Development and History

There will be two x 30 minutes presentations:

i) Anwesh will be presenting Media Bias and Information Bubbles: Evidence from Reporting of Pre-Election Polls on YouTube

Abstract: A large share of the economics literature on media bias focuses on framing or slant, rather than information selection. At the same time, growing concerns about information bubbles and the “polarisation of reality”, particularly in the US where media markets have strong partisan sorting, suggest that agenda setting may play an equally important role. I study the existence of such information gaps in the context of pre-election polling, where the underlying information is verifiable, but media outlets remain free to choose which polls to report. I construct novel data on poll reporting on YouTube, one of the most widely used news platforms in the United States. Using transcripts from 94 YouTube channels covering U.S. news and politics, together with an LLM-based extraction filter, I build a structured dataset of all polling-related information reported in each video. I document three main findings. First, at any given point in time, Republican-leaning channels report more information on polls where Trump is ahead relative to Democratic-leaning channels, establishing the presence of information bubbles even in a setting with hard, publicly verifiable information. Second, I find that reporting favourable information for the channel's preferred candidate generates noisy but generally positive effects on viewership. Third, I find that conditional on reporting about polls, these information bubbles are relatively more driven by the intensive margin -- channels selectively sampling from different ends of the distribution, than mechanically through the amount of information in each video.

ii) Yanjun will be presenting From Calories to Calcium: Reduced-Form and Structural Evidence on Soda–Milk Substitution from U.S. Scanner Data

Abstract: This paper examines the substitution patterns between milk and soda, with particular attention to demographic heterogeneity. Using the Nielsen Retail Scanner dataset, I estimate demand parameters through a novel share-to-share regression framework. The results indicate that while soda and milk appear nearly independent at the store level, they behave as strong substitutes at more aggregated market levels. Flavored milk, in particular, emerges as a close substitute for soda, consistent with its stronger appeal among younger consumers. I then adopt a structural approach by estimating a multinomial logit demand model using household-level scanner data. This demand model allows for richer individual heterogeneity, and the resulting structural estimates closely mirror the reduced-form findings. Taken together, these findings suggest that milk and soda are strong substitutes, especially flavored milk and particularly among households with children. Finally, I conduct a back-of-the-envelope policy simulation to evaluate how a one-cent-per-ounce sugary drink tax would affect the market shares of milk and soda, and how these effects differ across demographic groups. The results provide new insights into the evaluation of sugar tax policies

Title to be advised.

Two 30 minutes presentations.

Titles to be advised.

Title to be advised.

Title to be advised.

Two 30 minutes presentations.

Titles to be advised.