Selling Models
Can privacy technologies replace cookies? Ad revenue in a field experiment
Zhengrong Gu, Garrett Johnson & Shunto Kobayashi
Proceedings of the National Academy of Sciences, 12 May 2026
Abstract:
Digital advertising finances much of the open web, yet relies on tracking technologies that regulators increasingly seek to restrict. In response, industry has developed privacy-enhancing technologies intended to preserve advertising performance while limiting data collection, but their economic effects remain largely untested. We study this question using an open, industry-wide field experiment jointly overseen by Google and the UK Competition and Markets Authority, in which Chrome users were randomly assigned to browse with third-party cookies enabled, with cookies disabled, or with Google’s Privacy Sandbox replacing cookies. Combining this experimental variation with proprietary data from a major ad management firm, we analyze more than 200 million ad impressions across over 5,000 publishers worldwide. Removing third-party cookies reduces publisher advertising revenue by 29.1%. Privacy Sandbox recovers only 4.2% of this lost revenue; this estimate reflects observed adoption and performance during the study period and may reflect modest industry adoption. Privacy-preserving auctions also increase ad latency, reducing impression delivery and further limiting revenue performance. Together, these findings provide a large-scale experimental benchmark for evaluating privacy-preserving reforms and demonstrate the difficulty of reconciling privacy protection with the economics of online content provision.
Are Apologies Always the Best Policy? Apologies for Service Failures Backfire When Consumers Are Not Aware of the Failure
Mason Jenkins, Paul Fombelle & Mary Steffel
Journal of Consumer Research, June 2026, Pages 1-21
Abstract:
Technological advancements enable firms to anticipate service failures and apologize to consumers more effectively than ever. However, are apologies always the best policy? Five experiments, including a large-scale field experiment, demonstrate that apologies backfire when consumers are not aware of the failure. Apologies decrease consumer satisfaction, trust, recommendation intentions, and repatronage behavior. This is because apologies increase awareness that a failure occurred. Accordingly, apologies backfire for both mild and severe failures so long as there is room to increase awareness of the failure. Moreover, apologies backfire more than notifications about a service failure because although both increase awareness of the aspect of the service experience that was a failure, apologies uniquely increase awareness that this constituted a failure. Beyond increasing awareness, apologies backfire by decreasing perceived service quality and competence. However, apologies also increase perceived warmth and honesty. Consequently, apologizing can be beneficial when consumers are already aware of a failure because apologies are less likely to backfire and more likely to enhance positive perceptions of the firm. These findings challenge the conventional wisdom that apologies are always the best policy when something goes wrong and provide insight into when and when not to apologize.
When pursuing hedonic experiences leads to less hedonic outcomes: Hedonic sampling drives the persistence of false beliefs in reward-rich food environments
Niklas Pivecka, Sonja Kunz & Arnd Florack
Journal of Experimental Psychology: Applied, forthcoming
Abstract:
Existing research explains the widespread belief that unhealthy food is tastier than healthy food by how food is portrayed in the media and social networks and by the notion that food cannot serve two purposes (health and pleasure) at the same time. However, these explanations cannot explain why people do not change this belief despite the increasingly large variety of healthy and tasty food options. We argue that it is the abundance of tasty foods in the environment, combined with people’s motivation to eat palatable foods, that prevents individuals from changing their beliefs. In three online studies and one taste experiment (total N = 976), we show that individuals’ food choices in reward-rich environments with many tasty foods, but not in reward-poor environments with few tasty foods, led to a consolidation of initial beliefs, even when there was no relationship between food healthiness and taste or a relationship that contradicted their beliefs in the experimental food environment. Our findings suggest that a basic hedonic sampling mechanism contributes to the persistence of food beliefs, such as the belief that unhealthy food tastes better, and should be considered when trying to change unhealthy diets.
When Certification Backfires: Negative Spillovers on Noncertified Offerings in Airbnb Plus
Heeyon Kim, Qian Wang & Martina Montauti
Organization Science, forthcoming
Abstract:
Research on certification has largely emphasized its positive effects, showing that certification can enhance trust and legitimacy for certified actors and in some cases, generate positive spillovers that benefit uncertified counterparts. Yet, certification programs are frequently scaled back or discontinued, suggesting that their broader consequences may be more complex than commonly assumed. In particular, certification can produce negative spillovers by sharpening distinctions between certified and noncertified actors; through a contrast mechanism, juxtaposition with certified offerings heightens the salience of uncertainty and leads audiences to evaluate noncertified actors less favorably, even when underlying quality is unchanged. We test this argument using data from the launch of Airbnb Plus, a program in which Airbnb certified a subset of listings based on verified quality standards. Employing a difference-in-differences design comparing Los Angeles (where the program was launched) and San Diego (where it was not), we find a negative treatment effect on the revenues of noncertified listings in Los Angeles. This negative spillover is attenuated when alternative quality signals are available, when noncertified listings are less directly comparable with certified ones, and when listings are more difficult to substitute outside the platform. Our findings suggest how certification systems intended to enhance trust and legitimacy can under certain market conditions inadvertently disadvantage noncertified actors.
When Goods Were Odds: Do People Prefer Uncertain Goods After Uncertainty Is Resolved?
Beidi Hu, Siyuan Yin & Alice Moon
Journal of Consumer Research, forthcoming
Abstract:
Much of the uncertainty people face is eventually resolved (e.g., a person entered in a raffle eventually learns what prize they have received). How do people evaluate goods (e.g., a prize of a $50 gift card) resulting from uncertain promotions (e.g., raffles)? Seven experiments (total N = 12,128) provide evidence for an uncertainty spillover effect: People prefer goods originating from uncertain prospects compared to those that were always known. This effect appeared both with naturalistic scenarios (study 1) and with incentive-compatible decisions (study 2). The authors propose this effect arises because uncertainty induces a perception that the outcome is superior relative to salient downward counterfactuals (study 3). Supporting this idea, this effect: (a) weakened when downward counterfactuals were salient for certain goods (study 4), (b) weakened when the worst outcome from uncertainty was realized (study 5), and (c) reversed when uncertainty involved losses (study 6). Lastly, this effect carried over to products associated with previously uncertain goods (study 7). These findings demonstrate that the influence of uncertainty persists beyond its resolution, shaping the evaluation of goods derived from uncertain prospects.