Get-Out-the-Vote Contact Effectiveness: What a Contact Is Worth: GOTV effects by mode, priced

In brief

Priced per 100,000 people a campaign sets out to reach, a social-pressure letter buys more turnout than a door knock, and a knock buys more than a text, which beats a paid phone call. The ranking flips if you count per person actually reached, where the door comes first. Every tactic delivers between a third and two-thirds less in a high-turnout election than in a low-turnout one. Only one tactic, texting, has a published cost per vote.

How to use this

Before comparing two tactics, state the unit, because votes per person reached and votes per person targeted rank differently. Work out your own contact rate before trusting a per-contact figure, since a door knocked at realistic rates reaches far fewer people than its headline number suggests. Budget a high-turnout election against the high-salience prices and expect every contact to move fewer voters. Ask for the calculation behind any cost per vote before repeating it, as only one tactic has a traceable price. Treat a result from one pair of trials as a case to watch, not a price to plan on.

What the story is about

Mann and Haenschen (2024), in a peer-reviewed meta-analysis, pooled 240 randomised field experiments on knocking, calling, mailing and texting voters. The experiments are the set gathered for the 2019 edition of the standard textbook on the subject (Green and Gerber, 2019), so this is older evidence, sorted again in 2024. The authors split it by the kind of election each experiment ran in. Even-year US general elections with House seats on the ballot count as high salience, meaning big and loud. Primaries, local contests, run-offs and off-cycle specials count as low salience. One thing you will not find here is a price in money. Almost nobody publishes what a vote costs by tactic, so every price below is counted in votes, and the reason for that gap is worth its own paragraph after the price list.

The results come in two kinds of number. For canvassing and phone calls, the literature reports CACE: the effect on each person the campaign actually reached. For mail and texts it reports ITT: the effect on each person the campaign set out to reach, whether the message arrived or not. In a high-salience election, a conversation at the door raises turnout by 3.05 percentage points among the people reached, a volunteer phone call by 2.28 and a professional call by 0.74. Per person targeted, a social-pressure letter, one telling voters that their neighbours can see, or may be told, whether they voted, adds 1.38 points. Ordinary mail adds 0.28, and a cold text, sent to people the campaign has no connection with, adds 0.22. Moving from a low-salience election to a high-salience one, the effects shrink by between 33% and 62% (Mann and Haenschen, 2024).

Comparing the two kinds of number takes one more step. A campaign pays for every name on its list, whether it reaches that person or not. The share of the list it actually gets through to is its contact rate. CACE is ITT divided by the contact rate, so the per-person-reached figure is always the bigger one, and the lower the contact rate, the wider the gap. Mixing the two kinds of number has a price. A campaign that planned a high-salience election on the combined figures would come up between 19% and 40% short of the votes it expected. Social-pressure letters are also about five times as effective as ordinary mail, in big elections and small ones alike (Mann and Haenschen, 2024).

Real contact rates are low. Mann and Haenschen (2024) report that canvassers reach roughly 25% to 30% of the households on their list, and volunteer phone banks 10% to 12%. Professional call centres reached 36% of their lists in 1997 and 6% in 2018. For canvassing to match a social-pressure letter per person targeted, it would have to reach more than 45% of its list in a high-salience election and 53% in a low-salience one. Phone calls would need 60% and 67%. A professional call never catches up, even if every call got through. Put every tactic on one footing, then: votes gained per 100,000 people a campaign sets out to reach in a high-salience election. On that price list, a social-pressure letter buys about 1,380 votes, canvassing 762 to 915, warm texts (sent to people already on the campaign's own list) 680, ordinary mail 280, volunteer phone calls 228 to 274, cold texts 220 and professional calls 44. The conversion is our own arithmetic on the paper's figures, and it reproduces all four of the break-even points the authors state. Canvassing at realistic rates still beats a cold text, so a door worth knocking is worth knocking, and a household nobody can reach is worth a text.

