2020, Wooten
2020, Wooten
2020, Wooten
Volume 53 Number 3
June 2022
ABSTRACT
An increasingly studied auction format is the asymmetric procurement auction, which
features an advantaged bidder selling a good or service to a buyer. Such asymmetric
setups, although generally believed to be more realistic, are also more complex. We in-
vestigate one such setting where one bidder has a cost advantage. Our central question is
whether bidders who are cost disadvantaged can overcome their inferior cost position. In
a design that mirrors real construction procurement auctions, our laboratory experiment
tests whether two information asymmetries (more precise cost estimates and knowledge
about the cost estimating abilities of a competitor) allow cost-disadvantaged sellers to
better compete. We begin by testing bidder performance in a procurement auction where
the only asymmetry is due to a cost. In this setting, analytical benchmarks are known. We
find—consistent with other studies—that all bidder types submit aggressive bids well
below the Nash equilibrium predictions. Over time, subjects submit less aggressive bids,
moving closer to what theory would predict. We then extend our experiments to include
our novel, multiple asymmetry setup. We find that endowing cost-disadvantaged bidders
with higher cost estimate precision benefits the bidder, as one might expect. Notably,
providing market knowledge about a rival’s ability to estimate costs may not provide
a benefit; in fact, it seems bidders are not able to use this information effectively and
performance suffers. Finally, we show how bidders behave myopically when making
these decisions. These implications raise important questions about how asymmetries
interact in complex auction settings. [Submitted: August 9, 2019. Revised: August 5,
2020. Accepted: September 18, 2020.]
INTRODUCTION
In this paper, we study a particular type of sourcing or supply chain management
problem: first-price sealed-bid (FPSB) procurement auctions where one party has
a cost advantage. In procurement auctions, sellers compete to supply a product
or service to a buyer. Naturally, the ability to deliver at a reduced cost confers
† Corresponding authors.
486
15405915, 2022, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/deci.12498 by National Cheng Kung University, Wiley Online Library on [25/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wooten et al. 487
an advantage to that seller. Our central question revolves around whether that
advantage can be overcome by other bidders.
The majority of the literature—and the theoretical results that follow—makes
the strong assumption of seller symmetry, where all sellers are assumed to be
alike in terms of cost and information (Aloysius, Deck, Hao, & French, 2016).
In practice, however, the assumption of symmetric sellers may not be applicable
because sellers often differ in terms of their cost structure (Laffont, Ossard, &
Vuong, 1995), market experience (De Silva, Dunne, & Kosmopoulou, 2003), ca-
pacity availability (Jofre-Bonet & Pesendorfer, 2003), proximity to buyer (Flam-
bard & Perrigne, 2006), or access to information (Hendricks & Porter, 1988). In
sealed-bid settings, such asymmetries are the source of realized profits: Milgrom
and Weber (1982) suggest that a seller without private information can never earn
a positive payoff and Klemperer (1999) submits that any seller surplus is due to the
possession of private information, although incremental. Although prior research
has assumed an advantaged seller has a cost or information advantage, little or no
research has considered the case where one seller is simultaneously cost advan-
taged yet information disadvantaged competing against a seller who is simultane-
ously cost disadvantaged yet information advantaged.
Consider the scenario where a small, local construction company is bidding
against a large, national construction firm for a contracted project. In this case, the
national firm might benefit from lower costs due to economies of scale. It is also
reasonable to envision that the local company—due to its experience in the area,
reputation with subcontractors, and proximity to the project—might enjoy better
information than its national rival. The local company can invest time and resources
(e.g., visiting the project site, exploiting local knowledge, and working alongside
local subcontractors) in developing more accurate cost estimates. Further, given
that many online procurement auctions post all bids after awarding a contract, the
local company would have the ability to collect many data points about the na-
tional firm’s bid history. With some work, the local company could generate better
market knowledge (about the bidding patterns and cost estimating capabilities of
the national firm) than might be possible going the other direction (for the national
firm knowing the local company’s capabilities). In fact, this scenario played out on
our own campus when the business school took bids to design and construct a new
LEED platinum building and a number of smaller, local businesses lost opportuni-
ties to bigger, out-of-town contractors. A similar dynamic has been observed in the
construction industry for a variety of bidder types (e.g., domestic vs. international
and new vs. incumbent).
