On the Use of Proxies in Political Ad Targeting

Abstract

Detailed targeting of advertisements has long been one of the core offerings of online platforms. Unfortunately, malicious advertisers have frequently abused such targeting features, with results that range from violating civil rights laws to driving division, polarization, and even social unrest. Platforms have often attempted to mitigate this behavior by removing targeting attributes deemed problematic, such as inferred political leaning, religion, or ethnicity. In this work, we examine the effectiveness of these mitigations by collecting data from political ads placed on Facebook in the lead up to the 2022 U.S. midterm elections. We show that major political advertisers circumvented these mitigations by targeting proxy attributes: seemingly innocuous targeting criteria that closely correspond to political and racial divides in American society. We introduce novel methods for directly measuring the skew of various targeting criteria to quantify their effectiveness as proxies, and then examine the scale at which those attributes are used. Our findings have crucial implications for the ongoing discussion on the regulation of political advertising and emphasize the urgency for increased transparency.

Figures (8)

• Figure 1: Estimating the skew of targeting interests using the overlap between custom audiences and inferred interests. Step 1: Consolidate voter records from all states where voters self-report race. Step 2: Create Custom Audience out of party- or race-uniform samples of voters. Step 3: Use the delivery_estimate API to estimate how many users can actually be reached in each custom audience. Step 4: Use the delivery_estimate API to estimate what fraction of users in each custom audience share a given interest (here: Ted Nugent). Step 5: Use Eq. \ref{['eq:skew_def']} to calculate skew.
• Figure 2: Estimating the skew of targeting interests using the Audience Insights data. Step 1: Obtain the list of interests used by advertisers from the Facebook Ad Library. Step 2: Obtain Facebook Pages popular among the users associated with a given interest. Step 3: Obtain the domain of the external URL self-reported by that Facebook Page. Step 4: Look-up the audience skew of each domain. Step 5: Average the audience skews of known domains and assign it to the interest.
• Figure 3: The coverage of interests between populations are highly correlated. Still, a non-trivial fraction of interests reveal large skews (color-coded in the figures) and could be used to selectively target populations that are not explicitly targetable.
• Figure 4: The skews calculated by audience and by page are correlated
• Figure 5: Examples of ads whose advertisers appear to use proxies. Virginia House Republicans (on the left) target users interested in (among others) NASCAR, Big-game hunting, and Deer hunting. On the right, CA Black Power Network targets users interested in The Shade Room, Basketball Wives, Love & Hip Hop.