than the campaign calendar assumes. National exit poll, share deciding in the final week before Election Day.

The few who decide late are not noise. In each of the last three elections the late deciders broke the same way, by 3 points in 2016, 14 in 2020 and 12 in 2024, according to the exit polls as reported at the time.3, 4, 5 A small group that leans together can decide a close state. It is still a small group, and it is a group that has been receiving everyone’s advertising since the spring.

The few late deciders leaned the same way three times running. Margin among voters deciding in the final week, as reported from the national exit poll.

Where the money goes

Political advertising follows the attention of the audience and the cash flow of the campaign. Both peak late. Fundraising compounds through the fall, outside groups commit after the conventions, and broadcast stations sell the final weeks at the highest rates of the cycle. AdImpact’s review of 2024 records the result: more than half of the cycle’s spending in its last eight weeks, two thirds of all spending in thirteen states, and a single state, Pennsylvania, above $1 billion on its own.1, 6

The two shares that a campaign calendar should reconcile. Spending share from AdImpact; decision share from the national exit poll (September or later).

Put the two series side by side and the gap is plain. In the window that carried half of the money, roughly one voter in six was still deciding. For a persuasion budget, the arithmetic is unforgiving. If the persuadable share of the audience falls by two thirds between June and October while the price of reaching that audience rises, the cost of each persuaded voter rises faster than either.

Why late persuasion buys so little

The field-experimental record says the same thing from a different direction. A meta-analysis of 49 field experiments found that the average persuasive effect of campaign contact in the two months before a general election was indistinguishable from zero. The same paper found measurable effects earlier in the cycle and in contests without a party cue, such as ballot measures and some primaries.7

Television effects, where they exist, are fleeting. In a randomized experiment that varied a governor’s advertising across Texas media markets, the effect on vote intention was sizable in the week the ads ran and had largely faded within a week or two after.8 A broad study of television advertising across U.S. elections finds real but small effects, larger for down-ballot races than for presidential ones.9

The shape the experiments report: a television effect that is visible in the week it airs and largely gone a week or two later. Values illustrative; shape from Gerber and colleagues (2011).

Late money is not wasted on everything. It is well spent on turnout among voters who have decided and have not yet voted, and on answering an attack that lands in the final days. It is poorly spent on persuading people who decided in July.

A calendar built on decisions

The alternative does not need a larger budget. It needs the budget to arrive in the order voters decide.

  • Persuade early, against a control. Put the persuasion dollars where the undecided are most of the audience: late spring through Labor Day. Measure every flight against a sealed holdout so that the early spend has to prove itself.
  • Read who has decided. Behavioral signals separate the voter who has settled from the voter who has not: what they read, whether they engaged with the race at all, how they voted last time. Stop persuading the settled.
  • Turn the close into mobilization. In the last eight weeks the persuadable pool is small and the turnout pool is large. Sequence the money so that most of it reaches people whose decision is whether to vote.
  • Hold a reserve for the attack. The one late persuasion job that pays is answering a claim before it settles. Keep money and creative ready for it (WP-23 sets out the desk that does this).
A persuasion calendar sequenced to the voter, from the first terrain read to the close. The order is the method; dates move with the race.

Objections

Attention peaks late, so ads work better late. Attention peaks, and so does clutter. The voter who is paying attention in October has usually decided, and the one who has not is hearing every campaign at once.

Early money is wasted because voters forget. Early persuasion does decay, which is why it is measured against a control and repeated. The decision it helps form does not decay at the same rate: four in five voters report a decision made before September.

Our donors give late. Then the plan has to borrow from the future: commit early cash to persuasion and late cash to turnout. The order of spending is a choice the campaign can make even when the order of receipts is not.

A worked example: two calendars, one budget

Take an invented statewide race with $10 million to spend on persuasion and turnout between June and Election Day. The figures in this example are illustrative; the shape follows the record above.

A conventional plan spends 15 percent before Labor Day and 85 percent after, rising each week to the close. A plan sequenced to the voter spends 45 percent before Labor Day, almost all of it on persuasion against a control, and 55 percent after, most of it on turnout among voters who have already decided.

Spending by month under the two plans, millions of dollars. Illustrative.