We assess the impact of better information—in the form of more precise
cost estimates and knowledge about the cost estimating abilities of a competitor—
when a seller is cost disadvantaged. Can such information advantages allow a cost-
disadvantaged seller to successfully compete against their cost-advantaged rival?
We answer this through a series of controlled lab experiments designed to examine
bidder behavior.
First, we test the predictions from a normative model to see if cost-
disadvantaged bidders behave as theory would predict. Next, we endow the cost-
disadvantaged bidder with better information—first as a cost precision advantage
and also as a competitor knowledge advantage—and analyze the impact of those
15405915, 2022, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/deci.12498 by National Cheng Kung University, Wiley Online Library on [25/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
488 Overcoming Cost Disadvantages in Procurement Auctions
information—and how these two advantages work with or against each other in this
setting.
We find participants in the lab bid more aggressively than normative theory
and our numerical results would predict (a common finding). Bidders do move
closer to normative levels with experience, similar to some results where “super
experienced” bidders perform much better—but still below Nash equilibrium profit
levels—than their naïve selves (Kagel & Richard, 2001). Additionally, being cost
disadvantaged hurts one’s ability to make money, as expected. When assessing
whether more information helps cost-disadvantaged bidders overcome that hurdle,
we find that more precise cost estimates do help (because lower variability leads
to safer, more aggressive bids) but knowing about the cost estimating ability of
your competitor does not help. Having more complete market information results
in more aggressive bids, but in this case, they are not safer bids—knowing more
information about the market has not improved the distribution from which costs
are derived. Finally, we examine the behaviors surrounding these phenomena. We
find that the situation matters and bidders react myopically to the previous outcome
(in line with learning direction theory) and the prior opponent. Overall, we find
information and precision asymmetries to have an opposite effect than found in
prior literature. We reconcile this anomaly and conjecture that this is due to two
influences—the presence of two asymmetries concurrently (a novel setup) and the
negative impact of information transparency (an unrealistic artifact in some auction
research).
Cost Asymmetry
There is a growing body of literature on PV and CV procurement auctions where
sellers are asymmetric. In the asymmetric PV literature, a seller is said to be
cost advantaged if they enjoy a lower cost of production—where the distribution
of their valuation signal stochastically dominates the valuation signals of other
sellers (Maskin & Riley, 2000). Our experiment looks like a shift asymmetry (e.g.,
Maskin & Riley, 2000; where both ends of a common support are adjusted by a
fixed about amount) instead of a stretch asymmetry (e.g., Güth, Ivanova-Stenzel,
& Wolfstetter, 2005; where only one end is adjusted). The typical assumption is
that valuations are drawn from a uniform distribution, which allows for many of
the theoretical results reported. This uniform assumption has been criticized as
15405915, 2022, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/deci.12498 by National Cheng Kung University, Wiley Online Library on [25/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
490 Overcoming Cost Disadvantages in Procurement Auctions
Information Asymmetry
From the literature on asymmetric CV auctions, asymmetry is introduced when the
information available to one seller is superior to the information available to the
other. The result is that the advantaged seller, due to having superior information,
has less uncertainty in the value of the item resulting in a more accurate estimate.
Myerson (1981) refers to these differences in estimates as quality uncertainties.
