The role of geolocation data in U.S. political campaigning: How digital political strategists perceive it

Geolocation data and the perceived voter

Political science literature suggests that campaign messages have a measurable effect on election outcomes (Kenski et al., 2010). In determining which subsets of voters are more susceptible to influence from campaign messages, researchers propose a variety of definitions, such as ‘swing’, ‘floating’, or ‘persuadable’ for those who change their voting decisions between elections. Statistically, early in the election season such voters amount to 20%–25% of the electorate, but as campaigns shift into full gear, that proportion falls below 10% (Kenski et al., 2010). Today’s data-driven electoral campaigns use databases to reliably identify which voters are supporters and which may be persuadable, developing predictive models and targeted advertising to increase the likelihood that people will vote for a candidate or cause. The use of data to make decisions about how to interface with voters contributes to what Hersh (2015: 24) calls the ‘perceived voter model’, which is a product of formed impressions about voters based on input data. Hersh (2015: 8) critically points out that campaigns never really deal with actual voters, but with voter imaginaries, consisting of ‘zeros and ones in computers’.

In this study, we narrowed our research inquiry to focus on the role of one type of data: geolocation information. Geolocation data represent dynamic pieces of information carrying attributes that point to specific individuals and their activities in relation to time and place (Arminen, 2006). GPS-enabled devices have the capacity to accurately track people’s location within a 16-feet (4.9 m) radius (National Coordination Office for Space-Based Positioning Navigation and Timing, 2020) with a real-time frequency of as many as 14,000 times a day (Valentino-DeVries et al., 2018). Geofencing, in which a virtual boundary is drawn around certain physical points of interest, is an additional layer of a location-based apparatus that serves as a virtual tripwire, signaling an operator whether a target is entering or exiting a ‘geofenced’ polygon. While geography has always been integral to political campaigns, there are larger ethical and technological issues that heighten concerns about geolocation data. Historically, geography has determined which media markets receive television advertising (Schaffner, 2006) and whether jurisdictional policies alter campaign advertising (Hersh, 2015). As Kruschinski and Haller (2017: 3) put it, ‘geographical targeting is the oldest – and most coarse – approach. It is based on the analysis of precinct-level results from past elections to identify auspicious electoral constituencies’. In the current political landscape, geolocation data play an important role as they can be used for purposes beyond targeting advertising to different congressional districts. Not only do platforms and mobile applications engage in geolocation tracking, but individuals’ photos, videos, and social media posts are accompanied by geo-tagged metadata (Barreneche, 2012). These technological advances open the door for new uses of geolocation data by campaigns, allowing for even more granular targeting of voters.

Data-driven political campaigning

Modern political campaigns run on data. Campaigns procure databases of information layered from voter files, social media, census data, and consumer data vendors (Funakoshi et al., 2020; Hersh, 2015; Nickerson and Rogers, 2014; Nielsen, 2012). Geolocation data are just one layer among many within data-driven political campaigns. These data inform subsequent political communication and messaging (Dommett, 2019; Hersh, 2015; Kreiss, 2016), from A/B testing email headlines to gauging public opinion (McGregor, 2020). Data-driven campaigning commonly refers to the use of data sets to alleviate the targeting of political messages (Baldwin-Philippi, 2019). It is ‘not a homogenous construct but something conducted by different actors, using different data, adopting different strategies’ (Dommett, 2019: 12). Indeed, Römmele and Gibson (2020: 603) distinguish between at least two different philosophies and approaches to data-driven campaigning: A scientific approach rationalizes voters as well as ‘narrows and reinforces voter preferences’, whereas a subversive approach harkens to emotions such as fear and does not stray away from distributing false information to ultimately suppress voter participation. Gil de Zúñiga et al. (2020: 589) write that ‘modern political campaigning allows the proliferation of extensive databases for psychometrical profiling of the voters and micro-targeting, but also leaves an open space for cyber-attacks from the opposition, and beyond’. Researchers opposing the use of micro-targeting in political campaigning argue that at its core, it is antidemocratic as it violates people’s freedom of information rights and impedes voters’ ability to exercise informed decision-making (Bayer, 2020).