The two plans buy very different audiences. In June, a large share of the people a campaign can reach have not settled on a choice. By October, most have. A persuasion dollar spent in June is spent on an audience in which the undecided are common; the same dollar in October is spent on an audience in which they are rare and expensive to find.

Share of reachable likely voters still undecided, by month. Illustrative of the exit-poll pattern, not a measurement.

The sequenced plan is not cheaper. It buys the persuasion when persuasion is available, and it buys the turnout when turnout is the question. The control tells the campaign within weeks whether the early spending is working, which is the one thing a plan that saves its money for October can never learn in time.

Questions to ask of a media plan

  • What share of the persuasion budget is scheduled before Labor Day?
  • How does the plan decide that a voter has made up her mind, and what happens to her contact when she has?
  • What control will say whether the summer spending worked, and when will it report?
  • How much of the final eight weeks goes to people who have already voted?
  • What is held in reserve for a late attack, and who can release it?
  • If the race tightens in October, which voters does the plan expect to move, and what evidence says they still can?

Further reading

  • Kalla, J. L., and Broockman, D. E. (2018), cited above: the meta-analysis behind the finding that late persuasion in general elections averages near zero.
  • Gerber, A. S., Gimpel, J. G., Green, D. P., and Shaw, D. R. (2011), cited above: the randomized test of how quickly television effects fade.
  • Green, D. P., and Gerber, A. S. (2019). Get Out the Vote: How to Increase Voter Turnout, 4th ed. Brookings Institution Press: the evidence on what moves turnout in the final weeks.
  • Vavreck, L. (2009). The Message Matters: The Economy and Presidential Campaigns. Princeton University Press: why some campaigns can move a race and others cannot.

Implications

For a campaign. Write the calendar from the electorate, not from the cash flow. A dollar that reaches an undecided voter in July is worth more than one that reaches a decided voter in October.

For a committee or a donor. Ask for the share of persuasion spending scheduled before Labor Day, and for the control that will say whether it worked.

For both. The cheapest month to be heard is the one in which nobody else is speaking. In most cycles that month is June.

Notes and references

  1. AdImpact, Cycle in Review 2023–2024, December 20, 2024: total spending, media shares and the share in the final eight weeks; summarized by TV Tech.
  2. NBC News exit poll, 2024: “When did you finally decide for whom to vote in the presidential election?” National results as published.
  3. 2016 and 2020 shares deciding in the final week, Edison Research national exit poll, as reported by The Washington Post and NBC News on November 4, 2020.
  4. ABC News exit poll, 2020: national results by time of decision.
  5. NPR, November 12, 2024, reporting the 2024 exit poll result among voters deciding in the last week and the last few days.
  6. AdImpact, on Pennsylvania’s $1.2 billion in 2024 election advertising.
  7. Kalla, J. L., and Broockman, D. E. (2018). The minimal persuasive effects of campaign contact in general elections: evidence from 49 field experiments. American Political Science Review, 112(1), 148–166.
  8. Gerber, A. S., Gimpel, J. G., Green, D. P., and Shaw, D. R. (2011). How large and long-lasting are the persuasive effects of televised campaign ads? Results from a randomized field experiment. American Political Science Review, 105(1), 135–150.
  9. Sides, J., Vavreck, L., and Warshaw, C. (2022). The effect of television advertising in United States elections. American Political Science Review, 116(2), 702–718.

Basis of figures. Spending: AdImpact’s published review of the 2023–2024 cycle. Decision timing: the national exit poll as published by the networks; exit polls carry sampling error and are weighted to results. The decay, calendar and undecided-share exhibits are illustrative. No Brushworks program figures appear in this paper.

Next paperThe Review

Start a conversation

Brief us.

Tell us the race, the calendar and the margin. Within a week we will tell you whether we can move it and what it would take. Candidate work and independent-expenditure work are taken by separate teams; say which you are. Governments and institutions: tell us the mandate and the jurisdiction, and we will tell you what registration it requires before we say anything else.

Send a briefbrief@brwrks.com

  • Candidate practice
  • Independent-expenditure practice
  • International practice

Brushworks Strategies Inc., a Wyoming corporation · Washington, New York, Ottawa, London, Munich, Dubai, Tokyo