This project would turn a standard get-out-the-vote effort for the November 3, 2026 gubernatorial election into a controlled field experiment. The aim is to find out which voter-contact methods raise turnout most cost-effectively.

The target group is “Presidential-Only” Republicans: registered Republicans who voted in both the 2020 and 2024 presidential elections but skipped both the 2018 and 2022 gubernatorial elections. These voters already back Republican candidates when they turn out, so the task is mobilization rather than persuasion. That makes them the most efficient source of additional votes in a non-presidential year. Before the groups are drawn, the list is cleaned to remove moved or inactive voters, duplicates, and anyone who wasn’t eligible to vote in 2018 or 2022.

The universe is split at random, by household, into five equal groups. Four groups each receive one contact method, all carrying the same core message about the election date, early voting, and mail-in ballots:

• Group 1: peer-to-peer text messages.
• Group 2: direct mail.
• Group 3: a voicemail from a fellow Republican in the voter’s own community.
• Group 4: door-to-door canvassing with a printed leave-behind.
• Group 5: a control group that gets no contact.

A true random control group is recommended, rather than one built from voters who couldn’t be reached, to keep the comparison unbiased. Group sizes matter too: detecting a 2-point turnout lift takes roughly 9,500 voters per group. With a smaller universe, the experiment should test fewer methods.

After the election, we match official vote-history records to the group lists. The analysis compares each group’s turnout rate, its lift over the control group, and the number of additional voters mobilized. The most important measure is cost per additional vote, because a method with a bigger effect can still be the worse investment if it costs far more.

The result is local evidence about what works with the campaign’s own voters. Future turnout budgets can then go to the proven methods, and the same design can be repeated in 2028 and later cycles.

project details

[ DOWNLOAD PLAN PDF ]

2026 FIELD EXPERIMENT

Targeted Republican Voter Turnout Experiment

A measurable field test of four voter-contact methods among “Presidential-Only” Republican voters in the 2026 gubernatorial election

Executive Summary

This project aims to identify the most effective and cost-efficient methods for increasing turnout among Republican voters who reliably participate in presidential elections but have historically sat out non-presidential elections.

Rather than assuming which voter-contact method works best, the program tests four approaches at the same time against a no-contact control group, then measures actual turnout from official vote-history records after the election. The result is hard, local evidence—drawn from our own voters—about what each method delivers and what each additional vote costs.

1. Purpose and Background

1.1 The drop-off problem

Every party faces “surge and decline”: turnout rises in presidential years and falls in midterm and gubernatorial years. The voters who make up that difference are not unknown or disengaged citizens—they are registered, have a vote history, and have already shown they will turn out when the stakes feel high. For Republicans in a gubernatorial year, they represent the most efficient source of additional votes available.

Because these voters already support the party’s candidates when they vote, the task is purely mobilization rather than persuasion. The challenge is that no one knows, for this particular group of voters, which form of contact is actually effective, and campaigns tend to fall back on habit, vendor claims, or anecdote.

1.2 Why run an experiment

Published academic field experiments on turnout have produced useful general lessons—personal, face-to-face contact tends to outperform impersonal contact—but results vary widely by state, electorate, message, and year. A local experiment answers the question that actually matters to us: what works for our voters, in our election, at our costs.

• Evidence over assumption. Decisions about next cycle’s budget rest on measured results, not intuition.
• Cost discipline. The key output is cost per additional vote, which lets cheap-but-modest methods be compared fairly with expensive-but-powerful ones.
• Low marginal cost. The contact work is GOTV that would be done anyway; the added cost is mostly planning, randomization, and post-election analysis.
• Institutional knowledge. Findings persist beyond any single campaign and can be shared with allied organizations.

2. Identify the Target Universe

Develop a list of Republican voters who meet all four of the following vote-history conditions:

These voters can be characterized as “Presidential-Only” Republican voters: they have shown they will vote in presidential elections but have not participated in gubernatorial elections.

2.1 Data sources

• State voter file with party registration and individual vote history (via the state party, the state board of elections, or a party data vendor).
• Mobile phone append for the text and voicemail groups; landlines can be used for voicemail.
• Address standardization (NCOA / CASS) for mail and walk lists.

2.2 Clean-up and exclusions

Before any groups are drawn, remove records that would add noise or waste resources. Apply every exclusion before randomization so all five groups are treated identically:

• Voters flagged as inactive, moved, or deceased, and anyone not registered at a current in-state address.
• Voters who were not old enough or not registered in 2018 or 2022. Otherwise, newly eligible voters would be misclassified as having “skipped” those elections.
• Duplicate records.
• Optional: voters who have already requested or returned a mail ballot for 2026, if that data is available at the time the list is built.

