Voter Data, Targeting and Campaign Infrastructure

Behind almost every modern campaign sits a file. In the United States it starts with the public voter registration record: names, addresses and, in many places, whether each person voted in past elections. Campaigns and data firms attach more to it, from consumer data to canvassing notes, and then build models that give every voter a score. How likely are you to support the candidate? How likely are you to turn out? How likely are you to be persuaded? Those scores decide who gets a knock on the door, a text, an advert or nothing at all.

The same idea looks very different in Europe. Data protection law treats information about someone's politics as especially sensitive, and a campaign cannot simply assemble a scored file on every voter the way an American one can. In 2025 the two biggest sellers of online political advertising, Meta and Google, stopped selling political ads in the European Union rather than comply with its new advertising rules. Campaigns there have had to fall back on channels they own, such as their own email and supporter lists, and on the phones of their volunteers.

Why this matters to you

If you work on a campaign, the data layer decides where your time and money go, so you need to know how reliable it is. A support score is a model's guess, not a fact about a person, and nobody outside the campaign can check most of them. Matching a name to the right person in a file is itself a guess, and a wrong match sends the wrong message to the wrong voter. If you are a citizen, this chapter explains what campaigns can legally hold about you and how accurate it is likely to be. If you regulate, it sets out the practical consequences of the rules, including some nobody intended.

What you'll find inside

The chapter begins with the voter file itself and then the three scores that run most campaigns: support, turnout and persuasion. It compares the American and European data regimes, and follows what happened after the platforms left the European political advertising market. You will see how campaigns rebuilt around owned lists and how many voters those lists really reach. One piece tests the most famous claim in the field, that tailoring a message to each person persuades more than one good message for everybody. Others cover relational data, the contacts in your volunteers' phones, and synthetic electorates, simulated voters used to rehearse a campaign. The chapter closes with identity resolution: how often matching a person to a record goes wrong, and who pays when it does.

The honest note

The characteristic caveat of this chapter is that a score is a model, and the evidence that finer targeting yields more persuasion is thinner than the sales pitch. Much of what campaigns believe about their data has never been independently tested, and a fair share of the evidence here comes from practitioners and official records rather than from peer-reviewed studies. The articles say so where it applies. The data layer is powerful at deciding whom to contact. Whether it knows those people as well as it claims is a different question, and often the honest answer is that nobody has checked.

The nine pieces in this chapter

• P3-01 · The Voter File: The most important dataset in politics
• P3-02 · Support, Turnout, Persuasion Scores: Three models that run a campaign
• P3-03 · Two Data Regimes: What US campaigns can do and EU campaigns can't
• P3-04 · When the Platforms Left: Political ads vanish from Meta and Google in the EU
• P3-05 · The First-Party Revival: Owned channels replace rented reach
• P3-06 · The Microtargeting Claim: Precision without proof
• P3-07 · Relational Data: Your volunteers' phones as the file
• P3-08 · Synthetic Electorates: Rehearsing the campaign on silicon voters
• P3-09 · Knowing Who Is Who: Identity resolution and its limits