Public concern about data collected by campaigns generally focuses on geolocation data, or ‘tracking’. In a 2020 nationally representative survey of U.S. adults (N = 1010), we asked the following question: ‘As you may know, some companies gather your precise location through your digital devices, in real-time, and sell it to other companies, interest groups, and political groups. How concerned are you about the collection and sale of your location data?’ ¹ 45.8% of participants reported being extremely or very concerned, while 31.7% were moderately concerned. Only 15.8% were slightly concerned, and 6.4% were not at all concerned. The assemblage of vast stores of information has prompted strong public concern in the aftermath of the Cambridge Analytica scandal (Venturini and Rogers, 2019). Data leaks of nearly 200 million registered voters in the United States (Larson, 2017) have sparked significant media attention (Funakoshi et al., 2020). The media coverage triggered a wide range of disputes and assumptions about the overall efficacy of data-driven campaigning, pointing toward what Dommett (2019: 4) has described as a ‘gulf between possible and actual practice’. Hersh (2015) similarly has pointed out that it is false to assume that information in political campaigning is always accurate. Political campaigns tend to accumulate as much data as possible, assuming that ‘more data necessarily means more and better knowledge’ (Baldwin-Philippi, 2020: 469). In this context, understanding how the protagonists in data-driven political campaigns perceive the efficacy and accuracy of geolocation-focused efforts yields important insights into where political micro-targeting succeeds, and where it still fails.

Digital political strategists

The term data broker refers to organizations that engage in ‘the practice of buying, aggregating, selling, licensing, and otherwise sharing individuals’ data’ (Sherman, 2021: 2). Data brokers often create their own voter databases from local voter files. They clean them and merge them with additional data points, anything from emails, phone numbers or IP addresses to socio-demographic information and consumer preferences (Funakoshi et al., 2020). A report by Duke Sanford’s Cyber Policy Program concludes that, in the United States, ‘there are virtually no controls on the data brokerage industry’ (Sherman, 2021: 12). The capabilities of geolocation data justify why research is needed to understand how this information is used and by whom. Hersh (2015) analyzed a perceived voter database and surveyed campaign professionals to understand the data available to political campaigns. Kreiss and McGregor (2018) interviewed staffers at technology firms which contributed to campaign advertising and communication. Serazio (2014) interviewed campaign consultants about expanding the spaces for political communication online and segmenting audiences for more targeted communications. Dommett (2019) mapped actors involved in political campaigns along the axes volunteer versus professional and novice versus expert, thereby accounting for the vast pools of volunteer workers in data-driven campaigns as well as hired outside counsel. As Nielsen (2012: 141) holds, ‘predictive modeling is generally done by specialized outside consultants, and not campaign organizations themselves’. Our study explicitly focuses on experts in the domain of geolocation data – digital political strategists who aid campaigns. We posit the following research question:

RQ: How do digital political strategists perceive (a) the effectiveness, and (b) the propriety of uses of geolocation data in data-driven political campaigning?

Benefits of expert interviews

We chose open interviewing techniques based on topical guides, as these are particularly suitable for expert interviews (Meuser and Nagel, 2009). It is common practice for studies in computational politics to rely on interviews as the primary method of data collection (Kreiss, 2016; McGregor, 2020; Woolley et al., 2018). They often serve as ‘vehicles for exploring people’s explanations’ (Lindlof and Taylor, 2011: 174) of a phenomenon experienced through personal encounters. Expert interviews aim to discover knowledge held through ‘specific functions (…) individuals have with regard to problems – whether by virtue of a professional role, or as a volunteer’ (Meuser and Nagel, 2009: 24). By drawing a sample of top-tier industry professionals (cf. Kreiss, 2016; McGregor, 2020), there is an innate opportunity to gather insights from highly experienced strategists whose campaign-trail knowledge extends across several election seasons (see Appendix I for sample interview questions).