3. Experimental Design

The target universe is divided into five groups of approximately equal size. Four groups receive a different contact method; the fifth receives no contact. Because the groups are formed at random, they will be statistically alike in age, geography, and past behavior, so any difference in turnout after the election can be credited to the contact method itself.

3.1 Random assignment

• Randomize by household, not individual. Voters in the same home talk to each other and share mail. Assign every voter in a household to the same group to prevent one group’s contact from leaking into another.
• Stratify before randomizing. Sort by county or precinct, age band, and household size, then assign randomly within each block. This guarantees balance on the factors that matter most.
• Lock the lists. Record the random seed and freeze each group’s file before any contact begins. Group membership must not change afterwards.
• Check balance. Confirm that the groups have similar averages on age, gender, geography, and vote history before launch.

3.2 How big must each group be?

Turnout effects from a single GOTV contact are often only a few percentage points, so each group needs to be large enough to detect that difference reliably. The table below shows the approximate number of voters needed per group (to compare one treatment with control at 80% power and 95% confidence, assuming a baseline turnout near 40%):

Implication: If the universe is small, reduce the number of treatment arms (for example, test door-to-door and text against control) rather than spreading voters thinly across all four. Door-to-door is also limited by volunteer capacity; its group can be made smaller and geographically concentrated, as long as its matching control voters come from the same areas.

3.3 Protecting the control group

The control group is essential. It will be tempting to contact these voters, but without them there is no baseline, and none of the other results can be interpreted. Control voters must be kept off all program-generated lists: text, mail, calls, walk lists, and any additional “chase” efforts.

4. The Four Treatments

All four treatment groups receive the same core message, so that the experiment tests the delivery channel rather than differences in wording. Each channel is detailed below.

*Planning ranges only. Obtain real quotes from vendors; actual costs vary by market, volume, and volunteer vs. paid labor.

Group 1 — Text Messaging

Conduct a targeted text-message campaign using mobile telephone numbers appended to the voter file. Messages would emphasize the importance of voting in the gubernatorial election and provide information about voting options, including early voting and mail-in voting.

• Cadence: (1) a mail-ballot request reminder around the application deadline; (2) an early-voting reminder when early voting opens; (3) an Election Day reminder with a polling-place lookup link.
• Format: Peer-to-peer platform with a volunteer or staff sender name, a two-way reply capability, and a clear opt-out.
• Track: Numbers attempted, messages delivered, replies, opt-outs, and wrong numbers.

Group 2 — Direct Mail

Run a targeted direct-mail campaign to this voter group. The mail piece would deliver a concise turnout message and practical information about when, where, and how to vote, including early voting and mail-in voting options. The strategy may also include another form of written correspondence, such as a hand-addressed postcard or a personal letter.

• Format: An oversized postcard (quick to read) is recommended, followed by a second piece timed for early voting. A hand-written postcard variant can be tested as a sub-group if volume allows.
• Content: Election date, early-voting dates and locations, how to request a mail ballot, and a QR code linking to an official polling-place lookup.
• Track: Pieces mailed, in-home dates, and undeliverable returns.

Group 3 — Neighbor-to-Neighbor Voicemail

Use a recorded or personalized voicemail message delivered by a fellow Republican, ideally someone from the recipient’s own community. The message would encourage the recipient to vote and specifically encourage early voting or requesting a mail-in ballot.

• Messenger: Recruit local, recognizable voices—a neighbor, local volunteer, or community leader—and record region-specific versions so the speaker’s community matches the recipient’s.
• Delivery: Ringless voicemail or a voicemail drop through a compliant dialer. Keep messages under 30 seconds.
• Track: Numbers attempted and messages confirmed delivered.

Group 4 — Door-to-Door Contact

Conduct a targeted door-knocking campaign to voters in this group. The same core message would be delivered personally, with a printed leave-behind that provides information about early voting and mail-in voting.

• Walk lists: Generated only from Group 4 households, using a canvassing app to record every attempt and result.
• Conversation: Brief and friendly: confirm the election date, ask whether the voter plans to vote early, by mail, or on Election Day, and help them make a specific plan.
• Leave-behind: Left at every door, whether or not someone answered.
• Track: Doors attempted, conversations held, not-home, refused, and moved.

Group 5 — Control

No additional program contact. These voters may still hear from candidates, other groups, or the media, just as every other group will; that background activity is part of the baseline the other groups are measured against.

