Hacking the Electorate

1. Campaigns operate on "perceived voters," not encyclopedic knowledge of the electorate.

Campaigns do not perceive voters as voters perceive themselves.

Information fallacy. Contrary to popular belief and even some academic assumptions, modern political campaigns do not possess accurate, detailed, or encyclopedic information about every voter's preferences and behaviors. The idea that microtargeters "know you better than you know yourself" is a pervasive "information fallacy." Campaigns face significant informational hurdles in understanding the electorate.

Limited knowledge. Campaigns operate with a simplified and often distorted version of the electorate, which the author terms "perceived voters." These are avatars generated from available data, not intimate knowledge. Several factors prevent campaigns from achieving encyclopedic knowledge:

• Scale: Large electorates (hundreds of thousands to millions of voters) make individual-level intimate knowledge impossible.
• Flux: Voters constantly move, register, unregister, and die, making long-term tracking difficult.
• Privacy: Vote choice is secret, and voters' opinions are fluid and often candidate-specific.
• Bias: Campaign workers' perceptions of voter support are often overly optimistic.

Strategic decisions. All campaign decisions, from who to contact to what message to deliver, are based on these "perceived voters." Understanding campaign strategy requires looking at the data inputs and the inherent biases and limitations of those tools, rather than assuming campaigns know what voters know about themselves. This shift in perspective reveals why campaigns make the choices they do, given their informational constraints.

2. Public records are the most crucial and reliable "shortcuts" for campaigns to understand voters.

The data that inform their perceptions will affect the decisions they make.

Essential shortcuts. Since campaigns cannot have encyclopedic knowledge, they rely on "information shortcuts" or heuristics to form impressions about voters' political dispositions. Public records, primarily from the voter registration system, but also from the Census Bureau and state licensing agencies, serve as the most critical and useful of these shortcuts.

Key characteristics. Public records possess several qualities that make them invaluable to campaigns:

• Predictive power: Information like party registration or primary voting history is highly correlated with partisan support and turnout.
• Wide coverage: Voter registration lists include nearly every eligible voter in a jurisdiction.
• Low cost: Election authorities often provide these lists cheaply or for free to political entities.
• Accuracy: Data points like age, gender, and party affiliation (where collected) are highly accurate.

Foundation of perception. These records allow campaigns to segment the electorate into broad categories of likely supporters, opponents, and potential voters. For example, a voter's age is a public record in nearly every state, enabling campaigns to target young voters who strongly supported Barack Obama, even those not in college towns. Without such individual-identifying data, campaigns would be forced to use coarser, less effective strategies.

3. State data policies, not just technology, fundamentally shape campaign strategies.

The laws about the collection and dissemination of public data are often tailored so that politicians can repurpose the data in service of their campaigns.

Policy as a lever. The specific laws governing the collection and distribution of public records vary significantly across U.S. states. These variations act as "levers" that directly influence how campaigns perceive voters and, consequently, the strategic choices they make. This highlights a new link between American political institutions: state legislatures and election administrations directly impact political parties' electoral strategies.

Purposeful design. Public records are not merely incidentally useful for campaigning; they are often useful by design. Politicians, aware of the electoral value of personal data, have historically crafted laws to facilitate its collection and dissemination for campaign purposes. Examples include:

• Ohio Senate Bill 8 (2009): Proposed allowing voters to register with a party, explicitly to help campaigns "target our new voters that are more likely to lean your way or my way."
• California bill (2003): Sought to add a race option to voter registration, with the stated purpose of increasing "voter outreach efforts to these historically underrepresented populations."
• Rejection of privacy bills: Proposals in Utah (2012) and New Hampshire (2008) to allow voters to opt-out of sharing personal data with parties were rejected, with committees citing the "strong public interest in having candidates being able to directly contact potential voters."

Beyond registration. This repurposing extends beyond voter registration. The U.S. Census, initially for apportionment, now provides granular neighborhood data used by campaigns. Open-record laws, intended for government transparency, are leveraged by parties to obtain lists of licensed professionals (e.g., pilots, teachers, farmers) for targeted messaging. This demonstrates a systemic integration of administrative data into political strategy.

4. Party registration data dramatically alters how campaigns perceive and mobilize partisans.

When public records are available that provide a clear signal of partisanship, I hypothesize that campaigns focus more on mobilizing partisans and less on targeting geographies or persuading undecided voters.

Clear signals. States vary in whether they collect party registration and/or primary voting history. In states where these public records are available, campaigns gain a highly accurate signal of a voter's likely partisan support. For example, in 2012, 88% of registered Democrats voted for Obama, and 90% of registered Republicans voted for Romney. This allows campaigns to focus efficiently on mobilizing their base.