Upon completing each interview, researchers created thematic memos reflecting important takeaways from each conversation (Rossman and Rallis, 2017). Multiple researchers sifted through and repeatedly read 233 single-spaced pages of interview transcripts and memos to identify themes. As Meuser and Nagel (2009: 35) suggest, this analytical step focuses on the identification of ‘thematic units, that is passages with similar topics which are scattered about the interviews’. In the next step, we compared themes amongst each other, restructured and regrouped them accordingly, and ultimately clustered them into four overarching themes that represent generalizations at a theoretical level across interviews (Cresswell, 1994; Meuser and Nagel, 2009). While the methods applied in this study aim to identify commonalities in the data, we also allude to noteworthy tensions and contradictions in the data in the discussion section.

Geolocation data and political campaigning

Industry professionals emphasized that geolocation data were a foundational pillar for bolstering political advertising and campaign efforts. Interviewees reinforced the importance of geolocation data for identifying persuadable voters. Sam, a digital political strategist, noted that advertising to sway persuadable voters is pointless if those persons live outside the district where votes are needed. Jessie, a digital strategist with a leading position at an enterprise software platform, reflected on the process of how political campaigns go about locating their prospective voter-hopefuls. This process involves gathering mobile numbers, which provide organizations with information about where voters live and who they are. More importantly, this information can subsequently be used to match people with their voter registration records. Interviewees pointed out that one of the reasons why geolocation data became an important asset in political campaigning was because in the United States elections are increasingly won by very slim margins. Indeed, the 2020 election was the ninth consecutive presidential race with a single-digit victory margin (Skelley, 2019). This further underscores why political campaigns are constantly on the lookout for cutting-edge technology to harness their power of winning extra electoral votes. Percy, a digital strategist at a large advertising agency, confirmed that geolocation data tell a lot about people’s behavior. Once matched with people’s voter history, geolocation data did help them identify their core group of ‘persuadables’. Tyler, a digital strategist in a leading position at a voter data firm, explained what happens after political campaigns choose and subset their persuadable voters into a separate data set. Once that was done, they said, they would begin targeting them through a wide range of channels – mailed handwritten postcards, telemarketing, text messaging, or targeted Facebook advertising. The extent to which advertising firms themselves are involved in the implementation of these processes varied. Interviewees reported that political campaigns themselves had varying levels of technical know-how, which, in some cases, led to outsized expectations campaigns had toward advertisers. In other instances, political campaigns themselves had very little knowledge; therefore, they required ‘a little hand-holding’ (Hayden, a digital strategist in a leading position at a voter data firm).

As part of our inquiry, we were also interested in the merits and uses of geofencing as one particular geolocation technology. It refers to virtual perimeter boundaries erected around a certain locale, and which allows advertising to be distributed to devices that enter that area. We asked interviewees if they found geofencing to be helpful. Across conversations, the possibility to identify rally participants emerged as a relevant use case for political campaigns. Tyler said, geofencing would allow their team to identify cell phones of the rally participants, which further could be matched to phone numbers in the voter files. This, in turn, would give campaigns insights into participants’ previous voting behaviors and sometimes even disclose their political affinity. Illustrating a popular geofencing use case, Jessie said:

That’s why you see these great stats that come out like at Trump events, you see people saying, ‘Okay, (...) 25% of the audience was Democratic. And another 20% are people that didn’t vote in 2016.

Jordan, a digital strategist working in political consultancy, noted that a company focusing on preparing voter files would typically supply geolocation companies with people they would like to target:

We’ve kind of cleaned it up and spliced it. So, it goes down to (…) the census block, right. So, you can see where people are that way. (…) A lot of those geolocation companies, they sit on device ID information and IP and other sorts of digital touch points. And they’ll handle that match process.

