Lists of registered voters—augmented by data from consumer files—are the sine qua non of micro-targeting and data analytics in contemporary political campaigns. Although journalists often overstate the coverage and quality of these lists—and the use of lists is hardly a recent development—they are vitally important to parties, campaigns, and interest groups (Hersh Reference Hersh2015; Sigelman and Jewell Reference Sigelman and Jewell1986). In political science, lists of registered voters were essential to the development of “Get Out The Vote” field experiments, now a staple of mobilization and persuasion campaigns (Gerber, Green, and Larimer Reference Gerber, Green and Larimer2008; Issenberg Reference Issenberg2012). Lists have been proposed as a superior means of conducting “likely voter” screens in pre-election surveys and are widely used as sampling frames in political polling (Green and Gerber Reference Green and Gerber2006; Rogers and Aida Reference Rogers and Aida2013).
The databases maintained by partisan organizations (e.g., Catalist, Targetsmart, and the GOP Data Trust) and commercial vendors (e.g., L2) aggregate information from state voter files and commercial data vendors. These lists can reasonably purport to include every registered voter in the United States (i.e., registration information is a matter of public record), as well as millions of unregistered people found only by commercial data vendors. Indeed, because these voter-registration lists provide reliable residential information for so many Americans, the US Census Bureau considered using them to identify vacant housing units and to enumerate some households for the 2020 Census (Morris, Keller, and Clark Reference Morris, Keller and Clark2015; Robinson Reference Robinson2011).
Because these commercial voter files now play a larger part in political science and political practice, researchers have begun to use them not only as a tool for studying political behavior or as a lens through which to understand campaigns but also as a subject of study themselves (Berent, Krosnick, and Lupia Reference Berent, Krosnick and Lupia2016; Hersh Reference Hersh2015; Hersh and Nall Reference Hersh and Nall2015; McDonald Reference McDonald2007; Nyhan, Skovron, and Titiunik Reference Nyhan, Skovron and Titiunik2017). This reflects an understanding that whereas the files are a useful research tool, they also can provide a view of American voters that is subject to biases. In this sense, voter files are not unlike surveys: they are a valuable research tool for understanding American politics but come at the cost of inherent imperfections.
However, much like surveys, voter-file research has become even more central to the practice of politics than they have to political science. Campaigns use voter files to choose who should be contacted for registration, mobilization, and persuasion activities and to find the best contact information for these prospective contacts (Issenberg Reference Issenberg2012). Indeed, the use of commercial voter files—and the analytics technologies they enabled—became such a central part of Hillary Clinton’s 2016 campaign that journalists conjectured that the overuse of these data was a factor in her loss—a critique unimaginable for a presidential campaign only 12 years before (Kilgore Reference Kilgore2016). Because voter files now have an important role in campaigns, the biases inherent to their use will become manifest in political action.
These lists have made political campaigns more efficient, but at what cost? Some critics bemoan a “death by data” in political campaigning, contending that the increased importance of targeted appeals to individual voters crowds out public messaging and a more substantive political discourse (Brooks Reference Brooks2014). This article examines a different type of civic death: the political invisibility of Americans not reachable from lists.
This article documents three biases inherent to voter files, registration bias, reachability, and unlistedness and also explores the political implications of these differences in the context of the 2012 US presidential election. These biases motivate four categories of people: (1) registered voters (whom we term registered) listed at their correct address; (2) unregistered people appearing on consumer files at their correct address; (3) mislisted people, who appear on the files at an address other than the address at which they live; and (4) unlisted people, who do not appear in databases of registered voters or in consumer files. Categories 1, 2, and 3 are “listed” persons and categories 1 and 2 are reachable correctly listed persons—that is, people who can be contacted by mail or in person (i.e., the two mainstays of political outreach) using information from the files.
Two thirds of mislisted people are registered to vote but at an address where they do not reside. Much of the machinery of political mobilization entirely misses these voters. Furthermore, because their current address is not recorded in the files, voter-registration operations that target recent relocations also will miss these people.