Why is none of this priced in money? The effect side of the sum is published, tested and precise. The cost side almost never is. Exactly one tactic has a documented price: texting, in a 2018 programme run by Vote.org that cost $2.5 million. Of the 5,220,823 people assigned to receive its texts, 13,052 voted who otherwise would not have. The authors' own cost analyses put that at $28 to $60 per extra vote, depending on assumptions about targeting, staff time, message length and content (Haenschen, Mann and DeMora, 2026). That is one tactic, in one programme, from 2018. The per-vote figures that circulate for other tactics are older numbers taken from earlier editions of the textbook and repeated without the calculation behind them, and the organisations that evaluate campaigns keep their methods behind member logins. A number nobody can trace is not a price, which is why this article counts in votes.

Two very large recent trials have tested texting. In the 2019 European Parliament election, a pre-registered trial with more than 3,000,000 Swedish voters found that "a simple text message increases the likelihood of voting by 0.3 percentage points" (Lindgren, Nyman and Oskarsson, 2025). Pre-registered means the design and the analysis were published before the results came in. That 2018 programme was also a pre-registered trial, with 7.1 million voters in the US midterms. It found that any text raised turnout by 0.25 points, and by 0.43 when data, message and targeting were optimised together (Haenschen, Mann and DeMora, 2026). Both figures are for a single message. What happens when voters receive text after text is a separate question, covered in text fatigue.

Paid digital advertising is where the largest test found no effect on turnout at all. A US$8.9 million campaign ran social-media ads for eight months at 2 million persuadable voters in five battleground states during the 2020 US presidential election. On average it neither raised nor lowered turnout (Aggarwal et al., 2023). In the UK, three randomised digital campaigns to get under-registered groups onto the electoral roll did not move registration, "Despite wide reach and relatively high rates of engagement" (Unan et al., 2024). Roadside billboards did no better. After 298 billboards were randomised across the country in the 2020 US general election, the researchers concluded that "it appears that signage does little to raise turnout in high-salience elections" (Green et al., 2024).

The price list is the best available, and it has soft spots. The letter at the top rests on just five experiments in high-salience elections, so it could shift as new studies arrive. Every experiment in the pooled set comes from the United States, because the authors could not include studies from other countries. A large share of the set is unpublished reports by civic and political organisations, which are hard to find and may not be a fair sample. The door-knocking figures lean heavily on experiments run many years ago, and the authors warn that getting people to open the door has become harder since. Phone calls carry the same warning, because fewer people answer, more of them use mobiles and the rules on automated calls have tightened (Mann and Haenschen, 2024).

So what

So the answer depends on what is counted. Priced per 100,000 people a campaign sets out to reach, a social-pressure letter buys more turnout than a knock on the door, which buys more than a text, which buys more than a paid phone call. Each of them delivers between a third and nearly two-thirds less in a high-turnout election than in a low-turnout one. Among the tactics measured per person reached, the door comes first, which is why every turnout figure needs its unit stated next to it.

Every price here is votes per 100,000 people a campaign sets out to reach, and three things set it: the tactic, the share of the list the campaign actually gets through to, and the kind of election. All of these prices also come from strangers contacting strangers. Messages passed on by people the voter already knows are a different tactic, covered in relational organising.

For political parties

A party planning a general election should budget against the high-salience prices, and it will meet two ways of deciding what counts as high salience. The pooled prices go by the kind of ballot. A Norwegian study went by turnout itself: "We find the two experiments to be more effective in low/middle salience elections, with turnout around 60 percent, than in high salience elections with turnout higher than 75 percent" (Bergh and Christensen, 2024). That study is the source of the 75% turnout line. It is one country's measure, tested on texts and phone calls, and it is not the rule behind the pooled prices. Both definitions point the same way, though: the bigger the election, the less each contact moves. The turnout in all these experiments is counted from official records of who voted, matched to the campaign's list, which is how the voter file works.