Geofencing helps locate people by their device IDs. Besides tracking voters in real-time, our interviewees also emphasized that geofencing could be used for tracking voters’ long-term mobility. Among populations that physically moved within or out of state between election registration cycles, geofencing helps bring people back on a campaign’s radar and further reach them with targeted advertising.

Hype versus Realities Surrounding Geolocation

Interviewees pointed out that there was a discrepancy between how the public and them – digital political strategists – perceived the efficacy of location-based political campaigning. As noted in the previous section, interviewees found geolocation data an important asset in political campaigning as it contributes to the basic function of identifying persuadable voters. Yet, at the same time, they also indicated that geotargeted advertising was not always as precise as one would think. One interviewee, Sam, put voters’ efficacy assumptions about the technique into sharp relief, describing how micro-targeting was not as granular as people thought:

I only want left-handed women who have more than three children in the household and their number one issue is cannabis legalization, and they’ve got post doctorate degrees - it’s not real. It’s like, ‘Oh, we’ve micro-targeted this message and it’s only for like these 200 people’ – you’re not going to reach them. There’s not enough of them, that process isn’t that efficient.

While geolocation data are an important piece of information about voter behavior, Tyler said that they are not near the top in the hierarchy of data points available to political campaigns. Rather, door-knocking and face-to-face campaign communication were still ‘so overwhelmingly the most powerful techniques for getting your message in a democratic system to the voters’.

Location-based targeting is not cheap. George, a digital strategist working on political campaigns, said that ‘except for really specific things like “your town votes here” kind of things, the cost often offsets the benefit, because the more you target, the more you’re gonna pay’. The efficacy of geolocation-based targeting was put into question by interviewees who emphasized that the method was neither panacea nor exact science, somewhat inefficient, and that targeting was often based on approximations. Campaigns typically select a state or a district to send out targeted ads, assuming that the people viewing those advertisements will be within the desired geographical bounds. Sam noted that this was often not quite the case, for reasons such as outdated data or IP addresses not being bound by geography.

Geotargeting is not limited to politics; rather, this arena is one that draws a higher level of scrutiny than uses of geolocation data by other industries. River, a digital strategist in a leading position at a political campaign and advertising company, pointed out to us how the use of geolocation-based targeting in political contexts was quicker to draw public attention than commercial uses of the same technology. In other words, what drew public interest the most was not so much the technology itself, but who used it and toward what end.

Despite contradicting much of the hype surrounding traditional uses of geolocation data, interviewees also highlighted a range of use cases in which geolocation data greatly enhanced political campaigns or allowed targeting that otherwise could not have been realized. Those use cases seemed to validate voter concerns. This ostensible match between hype and reality came into sharp focus in discussions of geofencing. Jessie outlined how they were essentially able to geofence any particular area that their clients requested. Geofencing was done through placing a ‘geofence’ around any location, big or small. Clarke, a digital political strategist working as a media director, explained they were able to draw a circle around a particular home and target devices within the radius of that building. Other interviewees, however, said, geotargeting through geofencing has its own limitations that may impact the overall precision metrics. Using an example of a Trump rally, Jordan said:

Let’s say that President Trump had a rally somewhere [with] more than 6,000 people. And let’s say that it was in a particular arena or venue and you did a geofence process over, let’s say four to five hours. And you may get the parking attendant who no way in hell is gonna go for Trump or you may get a paramedic, or a security officer who’s not even registered to vote. There’s a number of different kind of variables in there that’s not going to give you 100% accuracy.

Combining geolocation data with other sources

Geolocation data are only one layer of behavioral data available to political campaigns among many others. But voter registration files are the prime point of departure for identifying who the voters are. Vendors supplying voter registration rolls to political campaigns put a great amount of effort in streamlining those files. Tyler said:

There’s a tremendous amount of cleaning that has to be done (…): standardization, cleaning, and then appending of a tremendous amount of consumer data, modeled data, to old telephone numbers, email addresses to make the data useful to campaigns.

