Case Study 29.2: Cambridge Analytica and the Limits of Psychographic Microtargeting

Overview

No episode in the recent history of political microtargeting has generated more controversy — or more confusion — than the Cambridge Analytica story. The company claimed to have developed a revolutionary psychographic targeting system that used Facebook data to build psychological profiles of individual voters and deliver uniquely persuasive messages calibrated to their personality types. The story was sensational: 87 million Facebook users' data obtained without informed consent, used to elect a president by manipulating voters' deepest psychological vulnerabilities.

The reality, as subsequent research has established, was considerably more complicated. This case study examines what Cambridge Analytica actually claimed, what the evidence says about those claims, and what the genuine lessons of the episode are for understanding political microtargeting.

What Cambridge Analytica Claimed

Cambridge Analytica's core marketing proposition was that it had developed a psychographic targeting approach based on the "OCEAN" model of personality (Openness, Conscientiousness, Extraversion, Agreeableness, Neuroticism). The claim was that:

1. Facebook "like" data could be used to build accurate OCEAN personality profiles for individual voters.
2. Different personality profiles responded differently to specific types of political messaging.
3. Cambridge Analytica could therefore identify the optimal message for each voter based on their personality profile and deliver that message at scale through Facebook advertising.

This claim was built on academic research — primarily by Michal Kosinski and colleagues at Cambridge University, who had published work showing that Facebook likes could predict OCEAN personality scores with some accuracy relative to self-reported survey measures. Cambridge Analytica's commercial proposition was that this academic finding could be translated into a political targeting product.

The Data Acquisition Problem

Before examining whether the psychographic targeting worked, it's worth understanding how the data was acquired — because the acquisition method is central to why the Cambridge Analytica episode became a major political controversy.

The Facebook likes data that underlies psychographic modeling was collected through a Facebook app called "thisisyourdigitallife," developed by researcher Aleksandr Kogan. Approximately 270,000 Facebook users took the app's personality survey and consented to sharing their Facebook data with the app. However, under Facebook's data sharing policies at the time, Kogan's app also collected the likes data of those users' Facebook friends — people who had never consented to share their data with a third-party political targeting firm.

This extended the data collection from 270,000 consenting participants to approximately 87 million people whose data was collected without their knowledge or consent. Cambridge Analytica then used this data (which it had purchased from Kogan) to build its targeting models.

The data acquisition violated Facebook's terms of service and potentially relevant data protection regulations. When the Guardian and New York Times published investigative reports on the data collection in March 2018, it triggered congressional hearings, regulatory investigations in multiple countries, and Facebook's eventual $5 billion FTC settlement for privacy violations.

Did It Work? The Evidence Problem

The more analytically interesting question — separate from the ethical and legal problems with the data acquisition — is whether Cambridge Analytica's psychographic targeting actually worked. This question is harder to answer than it might appear, for several reasons.

The black box problem: Cambridge Analytica did not publish its targeting methodology or results in peer-reviewed form. Its claims were commercial propositions, not scientific findings. The company had every incentive to describe its product as more effective than it was.

The causality problem: Cambridge Analytica worked on multiple Republican campaigns and the Brexit campaign. Some of those campaigns won; some lost. Drawing any causal conclusion about the role of the targeting in those outcomes requires controlling for all of the other factors that determine election results — a notoriously difficult task.

The validity problem: Independent researchers who have tested the psychographic targeting claim — using data from academic experiments rather than Cambridge Analytica's proprietary system — have found mixed and generally modest results. A 2020 study by Matz and colleagues found some evidence that personality-matched political messages outperformed mismatched messages, but the effect sizes were small (approximately 3–11% higher engagement rates) and the experimental context was quite different from a real campaign.

The replication problem: A substantial portion of academic research on psychographic targeting has not replicated cleanly across different studies, populations, and electoral contexts.

What the Episode Actually Reveals

The Cambridge Analytica story is important not primarily for what it tells us about psychographic targeting — the evidence for that technique's effectiveness is weak — but for what it reveals about several structural features of political microtargeting.

The role of scale and data advantage: Even if the psychographic profiles were imprecise, Cambridge Analytica was working with Facebook data at a scale that no campaign had previously accessed for targeting purposes. The sheer volume of data may have provided some targeting advantage through conventional (non-psychographic) matching techniques, even if the psychographic overlay didn't add much.

