Research

Abstract

Social media platforms have become vital channels for businesses to reach consumersthrough advertising. But in the United States, the digital advertising market in whichthese platforms operate is dominated by a few major players, raising concerns forantitrust regulators. In such a concentrated market, the entry or exit of a singleplatform can reallocate billions in ad spending, affecting businesses and users. TikTok’stemporary suspension in the United States in January 2025 provides a unique naturalexperiment to examine how the removal of a major player would shift advertisingdemand and supply on competitors, specifically Facebook and Instagram, revealing thedegree of substitutability across platforms and the intensity of competition. Using adifference-in-differences approach comparing advertising activity in the United Statesto other countries, we find that Meta ad volume and spend rose by 6.3% and 22.4%, as aresult of the outage, without a corresponding increase in ad impressions. Consequently,Meta ad prices, as measured by cost per thousand impressions, jumped by 12.1%.Shifts in demand were three times greater for larger advertisers relative to smaller ones, suggesting that Meta platforms and TikTok are closer substitutes for larger firms and that a TikTok ban would therefore impose greater challenges on smaller businesses.

Abstract

Asymmetric information can distort market outcomes and harm welfare. This paper examines how cheaper access to review platforms affects consumer behavior and firms’ incentives to improve product quality in markets where information is traditionally limited. I first develop a consumer search model with firms making endogenous quality decisions. The model predicts that lower search costs shift demand toward higher-quality producers, raising firms’ incentives to upgrade quality, especially for firms selling ex-ante lower-quality products. I then use online ratings as a proxy for reduced consumer search costs and estimate their impact on the restaurant industry in Rome, leveraging the abolition of mobile roaming charges in the EU in 2017 for identification. Using a unique dataset that combines monthly Tripadvisor information with administrative social-security records, I find that revenues and total employment in higher-rated restaurants increased by 4–10% after the internet price drop. The effects were particularly strong for less visible high-rated establishments, indicating a complementarity between ratings and navigation platforms. Meanwhile, the probability for lower-rated restaurants to exit the market doubled, while surviving lower-rated establishments hired workers with higher wages and better experience, ultimately improving their online reputation. These findings highlight the significant role of review platforms in mitigating asymmetric information and fostering quality upgrades.

Abstract

How does the internet affect young people's mental health? We study this question in the context of Italy using administrative data on the universe of cases of mental disorders diagnosed in Italian hospitals between 2001 and 2013, which we combine with information on the availability of high-speed internet at the municipal level. Our identification strategy exploits differences in the proximity of municipalities to the pre-existing voice telecommunication infrastructure, which was previously irrelevant but became salient after the advent of the internet. We find that access to high-speed internet has a harmful effect on mental health for young cohorts but not for older ones. In particular, internet access is associated with an increase in diagnoses of depression, anxiety, drug abuse, and personality disorders-- for both males and females - and of eating and sleep disorders - for females only. We find similar results for urgent and compulsory hospitalizations and self-harm episodes. These results suggest that the effect of broadband is driven by a rise in the underlying prevalence of mental disorders and not merely by increased awareness about these pathologies.

Abstract

I exploit the staggered introduction of 3G mobile Internet technology across South Africa to estimate its impact on political participation, electoral competition, voters’ preferences, and protests. Combining granular coverage data with administrative records on municipal elections, I show that in 2016 mobile Internet availability caused a 2-percentage-point increase in voter turnout and a 3-percentage-point reduction in the vote share of the ruling party. The main opponents gained from mobile Internet arrival. The number of parties running for election and the number of protests grew. I provide suggestive evidence that both information and coordination mechanisms could explain the observed results.

Abstract

This study uses a cluster randomized controlled trial to evaluate the impact of a nationwide malaria prevention advertising campaign delivered through social media in India. Ads were randomly assigned at the district level, and the study relies on data from two independently recruited samples (8,257 individuals) and administrative records. Among users residing in solid (concrete) dwellings, where malaria risk is lower, the campaign led to an 11 percent increase in mosquito net usage and a 13 percent increase in timely treatment seeking. Self-reported malaria incidence decreased by 44 percent. Consistently, recorded health facility data indicate a reduction in urban monthly incidence of 6.2 cases per million people, corresponding to 30 percent of the overall monthly incidence rate of malaria. Conversely, the study finds no impact on households living in non-solid dwellings, which face higher malaria risk, nor among rural settlements where such dwellings are more prevalent. To disentangle if this lack of impact stems from ineffective content or insufficient reach, an individual-level trial was conducted (1,542 individuals), ensuring campaign exposure for both household types. The findings indicate an increase in bed net usage and timely treatment seeking for both groups, underscoring the need for improved targeting in social media campaigns to fulfill public health goals.

Abstract

This paper introduces and validates a new methodology for recruiting highly stratified survey respondents using digital advertising platforms. Our approach leverages ad platform APIs with a novel algorithm that dynamically optimizes participant recruitment in real time by minimizing the variance of post-stratification weighted estimates. We demonstrate the method's efficacy through a 1,500-person U.S. study, benchmarking results against gold-standard surveys (GSS, CPS, Pew), a leading online panel provider (Prolific), and an LLM-based synthetic methodology (digital twins). We obtain a mean absolute deviations of 6.1 percentage points from the gold-standard benchmarks, improving on both the online panel provider and LLM-based approaches. We show that this methodology is both cost-effective in the U.S., with a total cost of $0.30 per question per respondent, but also deployable around the world in both developed and emerging economies, where we provide evidence of success in over 33 studies across 23 countries. We present the software as an open-source platform, deployable as a web application on both public and private clouds.

Abstract

We conducted a field experiment on Facebook Messenger to evaluate the effectiveness of two social marketing video series at reshaping gender norms and reducing social acceptability of violence against women in India. To test different formats, we randomly invited participants to watch either a humorous fictitious drama (implicit format) or a documentary with clear calls to action (explicit format), or a placebo series (control). Our results show that the fictitious drama was more successful in facilitating attitudinal change, reducing the acceptability of regressive gender norms and violence against women by respectively 0.20 and 0.17 standard deviations. In contrast, the documentary was more effective in encouraging behavioral change, influencing the willingness to share the videos and promoting information-seeking behaviors. Moreover, in the medium term, documentary viewers were 91% more likely to add a frame anti-violence against women to their Facebook profile picture, a public commitment to social action. The absence of heterogeneous effects across socioeconomic indicators suggests that social media could be a viable platform for conveying social marketing campaigns to marginalized communities.