Among other data fields available that get added to a voter profile are ‘things like [your] likely educational attainment, likely income, likely net worth (…) – fields that you would not find in the voter data at the Secretary of State’s office’. But this information gets merged with the voting history to ‘see if you’re somebody who needs more push (…) to get out and vote’, Jordan said.

While somewhat more ancillary to our focus on geolocation data, interviewees emphasized the critical importance of behavioral and attitudinal data in identifying and targeting persuadable voters. One participant, Percy, recalled a campaign where they used polling data to better learn about the political attitudes of their targets:

There are people who consistently vote in low turnout elections. You make sure you talk to those people first, if they’re doing a vote, you should talk to them. And then there was polling that indicated who had the greatest levels of dissatisfaction with the incumbent school board and so we were able to aim creative at those folks to deliver the win for the change slate.

Other behavioral patterns, like a voter’s online news consumption history, can be additional sets of cues for political campaigns about their potential targets. Describing examples of different voter profiles, Clarke said:

For example, if you’re someone who reads articles about all candidates, (...) if you’re someone who’s reading both sides, you’re probably at least a little bit more open-minded. You’re stepping outside of your own echo chamber a little bit more. And in some ways, you’re kind of raising your hand.

In that specific use case, as Clarke noted, the campaign ranked that voter type higher in priority. In Percy and Clarke’s examples, other behavioral data were important in identifying types of voters campaigns wanted to target. But there are instances when campaigns specifically would use voters’ geolocation data to target them with ads leading up to an Election Day. Describing the procedure, Jordan said, they would first rank people’s likelihood to come out and vote based on other demographic and behavioral data they have on file. Then, they would combine that scoring system with A/B testing:

I would have a treatment group and a control, which I may or may not have done (…) in 2018. And what I did was, I had 125,000 records of voters that we knew were Catholic based on behavior and that we knew were actual voters, you know, they were either high, mid or low propensity.

Based on the scoring system, Jordan said, there was a certain number of people who ‘were going to be able to find their way’ to the polling booth regardless of intervention, hence they were used as a control group:

We then took out 25,000 and treated that as a control. We give them no further treatment. And then we used mobile ads based on geolocation data, and we use peer-to-peer text. Then (…) we had to wait almost four months until after Election Day. And what we were able to do is then compare the treatment and the control. What we found was the treatment group had a higher rate of turnout, almost 12%, which is really significant.

In the case described by Jordan, it was relatively clear who their target audience was. Not only did they identify persuadable voters, but they were able to empirically test and validate the efficacy of their campaign measures while highlighting the importance of geolocation data in reaching their persuadable voters.

Potentials for misuse

Interviewees discussed several use cases for geolocation data that involved working around existing regulatory restrictions and loopholes. Some practitioners indicated that combining personal data from a variety of sources could sometimes serve as a workaround when regulatory restrictions would otherwise pose obstacles. These regulations pertain especially to certain demographics that companies are otherwise not allowed to use for targeting, such as race or health-related characteristics. To get around these limitations, interviewees like Sam reported the possibility of using other behavioral data as a sort of proxy for making probabilistic approximations about people’s social identities. More specifically, geolocation data can be combined with other sources of data, such as cookie targeting or Facebook lookalike audiences. With Facebook lookalike audiences, interviewees described how audiences could be targeted based on shared behavioral features. Murphy, a digital strategist working on political campaigns, described the complex use of data assembly through backdated geofencing:

We say, you know, ‘I want everyone at 1600 Pennsylvania Avenue between December 31 and January 1 of 2019 to 2020’, and they will actually pull all the mobile device IDs (...) unique to your personal cell phone (...) and then we can upload that list into Facebook or into other ad platforms to target them directly. Or use that mobile device ID to feed it to one of our data partners who can then pull data insights for users who were in that venue.