Moving from the entire citizenry to registered voters, the electorate becomes less racially diverse, wealthier, more likely to report being contacted by a campaign, and less supportive of the Democratic Party. An electorate that encompasses only listed people has policy preferences that are more conservative than that of the entire citizenry. In this way, the reliance on lists in contemporary American politics diminishes the visibility and political power of minorities and poor people, thereby tilting policy and election outcomes in a more conservative and more Republican direction.
DATA AND METHODOLOGY
The 2012 American National Election Studies (ANES) face-to-face survey was based on a random sample of households drawn from the US Postal Service Computerized Delivery Sequence File (i.e., a list of active addresses in the United States) and was augmented by field enumeration of households in rural census tracts (ANES 2014). Institutional addresses were excluded from the sample. Information from voter and consumer files was not used in forming the sample. As such, the ANES sample covered addresses and individuals that do not appear on voter files or in the databases of commercial data vendors. The sampling frame for the ANES spanned adult US citizens (not residing in group quarters), which is a close match with the voting-eligible population.Footnote 1 A randomly chosen individual was used as the respondent in households with more than one eligible adult citizen. In addition, the ANES in-person interview occurred at the sampled respondent’s home so that, in general, the address at which the interview took place would correspond to the address associated with the respondent in the databases and lists used by campaigns.
Of 2,054 ANES respondents, 2,006 provided sufficient name information to attempt a match to three prominent commercial voter files acquired by ANES; 1,693 were located in these lists (including both registered and unregistered people). ANES collected detailed identifying data and interviewed respondents in their home, bolstering our confidence in its matches of respondents to the lists (details of the matching procedure are available from ANES 2014; 2016). Of the 333 mislisted respondents, 223 matched with a full name and birthdate to a commercial voter-file record with an address other than the address at which they were interviewed. Another 59 respondents matched on birth year or age when full birthdate information was incomplete, and 51 cases were treated as mislisted matches when the commercial voter-file record was missing it. In 40% of these missing-birthdate cases, the match was made to a record with an address in the same city as the respondent’s residence, leading us to believe that the mislisting was due to an error in the file rather than an erroneous match to another person. Finally, in 31 cases, comprising slightly less than 10% of the total mislisted cases, a match was made to a record missing birthdate and age information with an address outside of the respondent’s city or town. In these cases, we used the vendor’s match—although it is possible that some of the mislisted individuals should have been classified as unlisted. Although some surely occurred, any misclassifications would attenuate differences among citizenship categories, leading us to observe smaller differences than actually exist in the American public. The reported results are highly unlikely to be an artifact of errors in the matching process.
After applying weights to make the ANES data representative of the sampling frame, correctly listed persons comprised 78% of the sample, with those mislisted accounting for an additional 12%. Therefore, 23% of the weighted sample was unlisted or mislisted in the databases available to political parties and campaigns ahead of the 2012 General Election. We considered this estimate of the unreachable population to be a lower bound on the true proportion because people who have not registered to vote at their current address also are less likely to be interested in participating in a lengthy political survey (Jackman and Spahn Reference Jackman and Spahn2019).
We were interested in testing hypotheses that illustrate how these groups differ in their socioeconomic status (SES) and political attitudes, focusing on how individuals become politically invisible and the possible consequences of their nonparticipation or noncontact. Our aim was to show that the absence of politically invisible people from voter lists—and thus from voter contact and research strategies that rely on them—is both substantively and normatively consequential. To support our claims, we identified the following interesting and statistically significant differences among these groups (Jackman and Spahn Reference Jackman and Spahn2021).
RACE AND ETHNICITY
People of color are considerably less likely to be correctly registered than whites. Table 1 shows that only 18% of whites were unreachable, compared to approximately 40% for Blacks and Hispanics. Among Blacks and Hispanics, about 20% were unlisted (i.e., not appearing on either voter or consumer files) and another 20% were mislisted (i.e., listed at an address other than their current residence). Near majorities of the minority population could not be targeted for contact by direct mailers or in-person canvasses.
