Research
Working papers
Visibility and Retail Demand: Evidence from Bike Sharewith Avi Goldfarb
Revise & resubmit, Marketing Science
Awarded best in track, Evidence from the Field — Marketing Science Conference 2025
Abstract | SSRN
Storefronts are central to how retailers convert passerby traffic into customers, yet rigorous evidence on the role of a storefront remains limited. Retail models tend to emphasize proximity to consumers as a predictor of retail traffic and sales. We examine the role of storefront visibility, rather than location per se, in driving retail traffic by linking the staggered expansion of bike share networks in New York City and Boston with high-frequency cellphone mobility data on more than 1000 retailers. New bike share stations create a plausibly exogenous increase in passerby traffic. Using exact bike share station locations, we measure line-of-sight observability between stations and storefronts. Our results show that only visible retailers experience significant gains in store visits after the opening of a bike share station. Storefront features like informative signage and legible fonts further amplify these effects. These findings highlight the nature of the marketing communications role of storefronts.
with Samsun Knight
Working paper
Abstract | SSRN
Restaurants attract residents, and residents attract restaurants. This two-sided sorting confounds observational estimates of consumer preferences and may lead standard restaurant entry models to systematically mispredict which areas are best for potential new entrants. We estimate a joint model of restaurant entry and household location choice in the Baltimore metropolitan area, exploiting the 2024 Francis Scott Key bridge collapse and Baltimore County’s four-year zoning cycle as quasi-experiments to identify consumer demand and household sorting preferences. The bridge collapse reveals that consumers are meaningfully sensitive to travel time, with demand decreasing by 3.9% for every 10% increase in travel time, while the zoning event study identifies a nonmonotonic preference for living near restaurants, with households valuing restaurant access at moderate distance but disliking living too close. We then embed these quasi-experimental estimates into a dynamic entry model and show that an endogenous market specification, which anticipates household sorting in response to area amenities, significantly reduces out-of-sample prediction error relative to a standard exogenous dynamics baseline. Counterfactual analysis of transit improvements further shows that the two specifications rank candidate road improvements differently, and project different long-run profitability effects for nearby business locations. Taken together, our results show that “crowded” markets are often not as saturated as they appear, as higher competitive density can make neighborhoods more attractive places to live and expand local demand in the long-term.
Papers in progress
Corporate Consolidation in US Veterinary Markets,
Job Market Paper, draft coming soon
Publications
An experimental investigation into whether choice architecture interventions are considered ethicalwith Daniella Turetski, Renante Rondina, Bing Feng, and Dilip Soman
Scientific Reports, 2023
Abstract | Replication files
Despite their increasing use, choice architecture interventions have faced criticism for being possibly manipulative and unethical. We empirically explore how an intervention’s acceptability differs by the type of intervention used, by the domain, and by the way in which its implementation and benefits are explained. We employ a 5×5×5 factorial design with three fully crossed predictor variables: domain, type of intervention, and explanation. We measure participants’ acceptance of the proposed intervention, perceived threat to autonomy and freedom of choice, and belief that the intervention will be successful. We hypothesized that acceptability of the intervention and perceived threat to autonomy will change as a function of the type of intervention used, the domain in which it is implemented, and the rationale for which its use is presented. We find that acceptability of the intervention, perceived threat to autonomy, and belief that the intervention will be successful differ by the type of intervention used and by the domain in which it is implemented. The rationale for the use of the intervention appears to change acceptability of the intervention depending on the type of intervention that is being used, and the domain in which it is implemented. Exploratory analyses were conducted to investigate differences between specific levels within factors, and interactions between factors. Given the variation in acceptability across the three factors, we believe that the discourse about the ethics of choice architecture should avoid generalizations and should instead be at the level of individual interventions in a specific situation. We conclude with a discussion about areas for future research.