Divergent perceptions. In states without public partisanship data, campaigns must rely on noisier proxies like:

• Geographic data: Precinct-level election returns, which are often unhelpful as most precincts are mixed-partisan.
• Microtargeting models: Predictive models (like Catalist's) that use demographics, consumer habits, and proprietary data. While useful, these models produce a less polarized distribution of "perceived partisans" compared to states with direct party registration.

Strategic shifts. This difference in perception leads to distinct strategies:

• Party registration states: Campaigns prioritize individual-level mobilization of known partisans, leading to higher turnout among self-identified partisans.
• Non-party registration states: Campaigns are more reliant on geographic targeting and microtargeting models, leading to more focus on persuasion and more accidental contact with non-supporters. Campaign workers in these states report lower morale due to frequent interactions with opposing voters.

Counterfactual evidence. A simulation applying a "no party data" model to a party-registration state like Pennsylvania shows that campaigns would perceive a significantly different, less accurate "Democratic coalition," engaging many more independents and out-party voters. This confirms that data availability, not just voter disposition, drives these strategic differences.

5. Public records of racial identity enable targeted outreach and influence turnout patterns.

When public records provide a clear signal of racial identity, I will show that campaigns focus more on mobilizing voters because of the voters’ race, and less on targeting geographic areas with homogenous racial groups.

Racial identifiers. Eight Southern states collect racial identity data on voter registration forms, making it a public record. This information is highly predictive of partisan support, especially for Democrats targeting African American voters. Campaigns in these states can "surgically" target minority voters with specific messages.

Perception challenges. In states without public race data, campaigns resort to less precise methods to perceive race:

• Geography: Targeting predominantly black neighborhoods, which misses a large portion of the black electorate living in mixed areas.
• Commercial predictions: Models using names and neighborhood demographics, which are significantly less accurate (e.g., 68% accuracy for black voters vs. 95% with public records).
• Networks: Church-based mobilization, which reaches only a fraction of black voters.

Strategic and turnout effects. The availability of race data leads to measurable differences:

• Targeting: Campaign workers in states with racial registration (e.g., North Carolina, Florida) report race as a much more important characteristic for direct contact than in similar states without such data (e.g., Virginia).
• Turnout patterns: In North Carolina (with race data), black turnout was higher and less geographically contingent than in Virginia (without race data). This suggests individual targeting in NC versus neighborhood targeting in VA.
• "Unlisted" voters: Voters in racial registration states who do not list their race on public records consistently vote at lower rates. This is likely because campaigns, unable to perceive their race, tend to ignore them, despite similar demographics or personality traits.

6. Campaigns struggle to accurately identify and persuade truly undecided voters.

In campaign databases, there is no identifiable group of voters who are predictably and consistently persuadable.

Asymmetry in data. Unlike partisanship, for which public records can provide clear signals, there is virtually no data—public, commercial, or proprietary—that accurately predicts a voter's "persuadability." Persuadability is a complex psychological disposition (e.g., cross-pressured, uninformed, pure independent) that is difficult to infer from demographic, geographic, or behavioral data.

Noisy approximations. Campaigns typically define "persuadable perceived voters" as those who are likely to vote but whose partisan leanings are unknown from public records. Common targets include:

• Regular-voter targets: Voted in 70%+ of eligible elections, but not registered with a party or in partisan primaries, and live in mixed-partisan precincts.
• Surge targets: Voted in the last presidential election but not the last midterm, with no clear partisan signal.
• New-voter targets: Recently registered, with no clear partisan signal.

Disconnect with reality. A comparison with survey-based definitions of persuadable voters (e.g., pure independents, uninformed, cross-pressured) reveals a significant disconnect:

• Ideology: Campaign targets are considerably more ideologically committed than survey-defined persuadables, who tend to be moderate.
• Opinion shifts: Survey-defined persuadables (especially pure independents) are significantly more likely to change their vote preference or be undecided pre-election than campaign targets.
• Targeting inefficiency: Campaigns' persuasion efforts are often inefficient because their "persuadable perceived voters" are a very noisy approximation of actual persuadable voters.

Implications for democracy. This inability to pinpoint persuadable voters means that those most susceptible to elite influence (e.g., uninformed, cross-pressured) are often not reached by targeted persuasion messages. This tempers concerns about elites manipulating fickle voters, as campaigns simply cannot find them at the individual level.

7. Commercial data and social networks are imperfect substitutes for public records in targeting.

The commercial data, proprietary data, and social network data that campaigns have utilized in the last few election cycles do not allow campaigns to perceive voters in the same way that public records allow.

