Research

Auctions fail when too few bidders participate. In oil and gas leasing, a common remedy is pre-auction nomination: a tract is offered only if some bidder pays a non-refundable fee and commits to buying at the minimum bid. This paper asks how nomination compares with seller-initiated auctions when markets are thin. I develop a two-stage nomination-auction game in which bidders' values are correlated but costly to learn, and each bidder chooses whether to learn early and nominate or to learn only after someone else nominates. Using novel data from North Dakota oil and gas lease sales, I find nomination raised seller revenue by about $3.05 million (44%) across the 728 tracts in the 2017–2018 estimation sample, relative to a seller who chooses arbitrarily which tracts to offer. Nomination is valuable mainly because it guarantees a bidder when attention is scarce: every nominated tract sells, against about 90 percent without nomination, and the average winning bid rises by about 30 percent through the competition it adds. The gains accrue almost entirely to the seller. Bidders gain mainly by avoiding learning costs on tracts they would not buy anyway: nomination steers their learning toward tracts already screened as worth at least the minimum bid.

We collect and analyze data from a survey of just under 2,000 U.S. citizens to study perceptions about the role of generative Artificial Intelligence (AI) in the workplace, as well as changes in career-relevant decision-making since the initial launch of ChatGPT. We report three main findings. First, most respondents report dedicating more time to learning now compared to before ChatGPT, and this trend is driven by those who use AI and believe that it is a useful tool for learning. Second, while time invested in learning tends to decrease with age, this trend is attenuated among those who believe in AI's potential as a learning tool. Third, workers who believe that AI is a complement to their own effort increase their effort in response to the introduction of AI while workers who believe that AI is a substitute show no such effect. These patterns are robust to the inclusion of a rich set of controls and demonstrate the importance of worker-level tradeoffs in shaping heterogeneous responses to the introduction of AI. They highlight career-long learning as a particularly important dimension of AI capabilities that is especially important in the eyes of many workers.

This paper uses a dynamic structural model to characterize and quantify cost interdependency in mineral rights using sequential auction data to exploit dynamic spatial incentives arising from sequentially realized winning outcomes. I find that bidders overbid in anticipation of future rent from winning close-by auctions. Structural estimates indicate that winning multiple nearby tracts results in cost synergy, with variation in proximity and size. Using the results, I show that bundling auctions improves both efficiency and revenue in a sequential auction setting. A back-of-the-envelope calculation suggests that revenue and efficiency improve if tracts 0.25 miles apart are bundled. Furthermore, the size of a bundle should be capped at 400 acres in total.