Shunto J. Kobayashi

Ph.D. Candidate in Economics
California Institute of Technology

I am a Ph.D. candidate in Economics at California Institute of Technology with research interests in Empirical Industrial Organization and Econometrics.

My main research agenda examines the interplay of firms’ strategic behaviors with market mechanisms and information structures, especially within digital markets. I have focused on the online advertising industry, which presents a rich landscape for my research due to its intricate market mechanism and reliance on online data for advertising.

I am also interested in analyzing consumer behavior by integrating economic experiments with structural econometric methods to account for deviations from classical economic models.

I’m on the economics job market during the 2023-2024 academic year.


Job-Market Paper

  1. Dynamic Auctions with Budget-Constrained Bidders: Evidence from the Online Advertising Market
    with Miguel Alcobendas (Yahoo)
    When price discovery is necessary for time-sensitive goods, a common practice is to conduct an auction for each item sequentially, but dynamic incentives can lead to behavior distinct from static settings. We provide a novel empirical analysis of a large-scale sequential market that employs auctions to allocate objects to firms with budget constraints, leveraging a unique proprietary dataset of the online advertising market. In this market, because of their short-run budget constraints, participants face a tradeoff between winning auctions immediately or holding out for later opportunities. This dynamic incentive prompts them to adjust their entry rates and bidding strategies accordingly. We develop and estimate a finite-horizon dynamic game between bidders with heterogeneous budgets facing a sequence of simultaneous auctions to quantify this incentive and analyze its implication in competition and auction design. We find that a substantial markdown occurs due to the dynamic incentives arising from budget constraints, and this markdown varies significantly among bidders with different budgets. Using the estimated structural model, we provide a counterfactual simulation comparing the first-price and second-price formats. Unlike the standard environment, we find that dynamics and heterogeneous budgets lead to a significant disparity in the welfare distributions under them. This highlights that even a seemingly simple mechanism choice can have competitive implications in such a dynamic environment.

Working Papers

  1. The Impact of Privacy Protection on Online Advertising Markets
    with Miguel Alcobendas (Yahoo), Ke Shi (Caltech), Matthew Shum (Caltech)
    Accepted for presentation at EC '23: ACM Conference on Economics and Computation
    resubmitted (October 2023)
    Online privacy protection has gained momentum in recent years and spurred both government regulations and private-sector initiatives. A centerpiece of this movement is the removal of third-party cookies, which are widely employed to track online user behavior and implement targeted ads, from web browsers. Using banner ad auction data from Yahoo, we study the effect of a third-party cookie ban on the online advertising market. We first document stylized facts about the value of third-party cookies to advertisers. Adopting a structural approach to recover advertisers' valuations from their bids in these auctions, we simulate a few counterfactual scenarios to quantify the impact of Google's plan to phase out third-party cookies from Chrome, its market-leading browser. Our counterfactual analysis suggests that an outright ban would reduce publisher revenue by 54% and advertiser surplus by 40%. The introduction of alternative tracking technologies under Google's Privacy Sandbox initiative would recoup part of the loss. In either case, we find that big tech firms can leverage their informational advantage over their competitors and gain a larger surplus from the ban.

  2. Robust Estimation of Risk Preferences
    with Aldo Lucia (Caltech, his JMP)
    Economic models have been successful at rationalizing specific empirical findings, but their ability to concurrently explain multiple behaviors remains limited. In this paper, we illustrate this issue by conducting an experiment with 500 participants that studies two classical behaviors inconsistent with Expected Utility: the common ratio effect and preferences for randomization. The lack of generalizability of leading economic models across these two behaviors calls for the development of new empirical strategies to make predictions. Motivated by this observation, we introduce a novel empirical approach that enables us to predict behavior under risk without leaning on specific decision models. We further demonstrate that this method offers more accurate out-of-sample predictions about behaviors under risk, both inside and outside laboratory settings, than leading economic models and machine learning algorithms.

Conference Presentations