Geofencing around specific venues of interest can allow targeting people based on their professional occupation. George described an example in which combining geolocation information with other data was used to work around limitations on public political ads in Washington, D.C. to overwhelm Capitol Hill employees with messaging:

You combine geo-targeting with employer targeting. You can really direct messages almost specifically to legislative staff or Hill staff (…) you’re going to pick up the reporters, the lobbyists who are visiting the Capitol and nearby. So, things like that could be really fun. You can create the illusion of ubiquity, right? Combine it with some posters, billboards, ads in D.C. The Metro won’t run political ads anymore, but we used it when I was at an environmental group two years ago, we used to buy ads at the Capitol Hill-facing Metro stops.

In the absence of a federal privacy law, there are ways political campaigns can access and mine voters’ personal data. However, our interviewees agreed that government regulation of the data broker industry was imminent. Once implemented, those ‘hurdles’ would effectively limit the political advertising business compared to its current state, said Billy, a digital political strategist working in business development. There was a sense among some interviewees that, in the absence of regulation around geolocation data, some individual companies’ moral compass provided protections while other industry members would compromise on how far they would be willing to go in their use of geolocation data.

Interviews revealed a wide range of ways of how campaign strategists decided to use geolocation data for targeting. Some use cases were more extreme than others. Jessie, for example, described the possibility of geofencing a governor’s mansion to identify and target people who visited the residence for private events and who would possibly be voting on a particular piece of legislation in the near future, or who might be relevant donors for a cause:

We can actually geofence that private residence, see what devices showed up on the night of the big fundraiser, and then magically see if we know exactly who they are. You can actually build donor lists. That way you can build likely voter issues.

Conversely, other interviewees emphasized that these types of granular identification of persons based on geolocation data were not permissible. Nicky, who is a digital strategist in a leading position at a political advertising company, said, at times, clients themselves (i.e., political candidates) would push back, finding the technology to be ‘creepy’ or not be worth the hassle: ‘It’s a matter of accuracy versus a matter of pushback (...) if you wanted to vote and persuade one person, theoretically, you can do that’. Still, other participants described voluntary forms of industry self-regulation that governed their uses of geolocation data and technologies. Clarke emphasized that they could not target a single house. Depending on the client they were working with, their fences would have to be drawn to include at least 10,000 addresses or more. These discrepancies in the size of a locale that is being targeted point to the lack of industry standards on what is reasonable and ethically acceptable. Companies themselves would often arbitrarily impose their own rules, for instance saying that at least 1000 people would have to be included in a targeted area to avoid isolating persons or public officials that could otherwise be harassed (Jessie). In other instances, campaigns requesting data would themselves dictate their own parameters for targeting. Charlie, digital political strategist in a leading position at a consulting firm, said:

We use it all the time. For the national political party I’m working for, we geo-target and geo-fence donors’ homes and deliver individual ads to them at their homes. We get down to that level.

Our interviews demonstrate that political micro-targeting using geolocation data opens up a new frontier for serious ethical concerns and potential for misuse of personal data. Some interviewees like Sam called for industry self-regulation as a solution to preempt actual regulatory action from legislators, because:

If Foursquare lets a bunch of Republicans harass Nancy Pelosi at home, it won’t be long before Nancy Pelosi introduces legislation that makes what Foursquare does completely illegal.

Geolocation data and political campaigning

Political campaigns are bound to operate based on geographically defined campaign districts. They often do so using rather ‘mundane tools’ (Nielsen, 2011: 759) in their geolocation pursuits – such as contacting people via mail or by going door-to-door. Geolocation data, however, are an emerging asset in campaign data ecologies that help political candidates reach and possibly mobilize their constituents when voter margins remain very slim. Simultaneously, as cohorts of the voter population become more and more mobile, geolocation data have been helpful in locating the transient voters and bringing them back into the campaign universe.

