My research is in empirical industrial organization, primarily focusing on auctions and online markets. I’m interested in what bidding data can tell us about the things we can’t see directly, from bidders’ costs in procurement auctions to seller misconduct in online auctions. Much of my work brings new kinds of data and tools from machine learning and natural-language processing into auction econometrics.

Working Papers

Dimensionality Reduction for Structural Auction Estimation: A Johnson–Lindenstrauss Approach

Public tenders come with pages of descriptive text that structural auction estimation usually ignores. I convert tender text into sentence embeddings and use them as covariates in Guerre–Perrigne–Vuong cost recovery. To get past the curse of dimensionality, I compress the embeddings with an ensemble of random projections rather than PCA. On procurement data from British Columbia and Québec, this improves bid-prediction accuracy over a text-free baseline by eight to fourteen percentage points more than PCA does.

“Link-Tax” Policy and News Media Market Concentration

Using monthly web traffic for 425 Australian news outlets, I study the distributional effects of Australia’s 2021 News Media Bargaining Code. Outlet-level concentration spiked temporarily, but concentration at the parent-company level rose and stayed up. Outlets whose parents secured platform agreements received roughly 82 per cent more organic traffic than non-agreement outlets.

Detecting Shill Bidding in eBay Auctions: Evidence from Collectible Coins

I construct six price-free behavioural metrics from eBay bid histories for certified graded coins and use hierarchical clustering to flag auctions consistent with shill bidding. Within-seller comparisons show that flagged auctions sell for about 22.8 percentage points of book value more than the same sellers’ other auctions.