Distribution of Citizen Types (Percentage), by Race and Ethnicity, ANES 2012 Face-to-Face Respondents (Weighted)
Note: p < 0.01, N = 2,006.
Only a small proportion of minorities not registered at their current address could be found using consumer information. Of the 42% of Blacks not registered to vote at their current address (i.e., those in the unregistered, mislisted, and unlisted categories), only 6% had a record with a correct address. For Hispanics, the comparable percentage was 13%. For whites, 29% of those without a current voter registration had a consumer record with an accurate address. Consumer files have low penetration into the unregistered minority population but encompass almost a third of unregistered whites. Thus, consumer files are a viable tool for the political incorporation of whites who have not registered to vote, but they are much less useful for other racial and ethnic groups.
AGE
The median age of an unlisted person was 29 years and the average age of accurately registered voters was 51. This age contrast follows differences in mobility and homeownership throughout the life cycle. Younger people are more prone to rent and to relocate, more likely to be unlisted or mislisted, and therefore politically invisible to list-based political campaigns.
INCOME, RESIDENTIAL TENURE, AND HOMEOWNERSHIP
Differences across citizen types with respect to three SES indicators are shown in figure 1. Specifically, unlisted people are more financially vulnerable than their listed peers. The median annual income for unlisted people was only $24,000, increasing to $41,000 for unregistered people and to $57,000 for accurately registered respondents. Unlisted and mislisted people are about half as likely to report owning their home as the correctly listed citizenry (i.e., 40% versus 75%), and they are more than twice as likely to report living at their current address for less than one year. High rates of residential mobility and low rates of homeownership explain much of the absence of correct address data for unlisted and mislisted people. Frequent relocation means that the contact information available on commercial voter files reflects previous addresses, leaving people mislisted or absent from the files. For unlisted and mislisted people, a low income means they generate less consumer information, allowing a further lapse in contact information to occur. Although the US Postal Service National Change of Address database facilitates the revision of outdated contact information, not everyone registers their move with the postal service. Indeed, unlisted people may have little or no commercially relevant information or even a voter-registration record.
Age, Income and Residential Stability
Notes: Percentage of citizen types with listed attribute (error bars for proportions are ± two standard errors), ANES 2012 face-to-face respondents (weighted). Voters have consistently higher SES than nonvoters. Of the four groups, unlisted people have the lowest income and are least likely to own their home. “Recently moved” refers to people who have lived at their current address for less than a year. All of the between-group differences are extremely unlikely to be due to chance alone; each has a p.
RESIDENTIAL MOBILITY
Voter registration is tied to an individual living at a certain address rather than to the individual. Residential mobility severs a citizen’s connection to the voter-registration system. An affirmative act by citizens is required to reregister each time they relocate (Schmidhauser Reference Schmidhauser1963; Squire, Wolfinger, and Glass Reference Squire, Wolfinger and Glass1987). The National Voter Registration Act of 1993 (i.e., “Motor Voter”) lowered the cost of reregistration, but our data confirm a pattern long noted in the literature: moving results in many citizens falling off of voter rolls as well as commercial databases (Highton and Wolfinger Reference Highton and Wolfinger1998).
Residential mobility is concentrated among poor and urban-dwelling people. In turn, this can explain why whites are less likely to be listed than minorities. Figure 2 demonstrates that duration of residential tenure is associated with political visibility. A majority of Blacks (59%) and Hispanics (52%) reported residing at their current address for fewer than five years; among whites, the corresponding percentage was 36%. For whites, 48% reported residing at their current address for 10 years or more but this was the case for only one in three Blacks and Hispanics.
Income and Residential Tenure
Note: Distribution of citizen types by income tercile (annual household income) and residential tenure (years at current residence); p < 0.01.
