Technology & Democracy

Series 3 of 5

The Next Political Arms Race Will Be Data, Not Television Ads

Cambridge Analytica was not an aberration. It was a preview. The data infrastructure being built for the 2026 cycle makes 2016 look like a warm-up.

March 7, 20266 min read

In the weeks after the 2016 election, the name Cambridge Analytica was everywhere. The firm's claims — that it had built psychographic profiles of 87 million Americans and used them to deliver precisely targeted political messaging — generated a wave of coverage, congressional testimony, and public concern about data privacy in political campaigns.

Then Cambridge Analytica collapsed. Its parent company dissolved. The coverage moved on. What didn't move on was the underlying practice. The voter data industry — the network of firms that build, maintain, and sell detailed individual-level data on American voters — continued to expand throughout the Trump and Biden administrations and into the 2024 cycle. The methods became more sophisticated. The data sources multiplied. The computational power deployed against the problem grew exponentially.

"Cambridge Analytica was not the story. It was a distraction from the story. The story is what the industry became after everyone stopped watching."

The Voter Data Ecosystem

To understand the current state of political data, you need to understand what voter files actually are and what gets layered on top of them.

The voter file is the starting point: a government record, publicly available in most states, that contains the name, address, party registration, and voting history of every registered voter. This is the spine of the system.

On top of the voter file, campaigns and data vendors layer consumer data: purchase history from loyalty programs, magazine subscriptions, automobile ownership records, credit card transaction patterns, app usage data, location data from mobile devices, and social media behavior. They also layer modeled data: variables inferred from the observed data through statistical modeling. Did this person probably attend church? Did they probably take a vacation last year? Do they probably own a gun? Are they likely financially stressed?

The resulting profiles can contain hundreds of variables per individual. The major political data vendors — NGP VAN on the Democratic side, i360 on the Republican side, and a growing ecosystem of independent vendors serving both — maintain and continuously update these profiles for virtually every registered voter in the country and many unregistered ones.

This data is not just used to target advertising. It's used to model persuadability — to identify which voters are genuinely movable and which are locked in, so campaigns can concentrate their limited resources on the people most likely to change. It's used to model turnout probability — which supporters are likely to stay home in a low-enthusiasm environment and therefore need proactive contact. It's used to model donation likelihood — which voters, if targeted with the right message, are most likely to give.

From Targeting to Prediction

The methodological frontier, as of the 2024 cycle, has shifted from targeting to prediction.

Targeting asks: given what I know about this person, which message should I send them? Prediction asks: given what I know about this person, what will they do before they've decided?

"A targeting system optimizes the communication you send. A prediction system identifies the intervention you need to make — and when — to change a behavior before it crystallizes."

Campaigns using prediction-focused modeling were, in 2024, identifying persuadable voters weeks or months before those voters would have considered themselves persuadable. They were reaching those voters in the medium, with the message, at the time most likely to produce the desired behavioral outcome — not based on what the voter said they wanted, but based on statistical patterns in their observable behavior.

The analogy to commercial advertising is instructive. Netflix doesn't just recommend shows you've told it you like. It predicts shows you'll like before you've articulated the preference. Political data systems are converging on the same capability: predicting political behavior before the voter has consciously formed an intention.

The Infrastructure Gap and What It Means

One of the most consequential and least-discussed aspects of the data arms race is the significant gap between what well-funded campaigns can do and what everyone else can.

A major party Senate campaign in a competitive state, with access to the national party's data infrastructure and budget for sophisticated vendors, can run what amounts to a precision behavioral influence operation. A down-ballot campaign — a state house race, a school board contest, a judicial election — is operating with a fraction of that capability, often relying on vendor tools that are years behind the frontier.

This creates an asymmetry that cuts in a specific direction: the candidates most able to leverage sophisticated data infrastructure are generally the candidates with the largest fundraising operations. Which means the data advantage compounds the money advantage rather than counteracting it.

There is also a partisan asymmetry, though it shifts direction over time. Democrats built an earlier lead in data sophistication, partly as a response to being outspent on television advertising and partly due to the Obama campaign's innovations in 2008 and 2012. Republicans responded aggressively in subsequent cycles and have invested heavily in their own infrastructure. By 2024, both sides were operating sophisticated data operations, though with different vendor ecosystems and different philosophical approaches.

The Regulatory Vacuum

The legal framework governing political data is, to put it plainly, inadequate.

Federal law imposes virtually no restrictions on how campaigns collect, purchase, or use voter data. The regulations that exist focus on disclosure of spending — which candidates and committees must report expenditures — but do not address the underlying data infrastructure.

State privacy laws, most notably California's CCPA, theoretically apply to some political data collection. In practice, enforcement against political actors has been limited, and the exemptions for publicly available data (like voter files) create significant carve-outs.

The FTC's general consumer protection authority could theoretically reach some practices in the voter data industry, but the agency has not pursued political data as a priority enforcement area.

The result is a significant regulatory vacuum in one of the most consequential areas of democratic practice. Commercial firms face substantial regulation when they collect and use personal data for advertising purposes. Political actors — exercising rights that courts have broadly construed as First Amendment-protected — operate in a largely unregulated space.

"The 2026 cycle will be built on data infrastructure that would have been unimaginable in 2016. The cycle after that will be built on infrastructure unimaginable today."

This is not an argument for any particular regulatory intervention. First Amendment concerns in this space are genuine and not easily dismissed. But it is worth being clear that the current status quo is not a neutral baseline — it is an actively permissive environment for practices that have significant democratic implications, built not through any deliberate policy choice but through regulatory inaction in a fast-moving space.

This is the second installment of "Technology & Democracy," a series examining where the builder's perspective meets democratic systems. Next: "Why Campaign Websites Are Still Terrible."

Spread the word
Share

Stay Informed

Get The ElectionSpend Brief

Political ad spending data and analysis, delivered to your inbox.