Tesary Lin

We examine how privacy restrictions affect online content creation and consumption by evaluating the impact of YouTube’s settlement with the Federal Trade Commission over violating the Children’s Online Privacy Protection Act. Under the settlement, YouTube limited personalization for made-for-kids (MFK) content starting in January 2020, and this included personalized ads and content engagement features such as subscriber notifications and playlists. We study the resulting impact on 5,066 top American YouTube channels by comparing the MFK content creators to their non-MFK counterparts using a difference-in-differences design. On the supply side, MFK content creators produce 18% less content and pivot toward non-MFK content production. MFK content creators also invest less in content quality: the proportion of original content falls by 9% and manual captioning drops by 28%, whereas viewer content ratings fall by 9%. On the demand side, views of MFK channels fall by 20%. The restrictions also affected market competition, increasing concentration of both content creation and viewership among top MFK channels. This paper was accepted by Jean-Pierre Dubé, marketing. Funding: This work was supported by the Questrom School of Business Digital Business Institute, the Program on Economics & Privacy at George Mason University, and the “Economics of Digital Services” initiative by the University of Pennsylvania’s Center of Technology, Innovation & Competition and Warren Center. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2024.05295 .

User interface designs that maximize data collection consent rates may lead to biased data samples; we provide empirical evidence and propose a better alternative.

We discuss the quantitative marketing and economics literatures analyzing the benefits and unintended costs of digital consumer privacy regulation.

In an era where firms are innovating and using more data than ever before for marketing purposes, there is a perceived need for enhanced regulation to protect consumers' privacy. We provide a perspective based on the academic marketing literature that evaluates the various benefits and costs of existing and pending government regulations and corporate privacy policies. We make two key points. First, regulators may want to avoid starting from the stance that data-based marketing and personalization are automatically harmful. Second, regulations and policies may have inadvertent consequences. On the demand side, privacy regulations and policies may exacerbate the digital exclusion of already marginalized segments of consumers. Further, consumers differ in whether and how they benefit from sharing versus not sharing specific data. On the supply side, regulation and policies may disproportionately disadvantage the competitiveness of entrpreneurs and small businesses. Technology platforms are proposing differential privacy solutions that mitigate some of these harms, but, again, in a way that might disadvantage small firms and entrepreneurs.

We study how choice architecture that companies deploy during data collection influences consumers' privacy valuations. Further, we explore how this influence affects the quality of data collected, including both volume and representativeness. To this end, we run a large-scale choice experiment to elicit consumers' valuation for their Facebook data while randomizing two common choice frames: default and price anchor. An opt-out default decreases valuations by 14-22% compared to opt-in, while a \$0-50 price anchor decreases valuations by 37-53% compared to a \$50-100 anchor. Moreover, in some consumer segments, the susceptibility to frame influence negatively correlates with consumers' average valuation. We find that conventional frame optimization practices that maximize the volume of data collected can have opposite effects on its representativeness. A bias-exacerbating effect emerges when consumers' privacy valuations and frame effects are negatively correlated. On the other hand, a volume-maximizing frame may also mitigate the bias by getting a high percentage of consumers into the sample data, thereby improving its coverage. We demonstrate the magnitude of the volume-bias trade-off in our data and argue that it should be a decision-making factor in choice architecture design.

Companies often deploy some form of "choice architecture" when collecting consumer data, designed to nudge consumers towards sharing more private information. This study examines when an emphasis on maximizing the volume of data shared when deploying choice architecture can alter the composition of the collected data, hence creating a trade-off between the quantity and representativeness of data collected. To this end, we ran a large-scale choice experiment to elicit consumers' incentive-compatible valuation for their private Facebook data while randomizing the choice frames they encountered. Within participants, we elicited WTA using a multiple-price list, followed by a free-text entry. Across participants, we randomized the choice default and the price anchor. The default varied between opt-in, opt-out, and active choice. Price anchor was the range of prices in the multiple price list, which was either $0--$50 (low) or $50--$100 (high).

I empirically separate two components in a consumer’s privacy preference. The intrinsic component is a “taste” for privacy, a utility primitive. The instrumental component comes from the consumer’s anticipated economic loss from revealing his private information to the firm and arises endogenously from a firm’s usage of consumer data. Combining an experiment and a structural model, I measure the revealed preferences separately for each component. Intrinsic preferences have seemingly small mean values, ranging from $0.14 to $2.37 per demographic variable. Meanwhile, they are highly heterogeneous across consumers and categories of data: The valuations of consumers at the right tail often exceed the firm’s valuation of consumer data. Consumers’ self-selection into data sharing depends on the respective magnitudes and correlation between the two preference components and often deviates from the “low types are more willing to hide” argument. Through counterfactual analysis, I show how this more nuanced selection pattern changes a firm’s inference from consumers’ privacy decisions and its data-buying strategy.

The Identity Fragmentation Bias

Consumers interact with firms across multiple devices, browsers, and machines; these interactions are often recorded with different identifiers for the same consumer. The failure to correctly match different identities leads to a fragmented view of exposures and behaviors. This paper studies the identity fragmentation bias, referring to the estimation bias resulted from using fragmented data. Using a formal framework, we decompose the contributing factors of the estimation bias caused by data fragmentation and discuss the direction of bias. Contrary to conventional wisdom, this bias cannot be signed or bounded under standard assumptions. Instead, upward biases and sign reversals can occur even in experimental settings. We then compare several corrective measures, and discuss their respective advantages and caveats.