Hype versus realities surrounding geolocation

Prior research has alluded to at times misleading narratives about data-driven campaigning (Baldwin-Philippi, 2020; Dommett, 2019). In our study, we find this perception to be mirrored in discussions with digital political strategists. Interviewees pointed out how the efficacy of geolocation-based targeting was not quite as precise as public narratives about it would lead one to believe. At other times, when the number of targets is too small, it may become too expensive to justify the total cost. It is reasonable to surmise that it is indeed not in the best interest of these practitioners to feed the public’s pre-existing privacy concerns about the technology’s location tracking capabilities.

Combining geolocation data with other sources

Geolocation data represent only one of many sources of information that political campaigns can access about their voters. On top of voter files, campaigns mesh a wide range of behavioral and attitudinal data to identify their core supporters or at least those they hope to sway to their side by means of campaign messaging. Voter profiles constantly get updated with polling data to provide more current insight into people’s shifting attitudes as an election season comes to a close.

Potentials for misuse

Industry professionals acknowledged that with only few rules in place, geolocation data use is largely unregulated. It could be used toward targeting very specific, small groups of individuals. What happens in the end often depends on individual organizations’ ethical considerations about when and how to use location-based targeting. Extremely narrow targeting, as interviewees acknowledged, can open the floodgates for targeted abuse and harassment of certain individuals. What emanated as a noteworthy contradiction in our study was the proclivity among interviewees to downplay media fears around geolocation data, arguing that such fears were largely unfounded due to poor underlying data quality or because certain forms of microtargeting using geolocation data didn’t make business sense. At the same time, interviewees described precisely how geolocation data could indeed be abused. Participants even provided examples for abuse such as the possibility of geotargeting devices at a governor’s home, or the specter of someone harassing Nancy Pelosi at her home through geotargeting, as pointed out by an interviewee, a threat that rings eerily prophetic after the violent assault on her husband, Paul Pelosi, in October 2022 (Karni and Arango, 2023).

Our study contributes to the literature on data-driven political campaigning in three distinct ways. First, we expand the scope of research on data-driven political campaigning (i.e., Baldwin-Philippi, 2019; Dommett, 2019, Hersh, 2015) by foregrounding the critical role that geolocation data play in political persuasion. Methodologically, we contribute to the scholarly discourse about industry experts using data-driven technologies in the political space to influence electoral outcomes (Kreiss and McGregor, 2018; Kreiss et al., 2018). Third, we add to the current policy debates on personal data privacy, fostering a conversation about data misuse. The possible harms from the abuse of personal data, described by our interviewees, mirror the concerns previously raised by data privacy advocates (Cahn and Laperruque, 2021). Many of the dangers that stem from data misuse, such as the targeted harassment of specific individuals, are not limited to the domain of geolocation data. In fact, they speak to the lack of blanket privacy protections in the United States. For example, the moral panic and nationalist politics fueling a potential ban of the social media platform TikTok in the United States have been described as misplaced in light of the fact that highly personalized user data still can freely and legally be purchased from U.S. data brokers (Angwin, 2023). Regulating the data sources and tracking-enabled technology is as important as reining in the larger data broker industry that continues to go unchecked.

Our study is not without limitations. While data-driven campaigning is not unique to the U.S. context, the interview subjects we spoke with detailed their experiences exclusively as they pertained to U.S. elections. Therefore, our findings may not generalize to other electoral contexts. Isolating our interviews to practitioners in the U.S. context, however, is important because the freedom to use geolocation data in other countries is significantly more limited (Bennett, 2020), and the U.S. example is highly idiosyncratic. Furthermore, as industry professionals detailed their perceptions on the efficacy and pervasiveness of geolocation within the larger domain of data-driven campaigning, it is important to interpret these statements with caution. It is plausible that perceptions of efficacy and perceptions of risk and concern regarding the abuse of geolocation data differ greatly between industry professionals and the voters who are targeted.