Superior information may come in the form of a more precise estimate of the CV
(Hausch, 1987; Kagel & Levin, 1999; Laskowski & Slonin, 1999; Povel & Singh,
2006). When the information advantage is due to a better estimate of cost, as the
degree of advantage decreases between the two types of bidders, both types will
submit bids that are increasingly aggressive, with the information-advantaged
bidder submitting bids that are more aggressive than the bids submitted by the
information-disadvantaged bidder (Hausch, 1987; Grosskopf, Rentschler, &
Sarin, 2018). Another example of superior information is when an advantaged
seller possesses information that is not available to a disadvantaged seller (Gilley
& Karels, 1981; Milgrom & Weber, 1982; Englebrecht-Wiggans, Hendricks, &
Porter, 1988; Hendricks & Porter, 1988; Hendricks, Porter, & Wilson, 1994; Kagel
& Levin, 1999; Campbell & Levin, 2000; DeSilva et al., 2003; Banerjee, 2005;
Kim, 2008). A general result from this literature is that any bidder profits are due
to the amount of private information they possess, and a bidder with no private
information will make zero profits (Milgrom, 1989; Kagel & Levin, 1999). When
the superior information is due to the possession of private information, the seller
15405915, 2022, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/deci.12498 by National Cheng Kung University, Wiley Online Library on [25/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wooten et al. 491
with inferior information will bid less aggressively if they knowingly compete
against a seller with superior information. In response, the seller with superior
information will bid less aggressively due to the conservative bidding strategy of
the seller with inferior information (Milgrom & Weber, 1982; Grosskopf et al.,
2018). In a study of drainage field auctions, Hendricks and Porter (1988) find
that information-advantaged bidders bid aggressively (closer to their estimated
values), often submitting bids below their expected value. DeSilva et al. (2003)
and Hendricks et al. (1994) find that information-disadvantaged sellers also tend
to submit aggressive bids and thus will realize lower profits when they do win. In
a lab experiment on multi-seller procurement auctions, Haruvy and Katok (2013)
similarly find that not having visibility into competitors’ bids or product quality
both increase sellers’ aggressive bidding and reduce profits.
Contrary to the results from the PV literature where the cost-disadvantaged
bidder resorts to aggressive bidding, in a CV auction, an information-advantaged
seller (who knows they are competing against an information-disadvantaged seller)
bids more aggressively than if the information were equal between the sellers
(Englebrecht-Wiggans et al., 1983; Hendricks & Porter, 1988; DeSilva et al.,
2003). In our study, we combine these two asymmetries. As such, we expect that
the bidder who is cost disadvantaged and information advantaged will submit much
more aggressive bids than their competitor. However, we do not know the impact
on profitability, because theory suggests the profitability moves in opposite direc-
tions for these effects. Our experimental design will allow us to test the impact of
these two asymmetries.
Buyer Cost
Finally, the behavior of the sellers in procurement auctions has implications on
the auction’s effectiveness. When the sellers are asymmetric due to a cost advan-
tage, buyer procurement costs tend to increase as the cost disadvantage increases
(Rothkopf et al., 2003). Thus, the ability of the buyer to extract surplus from the
auction is limited and the auction that minimizes buyer procurement costs is one
that discriminates in favor of the higher cost seller (Myerson, 1981; McAfee &
McMillan, 1987). In many auction settings, however, it may not be feasible for a
buyer to discriminate in favor of one type of seller over another. Research has also
shown that buyer procurement costs are impacted if the sellers are asymmetric due
to different information.
In a CV auction, reducing seller uncertainty over the CV will reduce buyer
procurement costs (Gilley & Karels, 1981; Milgrom & Weber, 1982; Hausch,
1987). In a PV auction, Rothkopf et al. (2003) also show that buyer procure-
ment costs decrease as uncertainty of seller PVs decreases. Kagel et al. (1987)
show that in a PV auction where one seller is better informed about their own
value compared to another seller, buyer procurement costs are higher than if all
sellers were equally informed. In addition, releasing information regarding a ri-
val’s value decreases procurement cost. In Laskowski and Slonin (1999) and Kagel
and Levin (2011), procurement costs increase if the quality of information for the
information-advantaged seller improved relative to the less informed seller such
that the level of asymmetry between the sellers increased.
15405915, 2022, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/deci.12498 by National Cheng Kung University, Wiley Online Library on [25/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
492 Overcoming Cost Disadvantages in Procurement Auctions
Theoretical Predictions
Our asymmetric settings feature one bidder competing against a second, cost-
advantaged bidder in a FPSB procurement auction. To gain insight into how bidders
behave in such auctions, we start with a known theoretical result. Rothkopf et al.
(2003) assess whether external subsidies are sufficient to overcome cost disadvan-
tages in asymmetric procurement auctions. Instead of examining subsidies, we re-
purpose the base case scenarios to answer whether information can overcome cost
disadvantages. The model assumes that a cost-advantaged seller competes against
a cost-disadvantaged seller and that each seller is uncertain about their private cost.
