Not knowing the competition: evidence and implications for auction design

• DOI: http://doi.org/10.1111/1756-2171.12342
• Published date: 01 September 2020
• Date: 01 September 2020
• Author: Yunmi Kong

RAND Journal of Economics

Vol.51, No. 3, Fall 2020

pp. 840–867

Not knowing the competition: evidence

and implications for auction design

Yunmi Kong∗

In a government auction program wherefirst-price auctions generate significantly higher revenue

than English auctions, I document evidence that bidders are uncertain about the number of auc-

tion entrants. Motivated by additional data evidence, I estimate a structural model of auctions

in which rivals’ participation is stochastic, allowing for bidders’ risk aversion and asymmetry.

Counterfactual simulations reveal that bidders’ uncertainty about the number of entrants, com-

bined with risk aversion, substantially softens the revenue impact of lowcompetition in first-price

auctions. This explains the observed revenue patterns and uncovers an empirically important

reason for sellers to favor first-price auctions over English auctions.

1. Introduction

This article presents empirical evidence that uncertainty about the number of entrants com-

bined with risk aversion, a very plausible feature of real-world auctions, is of first-order impor-

tance for auction design. In a government auction program where first-price auctions generate

significantly higher revenue than English auctions, I document evidence that bidders are uncer-

tain about the number of auction entrants. Motivated by additional data evidence, I estimate a

structural model of auctions in which rivals’ participation is stochastic from the bidder’s point of

view, allowing for bidders’ risk aversion and asymmetry. Counterfactual simulations using the

estimated model reveal that bidders’ uncertainty about the number of entrants, combined with

risk aversion, substantially softens the revenue impact of low competition in first-price auctions

but has no effect in English auctions. This explains the observed revenue difference betweenauc-

tion formats and uncovers an empirically important reason for sellers to favor first-price auctions

∗Rice University; yunmi.kong.01@gmail.com.

This article incorporates content that was previously circulated under the title “Selective Entry in Auctions: Estima-

tion and Evidence.” Kong would like to thank the New Mexico State Land Officefor generous assistance in collecting

and interpreting the data. Special thanks are due to Stephen Wust, Joe Mraz, and Dan Fuqua. Chris Barnhill, Kevin

Hammit, Tracey Noriega, Philip White, and Lindsey Woods also provided valuable industry insights. The author would

like to thank Xiaohong Chen, Jeremy Fox, Emmanuel Guerre, Kei Kawai, Vijay Krishna, Laurent Lamy, John Lazarev,

Alessandro Lizzeri, Xun Tang, anonymousreferees, and par ticipants of the NYU CRATE lunch (2014), NYU Stern Fri-

day seminar (2015), SITE Workshop(2015), World Congress of the Econometric Society (2015 Montreal), NYU alumni

conference (2017), and Midwest Econometrics Group (2017 TAMU)for helpful comments and suggestions. The author

is deeply indebted to Isabelle Perrigne and Quang Vuong for their invaluable guidance and support. All remaining er-

rors are the author’s.Financial support from the NYU Center for Research in Applied and Theoretical Econometrics is

gratefully acknowledged.

840 © 2020, The RAND Corporation.

KONG / 841

over English auctions. To my knowledge, this is the first study to show the empirical importance

of uncertainty about entrants and risk aversion.

I study auctions run by the New Mexico State Land Office (NMSLO) in the Permian Basin

between 2005 and 2014; the Permian Basin is one of the most prolific and economically impor-

tant oil and gas basins in the world.1I document the following data patterns. (i) The first-price

sealed-bid auction generates about 30% higher revenue than English auctions, controlling for

characteristics of the auction item. (ii) Bidders in the sealed-bid auction bid several multiples of

the publicly announced reserve price even when they are the only bidder, whereas single-bidder

English auctions yield exactly the reserve price. Pattern (ii) reveals that bidders face uncertainty

about the number of auction entrants; if a bidder knew for certain he was the only entrant, he

would have bid the reserve price, which is the minimum acceptable bid. Meanwhile, I note that

the majority of bidders are small, local independent operators, which makes it likely that at least

some bidders are risk-averse. Whether bidders are risk-averseis important because we knowfrom

auction theory (McAfee and McMillan (1987) and Matthews (1987)) that, although uncertainty

about entrants is revenue-neutral under risk-neutrality, it breaks revenue equivalence in the same

direction as pattern (i) under reasonable forms of risk aversion. Intuitively, risk-averse bidders

in first-price auctions insure themselves against the possibility of a large number of entrants by

bidding relatively high, even when the expected number of entrants is low. Bidders also exhibit

asymmetry, which, as shown inMaskin and Riley (2000a), can contribute to patter n (i) as well.

To understand the cause of the revenue difference between auction formats, I estimate an

auction model in which bidders are uncertain about the set of entrants and allowed to be asym-

metric and risk-averse. From a bidder’s perspective, each potential competitor enters the auction

with some known probability,resulting in uncertainty about the set of entrants. I allow these entry

probabilities to differ by auction format, so that different entry rates can also be an explanation for

the observed revenue difference. Exploiting the presence of two auction formats, I nonparamet-

rically identify and estimate bidders’ value distributions and utility functions. Estimates indicate

that bidders are moderately risk-averse, at a level similar to what has been measured in previous

studies of risk aversion.

Counterfactual simulations that isolate the effect of each model feature reveal that uncer-

tainty about entrants combined with risk aversion is the main cause of the observed revenue

difference between first-price and English auctions. This finding is robust to alternative speci-

fications of the details surrounding auction entry, such as the number of potential bidders and

whether entry is selective. Uncertainty and risk aversion yield this outcome by boosting revenue

in first-price sealed-bid auctions although not affecting English auctions. The boost is largest

when the number of realized entrants is low, softening the negative effect of low competition.

Empirically, the magnitude of this boost is often similar to that of having one additional entrant.

As such, this is a first-order consideration for sellers choosing between first-price and English

auctions in low competition environments.

The qualitative patterns documented in this study are not particular to the NMSLO auctions;

they are consistent with early intuition in natural resource policy that preceded—and did not have

the benefit of—the auction theory used in this article. Surveying federal and state natural resource

auctions, Mead (1967) remarks anecdotally that even when“a lack of bidder interest [...] results in

one-bidder sales under sealed bid procedures, such sales may yield a price close to a competitive

price,” and where “competition is unreliable, sealed bidding is the more appropriate method as

it introduces a measure of uncertainty about who may appear as a bidder.” The findings are also

important because, as Bulow and Klemperer (1996) have noted, low competition is a critical

concern, and it is common in government auctions; Haile, Hendricks, and Porter (2010), in their

description of U.S. offshore oil and gas lease bidding, remark that “93% of tracts sold attracted

one or two bids” in 1998–2002.

1https://www.britannica.com/place/Permian-Basin

C

The RAND Corporation 2020.