The credibility of vendor claims: Cambridge Analytica's commercial success — it raised substantial fees from major campaigns on the basis of its psychographic claims — demonstrates that campaign managers and donors are susceptible to sophisticated-sounding targeting claims that have weak empirical foundations. The political consulting industry has strong incentives to claim effectiveness, and campaigns have limited capacity to independently validate those claims.

The regulatory gap: The Cambridge Analytica episode exposed a significant regulatory gap. The data was collected from Facebook with inadequate consent mechanisms. It was transferred to a political targeting firm without users' knowledge. It was used in campaigns that had no obligation to disclose how they targeted voters. None of this was clearly illegal under applicable US law at the time — though it violated Facebook's terms of service and, as regulators eventually determined, was deceptive under the FTC Act.

The public concern about targeting: Perhaps most significantly, the Cambridge Analytica episode crystallized public concern about the use of personal data for political targeting in a way that previous, less dramatic revelations had not. The story of 87 million people's data used to manipulate a presidential election was compelling precisely because it was frightening — and it was frightening because it captured something that felt true even if the specific psychographic mechanism was overstated: that campaigns were using personal data to influence elections in ways that voters didn't know about and couldn't evaluate.

The Aftermath: Platform and Regulatory Responses

The Cambridge Analytica revelations prompted significant changes in the political targeting environment.

Facebook responded by restricting the targeting categories available for political advertising, requiring identity verification for political advertisers, creating a public ad library, and strengthening data sharing restrictions on third-party apps. These changes meaningfully reduced the precision of political targeting available through Facebook, particularly for smaller campaigns that had relied on detailed interest and behavioral targeting.

Several states passed enhanced consumer privacy legislation that affects political data use. California's CCPA and its subsequent strengthening through the CPRA created new restrictions on how consumer data can be collected and used in political contexts. Other states are in varying stages of similar legislation.

The Federal Election Commission considered but has not implemented significant additional disclosure requirements for digital political advertising. The FTC's enforcement action against Facebook addressed the data collection violations but did not directly regulate political targeting practices.

Implications for Understanding Microtargeting

The Cambridge Analytica case study offers several lessons that extend beyond the specific episode.

Extraordinary claims require extraordinary evidence. When a vendor claims to have a targeting breakthrough — psychographic modeling, AI-powered message optimization, unique data assets — the claim should be evaluated with the same skepticism any scientific claim deserves. Campaigns that pay premium fees for unvalidated targeting methods are subsidizing vendor profits, not electoral victories.

The ethics of data use are separable from effectiveness. Even if Cambridge Analytica's psychographic targeting had worked well, the data acquisition methods would still have been ethically and legally problematic. The effectiveness question and the ethics question are distinct.

Regulatory gaps in political targeting remain significant. The specific violations in the Cambridge Analytica case have been partially addressed. The underlying structural issues — opaque targeting practices, limited disclosure requirements, asymmetric information between campaigns and voters — remain largely unresolved.

Public concern is itself politically significant. Whatever the empirical validity of Cambridge Analytica's claims, the episode created genuine political constraints on how campaigns discuss and deploy microtargeting. Campaigns that are seen as aggressive users of personal data for political manipulation face reputational risks that are operationally relevant.

Discussion Questions

1. Cambridge Analytica's claims about psychographic targeting were commercially successful even if empirically weak. What does this tell you about how campaigns evaluate vendor claims? What institutional changes would improve campaigns' capacity to evaluate data analytics vendors rigorously?

2. The psychographic targeting literature shows small positive effects in controlled experiments but has not demonstrated large real-world effects on vote choice. How should this gap between lab findings and real-world application affect how campaigns invest in novel targeting techniques?

3. Facebook's policy changes in response to Cambridge Analytica restricted both the data-sharing that enabled the scandal and some legitimate targeting capabilities that campaigns had relied on. How would you design platform policy rules that address the genuine concerns while preserving legitimate campaign uses?

4. The case study notes that the data acquisition violated Facebook's terms of service rather than US law. Is this an adequate legal framework for governing political data use? What alternative legal frameworks would better serve democratic interests?

5. The chapter argues that the line between targeting and manipulation is enforced primarily by "professional norms, legal requirements, and imperfect scrutiny." Does the Cambridge Analytica episode support or challenge this assessment? What stronger enforcement mechanisms are available?