For government

For anyone who could pay for better evidence, the gaps are the finding. Beyond the one texting figure, no source since 2022 publishes a traceable cost per vote for any of these tactics. The largest spend on record, the digital campaign above, produced no measurable average effect on turnout (Aggarwal et al., 2023). No large door-to-door turnout experiment has been published since 2022 either, so the most recommended tactic rests on the oldest evidence. Every comparison here runs across studies, and nobody has tested two tactics at equal budget inside one campaign. The letter at the top is also the tactic most exposed to backlash, and to changing views about what a public body may say about someone's voting record. A public body commissioning turnout research could close any of these gaps.

Case studies

Who sends the message turns out to be part of the message. Two randomised trials in the UK, run between 2019 and 2020, sent behaviourally designed texts encouraging people to register to vote. When a local authority sent them, the result was "an increased registration rate of eight percentage-points, which translates into a three percentage-point increase in voter turnout". When an issue-advocacy organisation with no previous involvement in registration sent them, its texts "neither increased voter registration, nor turnout, no matter whether the text message offered a personal follow-up conversation, or not" (Cheng-Matsuno et al., 2023). The authors say "this raises questions about the scope conditions of existing findings". It is one pair of trials and has not been replicated, which is why it sits here as a case rather than as a price.

References

Aggarwal, M. et al. (2023) 'A 2 million-person, campaign-wide field experiment shows how digital advertising affects voter turnout', Nature Human Behaviour, 7(3), pp. 332–341. Available at: https://doi.org/10.1038/s41562-022-01487-4 (Accessed: 10 September 2026).

Bergh, J. and Christensen, D.A. (2024) 'Getting out the vote in different electoral contexts: the effect of impersonal voter mobilization techniques in middle and high salience Norwegian elections', Journal of Elections, Public Opinion and Parties, 34(1), pp. 79–95. Available at: https://doi.org/10.1080/17457289.2022.2051149 (Accessed: 10 September 2026).

Cheng-Matsuno, V., Foos, F., John, P. and Unan, A. (2023) 'Do text messages increase voter registration? Evidence from RCTs with a local authority and an advocacy organisation in the UK', Electoral Studies, 81, 102572. Available at: https://doi.org/10.1016/j.electstud.2022.102572 (Accessed: 10 September 2026).

Green, D.P. and Gerber, A.S. (2019) Get Out the Vote: How to Increase Voter Turnout. 4th edn. Washington, DC: Brookings Institution Press. Available at: https://doi.org/10.5040/9780815750642 (Accessed: 10 September 2026).

Green, D.P., Ong, L., Rutherford, K. and Schein, A. (2024) 'Do billboard advertisements increase voter turnout? A large-scale field experiment', Quarterly Journal of Political Science, 19(3), pp. 307–330. Available at: https://doi.org/10.1561/100.00022062 (Accessed: 10 September 2026).

Haenschen, K., Mann, C.B. and DeMora, S.L. (2026) 'The effects of SMS mobilization, message, and timing on voter turnout', Political Behavior, published online 23 June 2026. Advance online publication. Available at: https://doi.org/10.1007/s11109-026-10121-2 (Accessed: 10 September 2026).

Lindgren, K.-O., Nyman, P. and Oskarsson, S. (2025) 'Examining voting spillover effects of text message reminders', British Journal of Political Science, 55, e21. Available at: https://doi.org/10.1017/s0007123424000504 (Accessed: 10 September 2026).

Mann, C.B. and Haenschen, K. (2024) 'A meta-analysis of voter mobilization tactics by electoral salience', Electoral Studies, 87, 102729. Available at: https://doi.org/10.1016/j.electstud.2023.102729 (Accessed: 10 September 2026). Calculations from this data were made for this article and are available upon request.

Unan, A., John, P., Foos, F. and Cheng-Matsuno, V. (2024) 'Null effects of social media ads on voter registration: three digital field experiments', Research & Politics, 11(1). Available at: https://doi.org/10.1177/20531680231225316 (Accessed: 10 September 2026).