Income powerfully shapes the relationship between residential mobility and being listed. Figure 2 shows that people with a higher income are more likely to be politically visible, consistent with a resource model of political participation, in which civic knowledge, time, and money are three key resources (Brady, Verba, and Schlozman Reference Brady, Verba and Schlozman1995). Our analysis points to a fourth element, closely associated with wealth: being listed. Because unregistered records are sourced from commercial voter files, the probability of being listed increases with wealth. Figure 2 demonstrates that the poorest respondents are significantly more likely to be unlisted and significantly less likely to be registered. Residential mobility elevates the risk of becoming (or remaining) unreachable, even for wealthy people.
CAMPAIGN CONTACT
Being reachable—and especially being registered—greatly facilitates contact with campaigns. As figure 3 illustrates, registered voters with accurate contact information were significantly more likely to report being contacted by a campaign in 2012. Unregistereds and politically invisible people reported less than half the rate of contact as registered respondents. Only 33% of registered mislisted people who voted in 2008 reported being contacted in 2012—less than the 46% contact rate for correctly listed registered respondents overall and the 52% contact rate for correctly listed registered respondents who voted in 2008. The 20-point gap between mislisted respondents and registered voters confirms the importance of accurate address information for contact by campaigns. When address information is inaccurate, a campaign’s outreach often fails to reach its intended target.
Campaigns and Issues
Notes: Percentage of each citizen type reporting contact by campaigns and parties, expressing support for more federal welfare spending, and expressing support for President Obama in an interview after the 2012 presidential election. ANES 2012 face-to-face respondents (weighted);. p < 0.01.
The situation is even more stark for unlisted people. Both unregistered and unlisted respondents were not registered to vote; however, almost 20% of unregistered respondents and only 10% of unlisted respondents reported contact from a campaign or political organization in 2012. Unregistered-but-listed respondents reported twice as much attention from campaigns as those unlisted. Because direct contact between candidates and citizens is an important avenue through which politicians learn about citizen preferences and make policy decisions, the extra attentiveness of politicians to listed versus unlisted people reinforces economic and social inequality (Fenno Reference Fenno1978). Although unlisted and mislisted people might be stimulated by broadcast advertisements, the ads do not funnel information back to the campaign as would an in-person canvassing.
By expending fewer political resources on unlisted people, politicians and politically engaged volunteers rarely come into contact with this group, potentially leading to policy issues of particular importance to them receiving less attention. The 11% of the population that is unlisted is mostly people of color with a low median income, and a high rate of residential mobility—traits that distinguish them from the more white, wealthier, and residentially stable registered population. The median income among those who reported contact from a campaign was $62,500, whereas among those not reporting contact, it was only $37,500—a gap partly explained by the differences in listed status between low and high earners.
The effect of being listed overshadows the effect of income, explaining more than three times as much variance in contact rates than income alone. After poorer people become listed (and especially when they start voting), they can be contacted for continued mobilization efforts (Nickerson Reference Nickerson2015). This suggests that voter-registration efforts may have important secondary effects, exposing those newly listed to contact opportunities from various political organizations using lists to identify targets for mobilization efforts.
POLITICAL ATTITUDES
If registration and mobilization activities truly affect the composition of the electorate, how would the political views of the electorate change if everyone was listed and registered and turned out to vote? In general, the policy positions of the unregistered, mislisted, and unregistered respondents in our analysis were more liberal than those of registered voters. For example, the middle panel in figure 3 displays variation in preferences about federal welfare spending. Only 13% of those registered thought that spending should increase, compared to 22% to 25% of those unregistered and unreachable. The differences across citizen types on policy matters were so stark that on a related issue—federal spending for childcare—the median position of the electorate moved from supporting the status quo level to supporting an increase when we shifted focus from voters to the entire citizenry.
OBAMA VOTE CHOICE AND PARTISANSHIP
Significant disparities in SES across the four citizen groups led to considerable variation in partisanship and vote choice. The bottom panel of figure 3 displays Obama’s share of the two-party vote-choice item across the four citizen types. The weighted ANES data closely reproduced the national 2012 two-party result, with respondents known to have turned out in 2012 favoring Obama versus Romney 52 to 48. Correctly listed registered voters supported Obama versus Romney 53 to 47, while those unregistered were split 50–50 between the two candidates.
