- Garrett Johnson, Assistant Professor, Marketing, Questrom School of Business, Boston University
- Tesary Lin, Assistant Professor, Marketing, Questrom School of Business, Boston University
- James C. Cooper, Professor of Law, Antonin Scalia Law School, George Mason University
This blog article is derived from the authors’ research paper titled COPPAcalypse? The YouTube Settlement’s Impact on Kids Content Creation, a project of the Economics of Digital Services initiative led by Penn’s Center for Technology, Innovation & Competition (CTIC) and The Warren Center for Network & Data Services. CTIC and the Warren Center are grateful to the John S. and James L. Knight Foundation for its generous support of the EODS initiative.
As the Internet has evolved, online display advertising has grown into a multi-billion-dollar industry, surpassing television advertising for the first time in 2017. Online content creators rely on the revenue generated by display advertising to monetize the services that they provide to consumers, often for free. The display advertising that consumers see can be based on the content they are viewing (contextual ads) or on personal information used to predict likely interests (interest-based advertising (IBA)), which is gleaned from persistent identifiers that track users online. Because ads served based on information about the viewer, not merely context, are more likely to lead to conversions, IBAs sell at a premium compared to impressions based only on context. Indeed, the weight of the empirical evidence suggests that IBAs generate between two to three times more revenue than display ads that are served based only on context.
While consumer tracking enhances monetization, it has long been a focal point for privacy concerns. Indeed, the General Data Privacy Regulation (GDPR) and the recently enacted California Consumer Privacy Act (CCPA) and California Privacy Rights Act (CPRA) have set up various regulatory requirements to curb this type of tracking. What is more, Congress is considering major privacy legislation that would severely restrict the use of IBA, and the Federal Trade Commission (FTC) recently announced the beginning of a rulemaking proceeding aimed at curtailing “commercial surveillance.”
Given the privacy concerns associated with online tracking, an important empirical question is to what extent do consumers share in the benefits of IBA by enjoying more—and higher quality—content than they would have access to if content providers were allowed to sell impressions based only on context? Providing a better understanding of the extent to which putting limits on the type of tracking that makes IBA possible will degrade the quantity and quality of online content can help policymakers strike the right balance between privacy and free content generation.
Addressing this important question empirically, however, is difficult. Because online content providers in the United States are not barred from collecting consumer data used in targeting, the choice to serve IBA or contextual ads is endogenous. One exception to this general rule, however, is the Children’s Online Privacy Protection Act (COPPA), which broadly prohibits operators of online services directed at children under 13 from collecting personal information without providing notice of its data collection and use practices and obtaining verifiable parental consent. Importantly, a 2013 Federal Trade Commission rule defines persistent identifiers (such as cookies) used to implement IBA as personal information subject to COPPA. Obtaining verifiable parental consent is rarely cost-justified for ad-supported content, so as a practical matter this means that creators of child-directed content are barred from using IBA.
While COPPA provides an exogenous restraint on publishers’ ability to use IBA for child-directed content, the FTC regulation has been in effect for nearly a decade. As a consequence, the data needed to identify a causal impact of the regulation on content creation are unlikely to exist. Fortunately, the 2019 YouTube COPPA settlement with the FTC provides a natural experiment to measure the impact of IBA on content. In September 2019, YouTube entered into a consent decree with the Federal Trade Commission (FTC) to settle charges that it had collected personal information in the form of persistent identifiers from children to serve IBAs against child-directed or “Made for Kids” (MFK) videos. Prior to the settlement, YouTube channel owners who provided MFK programming—most of whom are small businesses or hobbyists—could use IBA, presumably because (1) YouTube was a general audience Web site, and (2) someone has to be at least 13 to set up a Google ID (which is what makes tracking possible). The consent decree, however, required YouTube to identify all MFK programming and to stop collecting any personal information—including persistent identifiers that are used to implement IBA—from visitors to these channels. The upshot was that after January 2020, no MFK programming could employ IBA; all MFK advertising was based only on context.
In this manner, the 2019 consent decree serves as an exogenously-imposed ban on IBA for MFK programming on YouTube. This natural experiment allows us to implement an empirical strategy that uses non-MFK channels as a control group to identify the effect of eliminating IBA on MFK content creation on YouTube. We sample 5,441 YouTube channels—MFK, non-MFK, and mixed—from July 2018 through January 2021. We use YouTube’s API to collect rich video-level data from the 1.8 million videos these channels uploaded during the sample period, including whether it is marked as MFK, the date and time it was uploaded, number of views and likes, and length.
Results and Policy Implications
Using various econometric models, we find a robust and statistically significant negative impact on the number of MFK videos available for viewing as a result of the IBA ban.
Specifically, we estimate that compared to non-MFK channels, pure MFK and mixed channels uploaded about 11% and 9% fewer videos after January 2020, respectively. Digging a little deeper into these numbers, we also find evidence that some channels altered their mix of MFK and non-MFK videos toward the latter. For example, pure MFK channels on average reduced their share of MFK videos by about 2%, compared with a 7% average reduction in the share of MFK videos for mixed channels. This pattern is consistent with what one would expect—channels that are already posting a mix of MFK and non-MFK programming are likely to find it easier than pure MFK channels to pivot away from MFK programming after it became relatively less lucrative. Indeed, at the channel level, we see that almost 42% of mixed channels move to a no-MFK model after the consent decree takes effect. This compares to only 11% of pure MFK channels that move to mixed or no-MFK after the consent decree. In future work we plan to examine how the elimination of IBA for MFK programming impacted dimensions of quality, such as views and likes.
Overall, our preliminary results are consistent with what the basic economics of online content provision would predict: if the price advertisers are willing to pay for access to the eyeballs that online content attracts falls, at the margin, there will be less (and lower quality) content.