5. Message Framework

A shared message across all four channels ensures the experiment measures the delivery method. Research on turnout messaging points to several elements that generally help:

• Specific election, specific date. Name the gubernatorial election and November 3, 2026.
• Make a plan. Ask the voter how and when they will vote—early, by mail, or on Election Day. Plan-making prompts are among the better-supported techniques.
• Practical logistics. Early-voting dates and locations, mail-ballot request deadlines, and how to find a polling place.
• Positive identity and community. “Republicans in [community] are voting early this year” rather than scolding or shame-based messages.
• One clear call to action. Each contact should ask for one next step.

Sample scripts for each channel are provided in the Appendix.

6. Measure the Results

After the election, the state updates its voter file with 2026 vote history. Those records are matched to the locked group lists, and turnout is calculated for every group. Because the outcome comes from official records rather than surveys, it is objective and cannot be inflated by self-reporting.

6.1 Core metrics

The final measure is the most important. A method that produces a larger increase in turnout may still be the less useful choice if it costs substantially more to implement.

Hypothetical data for illustration only. Actual results will be substituted after the election.

6.2 Analysis approach

• Intent-to-treat (primary). Compare every voter assigned to a group with the control group, whether or not they were actually reached. This is the honest, program-level answer to the question, “If we run this method again, what will we get?”
• Effect per voter reached (secondary). Divide the intent-to-treat lift by the contact rate to estimate the effect on voters who were actually reached. This is useful for comparing methods with very different contact rates.
• Statistical confidence. Report each lift with a confidence interval, and say plainly when a difference is too small to distinguish from zero.
• Sub-group results. Where group sizes allow, break results down by age, geography, and past vote frequency to see whether different methods work for different voters.
• Vote method. Record whether mobilized voters voted early, by mail, or on Election Day to see whether the early-voting and mail-ballot messaging changed how people voted as well as whether they voted.

6.3 Worked example

Suppose each group contains 10,000 voters and the control group turns out at 30.0%:

All figures hypothetical.

In this hypothetical, door-to-door produces the largest lift but text produces the cheapest additional votes—exactly the trade-off the experiment is designed to reveal.

7. Budget Framework

Costs must be tracked separately by group so that the cost-per-vote figure is accurate. Include both vendor and staff/volunteer costs. Use the template below once vendor quotes are in hand:

Shared costs (voter file, analysis) should be reported separately rather than divided among the groups, because they would be incurred no matter which method is chosen in the future.

8. Timeline

With the general election on November 3, 2026, the list build and randomization need to be done quickly so that mail-ballot request reminders reach voters before request deadlines. The schedule below is shown in relative weeks and should be aligned to the state’s official early-voting and mail-ballot calendar.

9. Compliance and Data Handling

• Texting and calling rules. Use a compliant peer-to-peer texting platform and voicemail vendor, honor opt-outs immediately, and confirm that current federal (TCPA) and state rules are followed. Have counsel or the vendor confirm the approach.
• Disclaimers. Include the required “paid for by” / authority line on mail, texts, leave-behinds, and recorded messages as required by federal and state campaign-finance law.
• Accurate information only. All dates, locations, and deadlines must come from the official state or county election authority and be double-checked before release.
• Data security. Voter-file use must follow the state’s permitted-use rules. Limit access, keep files off personal devices, and delete working copies at the end of the project.

10. Risks and Mitigations

11. Value of the Experiment

The principal benefit is that a traditional get-out-the-vote operation becomes a measurable field experiment. Instead of relying on assumptions about which form of voter contact works, the campaign generates its own evidence from its own voter universe.

• Smarter spending. Future turnout programs can concentrate resources on the methods with the greatest measurable effect relative to their cost.
• Repeatable. The same design can be rerun in 2028 and later cycles to test new messages, timing, or channels.
• Shareable. Clear, data-backed findings strengthen fundraising appeals and coordination with candidates and allied organizations.

Recommended next steps

1. Approve the design and confirm the treatment arms based on universe size and volunteer capacity.
2. Obtain the voter file and run the universe query and exclusions.
3. Collect vendor quotes and finalize the budget.
4. Randomize, check balance, and lock the group lists.
5. Finalize the message and scripts; complete compliance review.
6. Launch contact, track activity weekly, and audit control suppression.
7. After the election, match vote history and deliver the findings report.

Appendix: Sample Scripts

Bracketed items are filled in from official election calendars. Scripts are starting points for review.

Text message (early-voting wave)

Direct mail (postcard front / back)

Neighbor-to-neighbor voicemail

Door-to-door