Those who were unreachable, conversely, broke strongly for Obama: 62% of mislisted and 74% of unlisted people indicated support for Obama. These margins mirror the exceptionally strong support that Democratic candidates earn from non-whites.
Table 2 shows that the unreachable population is more likely to identify with the Democratic Party than their listed peers. This affiliation aligns with their higher rate of support for Obama and for the welfare state but begs a question: Why do they remain unreachable when Democrats would benefit greatly from getting them registered and voting? The answer likely lies with the institution of voter registration. The American voter-registration system was established as a tool to limit voting to landed men; ironically, it still serves those ends today (Cunningham Reference Cunningham1991). The institution of voter registration simply is not designed to accommodate itinerant citizens. Even if unreachable people were to be registered—and the Democratic Party appears to have a clear incentive to do so—their small commercial footprint and frequent moves would result in them again becoming politically invisible.
Percentage Identifying with Each of the Two Parties or Identifying as Independent, by Category
Note: p < 0.01.
POLITICAL REPRESENTATION IN THE LIST-BASED ERA
Mislisted and unlisted people are marginalized in other domains of American life. They are poorer, more financially vulnerable, younger, and more likely to be non-white than registered voters. They also report more liberal policy preferences and political attitudes, and they express less satisfaction with America’s political system.
Political parties and interest groups often are seen as brokers in American politics, connecting citizens and candidates, voters with vote seekers (Aldrich Reference Aldrich1995). That at least 11% of the citizenry is unlisted indicates a market failure of sorts. Mislisted and unlisted people report much less contact with political parties and electoral campaigns. Political parties and interest groups—the dominant agents of political mobilization—are either unaware of the substantial unlisted and mislisted segments of the citizenry or have calculated that mobilizing them is simply not worth the effort.
Some scholars might not perceive a great normative issue in large proportions of the citizenry being mislisted or unlisted. In a free society, citizens are not compelled to register to vote or to turn out—being unlisted or unregistered is a choice.
Disenfranchising any particular person is unlikely to be electorally pivotal and probably has infinitesimal direct effects on that person’s welfare. However, the burdens of the voter-registration system—as low as they might be—are experienced disproportionately by poor people. Low SES is associated with significantly lower levels of contact with candidates, parties, and interest groups. Low levels of contact between parties and candidates and poor people, coupled with their disproportionately low turnout, may explain why public policy tends to favor wealthy people (Bartels Reference Bartels2009; Gilens Reference Gilens2012).
The inequality in political visibility documented in this article has parallels with inequality in other domains of American life. Economic and social disadvantage predict diminished political visibility, participation, and representation. Economic disadvantage and its concomitant high rates of residential mobility result in minorities, young people, and those who are financially vulnerable—people more liberal than most voting Americans—being less likely to cast a ballot and far less visible to list-driven campaigns.
Moreover, without a positive, affirmative act by citizens to become registered to vote, the way they move from being unlisted to listed is via their behavior—not in the realm of politics but rather in the economy through credit card usage, active bank accounts, and homeownership. In no small measure, political visibility is premised on a person’s visibility as a consumer. Inequality in economic consumption is reflected in the unequal political visibility discussed in this article. As a formal, legal matter, political representation is a constitutional guarantee available to all citizens; as a practical matter, economic and social inequality generates inequality in political visibility.
Are lists good for American democracy? Although lists surely make campaigns more efficient, they do so at the expense of unlisted and mislisted Americans. The listed electorate is whiter, older, wealthier, and more conservative than the general citizenry. Those well off and already powerful are the beneficiaries of this new political institution.
DATA AVAILABILITY STATEMENT
Replication materials are available on Harvard Dataverse at https://doi.org/10.7910/DVN/PUYM1K.
