Information Provision in Procurement Auctions

• Date: 01 April 2017
• Author: DANIEL GARCIA,JOAQUÍN COLEFF
• DOI: http://doi.org/10.1111/jpet.12217
• Published date: 01 April 2017

INFORMATION PROVISION IN PROCUREMENT AUCTIONS

JOAQU´

IN COLEFF

Universidad del Rosario & UNLP - CEDLAS

DANIEL GARCIA

University of Vienna

Abstract

We study the optimal provision of information in a procurement auc-

tion with horizontally differentiated goods. The buyer has private in-

formation about her preferred location on the product space and has

access to a costless communication device. A seller who pays the en-

try cost may submit a bid comprising a location and a minimum price.

We characterize the optimal information structure and show that the

buyer prefers to attract only two bids. Further, additional sellers are in-

efficient since they reduce total and consumer surplus, gross of entry

costs. We show that the buyer will not find it optimal to send public in-

formation to all sellers. On the other hand, she may profit from setting

a minimum price and that a severe hold-up problem arises if she lacks

commitment to set up the rules of the auction ex ante.

1. Introduction

Ann would like to purchase a unique good. There is a number of sellers with the tech-

nology to produce it, but they are uncertain about Ann’s preferred design. How should

Ann communicate with those potential sellers? How many sellers should she try to at-

tract? In this paper, we attempt to answer these questions in the context of a horizontally

differentiated market with entry costs. More precisely, we analyze the problem of a buyer

whose ideal product is somewhere on a Salop circle, with linear transportation costs. She

can commit to send a number of messages informing about her location to any number

of sellers. Each of these messages may be either private (observed only by one seller)

or public (observed by all sellers). If a seller decides to enter, he has to pay a positive

cost and submit a bid comprised of a location in the circle and a price. We assume that

all potential sellers have the same production cost, which we normalize to zero. The

contract is then awarded through a reverse second price (RSP) auction to the highest

Joaqu´

ın Coleff, Universidad del Rosario, Bogota, Colombia (jcoleff@gmail.com). Daniel Garcia, Uni-

versity of Vienna, Vienna, Austria (daniel.garcia@univie.ac.at).

Daniel Garcia acknowledges the hospitality of Universidad del Rosario and financial help from the

Hardegg Foundation Grant. Both authors would like to thank Carlos Ponce, who eventually declined

a co-authorship but whose insightful comments pervade the paper. We also thank Federico Boffa and

Juanjo Ganuza for their insighful comments on earlier drafts as well as various audiences at EARIE 2013

in Evora, Jornadas de Econom´

ıa Industrial 2013 in Segovia, Universidad del Rosario and VGSE.

Received May 26, 2016; Accepted June 21, 2016.

C2016 Wiley Periodicals, Inc.

Journal of Public Economic Theory, 19 (2), 2017, pp. 426–444.

426

Information Provision in Procurement Auctions 427

surplus-creating bidder who receives the maximum price that would have allowed him

to win.

Ann faces a simple tradeoff. On the one hand, she would like to communicate

with each seller as precisely as possible in order to guarantee that each of them offers

a product she finds attractive. On the other hand, more precise signals induce harsh

competition and severely limit the profits that a given seller can expect to earn. Being

forward-looking, sellers decide not to enter the auction if they do not expect to extract

positive rents. In order to balance these two concerns, Ann finds it optimal to introduce

some noise into the information she communicates to each seller.

We solve Ann’s problem using a mechanism design approach. We first ask which is

the distribution of locations that maximizes expected consumer surplus conditional on

every seller getting an expected profit equal to his entry cost. We show that this prob-

lem is equivalent to maximizing total surplus subject to free entry. The solution requires

independent and identically distributed locations with maximal dispersion. Using these

insights, we characterize the optimal distribution of locations given the number of en-

trants. We also show that the buyer will optimally solicit only two bids, and that every

additional bid reduces consumer surplus net of the entry cost.

We then ask whether the buyer can use the release of information to implement the

optimal distribution of locations in a perfect Bayesian equilibrium (PBE). We show that

the buyer can do so by communicating privately with each seller, using a distribution

of signals akin to the optimal distribution of locations. The buyer never finds it optimal

to supply public information to the sellers. We further show that if the entry cost is not

too high, the buyer can fully implement the optimal distribution (i.e., implement the

optimal distribution of locations in any PBE).1

In our benchmark model, we simply assume that Ann must use a simple auction

to allocate the contract. Nevertheless, it should be noticed that if all sellers are (ex post)

homogeneous, she can attain her second-best payoff by setting a minimum price and com-

mitting to reveal perfectly her location. We show that if sellers are ex post heterogeneous

regarding their production costs, Ann may find it optimal to use a random communi-

cation mechanism together with a minimum price. Further, we show that if Ann cannot

commit to set up the rules of the auction before communication takes place, she faces

a severe hold-up problem. Once sellers have entered the auction she has incentives to

set up a maximum price, limiting the profits of sellers. To encourage entry, she has to

introduce more noise in her communication so that the ex post optimal auction yields

enough rents to entrants.

In our view, the interest of these results for the study of procurement auctions is

threefold. First, anecdotal evidence suggests that communication between buyer and

sellers preceding procurement auctions is often a key determinant of the outcome. In

most scoring auctions, bidders are able to make formal inquiries and enter in infor-

mal discussions with the referees, prior to making their bids. This process enables the

bidders to better gauge the potential value of the project but it also helps them fine-

tune their bid to the preferences of the buyer.2Second, attracting bidders to an auc-

tion is of primary concern to the auctioneer. In particular, it ensures competition and

1We focus on the case where communication arises from signals that are not negatively correlated,

which is the natural result when the sender sends a combination of private and public independent

signals (see Section 5).

2For instance, Design-Build Auctions in Florida allow preselected sellers to meet with the referees

before submitting their bids. Details can be found at www.dot.state.fl.us/construction/designbuild/

DBRules/DBRulesMain.shtm.