Limits of commercial data. While campaigns can access hundreds of commercial variables (e.g., cat ownership, interest in golf), these are rarely used directly for targeting. Their utility is limited because:

• Inaccuracy: Commercial data often lags or relies on rough predictive models (e.g., only 25% accuracy for Jewish voters, 38% for Catholic).
• Irrelevance: Most consumer habits are not strongly correlated with political preferences or turnout once basic demographics are controlled for.
• Cost: Acquiring and linking commercial data is expensive.

Proprietary data's role. Proprietary party records, such as field IDs from past campaign contacts, also have limited predictive value. Campaign workers are often overly optimistic about voter support, leading to inaccurate data. Even sophisticated models like Catalist's, which incorporate commercial and proprietary data, still rely heavily on public records (like party registration) to accurately predict partisanship. Without public party data, these models show voters as much more uncertain in their partisan leanings.

Social network challenges. Network-based targeting (offline or online via Facebook) faces significant hurdles:

• Volunteer preference: Volunteers often prefer contacting strangers over personal acquaintances about politics.
• Limited reach: Networks of activists are often too small and politically homogenous to reach a broad, diverse, or persuadable electorate.
• Data gaps: Campaign databases typically lack detailed information about voters' social connections beyond households.

Public records remain supreme. These alternative data sources cannot fully compensate for the absence of key public records. The strong effect of public records on campaign perceptions and strategies persists because other data sources are either less accurate, less predictive, or harder to leverage for large-scale, individual-level targeting.

8. The political repurposing of public data creates inherent conflicts of interest.

There is a conflict of interest inherent when governments collect administrative data and repurpose those data for political ends.

Dual incentives. Politicians, as both administrators and campaigners, face a conflict of interest when overseeing public records. They can shape data collection and dissemination policies to benefit their campaigns, often without public scrutiny. This repurposing of administrative data for political ends is problematic.

Inappropriate uses:

• Legislative proposals: Lawmakers propose collecting data (e.g., party affiliation, race) primarily for campaign targeting, not administrative necessity.
• Privacy rejections: Bills to allow citizens to opt-out of sharing personal data with campaigns are rejected to protect political access.
• Special access: Elected officials may leverage their positions to gain special access to government data (e.g., public housing resident lists) for campaign advantage, raising concerns about legality and coercion.

Campaign data in government. Conversely, campaign-derived data is increasingly used in official government functions, particularly in congressional offices for constituent services. Vendors merge constituent contact information with voter registration and commercial data, allowing offices to microtarget responses. This raises ethical flags:

• Bias risk: Staffers can see a constituent's party, race, or voting history, potentially leading to differential treatment based on political considerations, violating the ideal of equal representation.
• Lack of oversight: Unlike franked mail, which is reviewed by a commission, the targeting of constituent communications is not policed, making abuse difficult to detect.

Need for transparency and safeguards. To mitigate these conflicts, greater transparency in data policy debates and stricter protocols for data use are needed. This includes restricting the types of politically relevant data congressional offices can use in constituent databases and auditing how such data informs their interactions.

9. Microtargeting's democratic impact is a policy debate, not just a technological one.

A normative discussion about microtargeting need not take the form of vague laments or hopes about the inevitable uses of personal data in politics; rather it can take the form of a clear set of expectations about how specific policies lead to one or another form of campaigning.

Reframing the debate. The debate over whether microtargeting is good or bad for democracy is often framed as an inevitable consequence of technology. However, the book demonstrates that microtargeting is largely a product of public policy—specifically, laws governing public records. This reframing allows for a more concrete discussion about policy choices and their democratic ramifications.

Arguments for microtargeting:

• Better representation: Data helps politicians understand diverse constituent interests, enabling them to build coalitions around specific issues (the "long tail" of the electorate) and provide more tailored representation.
• Voter engagement: Campaigns can identify and mobilize inactive but supportive constituencies, increasing participation among those who care about niche issues.

Arguments against microtargeting:

• Distorted perceptions: Data can lead politicians to focus only on likely supporters (e.g., Romney's "47 percent" comment), narrowing their view of whom they represent and encouraging factional appeals.
• Limited exposure: Precise targeting, enabled by public records like party registration, reduces unplanned encounters between voters and opposing viewpoints. This can hinder "by-product learning" and make it harder for voters to make reasoned decisions.

Policy solutions. To resolve this debate, a research agenda should empirically test the effects of specific data policies on democratic outcomes. The author proposes a "free credit report" for campaign database entries, allowing voters to:

• Access their profiles: See what campaigns know about them.
• Provide feedback: Potentially correct information or indicate contact preferences.
  This policy would increase transparency, empower voters, and could even provide campaigns with more accurate, self-reported data, shifting the balance of power in data collection.

Last updated: February 21, 2026