We replicate their multiplicative strategy equilibrium for the two-seller case. Sell-
ers’ uncertainty over costs, u, is the standard deviation of cost estimation errors
that are randomly drawn from a two parameter Weibull distribution:
2 1 −1/2
u= 1+ − 2 1 + , (1)
m m
mX (d) [1 + X (d)]1/m
Pa = (2)
[mX (d) − 1]
and
1 Discussions with one of the authors of the original model (Rothkopf et al., 2003) revealed that they had
worked on closed form solutions that included more asymmetries but were unable to develop a solution.
In fact, they recommend a numerical solution even for simpler deviations such as increasing the number of
sellers beyond two.
15405915, 2022, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/deci.12498 by National Cheng Kung University, Wiley Online Library on [25/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
494 Overcoming Cost Disadvantages in Procurement Auctions
permits the high-cost bidder to mark up their estimate more aggressively, but their
profit remains surprisingly flat—this simply punishes the cost-advantaged bidder
and transfers the surplus to the auction buyer. Although increased precision may
not improve the profitability of disadvantaged bidders all that much, it should level
the playing field between the two sellers in that both can expect equal profits when
bidding.
We take our full set of equilibrium strategies (based on the theoretical model
and our numerical predictions) and move to the lab in order to generate experimen-
tal results in similar scenarios. In this way, we address our basic research questions
and benchmark our results against equilibrium predictions.
EXPERIMENTAL DESIGN
We conduct a series of experiments to explore our three questions about cost asym-
metric procurement auctions: (1) Do sellers bid as theory would predict? (2) Can
cost-disadvantaged sellers better compete through superior information? (3) How
does bidder behavior change under information asymmetries? In our experiments,
each of two bidders competes as a road construction contractor attempting to win
contracts in a series of 25 first-price sealed-bid auctions. We randomly match bid-
ders each round from each of a session’s two treatment groups. Bidders remain in
the same treatment group (and bid against the opposite group) for the entire ses-
sion. The treatments consider three factors—cost advantage, estimate precision,
and understanding (of competitor estimating ability)—and are built to mirror the
setup from the multiplicative strategy model and simulations in THEORETICAL
PREDICTIONS section.
set of 3,960 projects from 2003 to 2015 and is fairly uniform.2 After drawing Ca ,
the cost for the cost-disadvantaged seller was determined by multiplying Ca by the
chosen cost disadvantage ratio d. Subjects submitted bids without knowing Ca or
Cd . Instead, each participant received a private signal, ca or cd , which represented
their estimated cost to complete the project. After receiving the estimate, each par-
ticipant submitted a bid. The winner was determined using the FPSB format (low
bid wins) and the results were revealed. Winning bidders observed both submitted
bids, their actual project cost, their profit (or loss), and the change to their operating
cash balance. Losing bidders observed both submitted bids and their (unchanged)
cash balance. The actual cost to complete the project and the associated profit were
known only by the winning bidder and were not common knowledge. See Online
Appendices A and B for example screen samples and instructions.
All experiments were run using the z-Tree software platform (Fischbacher,
2007). Each session lasted 75 minutes on average and contained an even number
of participants who were randomly split into the two seller types. In total, 220
participants went through the experiment. Each bidder’s opponent was randomly
assigned from the other seller type for each of the 25 rounds. Thus, participants
did not face the same competitor every round (and knew this), which prevented
the development of any idiosyncratic patterns between sellers. All sellers were
provided with a calculator within the platform and scratch paper to perform any
calculations necessary. The experiment rounds proceeded in lockstep and results
were provided only after all participants had submitted bids.
All participants were paid $10 for participating, with additional prizes of $10,
$20, and $30 awarded to the top three performers in each of a session’s two groups.
As an added enticement, all participants were allowed to complete the experiment
in lieu of one of their class’s online assignments. No subject participated in more
than one session. Prior to the actual experiment, we ran a pilot test with several
graduate students to insure our experiment protocol worked as planned.
Treatments
In total, we run 10 different treatment combinations (Table 1) to investigate the
impact of cost and information advantages in procurement auctions. We con-
struct the treatments in tiers to address increasingly complex questions. First,
we test the predictions from the normative model (Rothkopf et al., 2003) to see
if cost-disadvantaged bidders behave as predicted. Next, we endow those cost-
disadvantaged bidders with better information in the form of a more precise cost
estimate and test our numerical results. Finally, we endow the cost-disadvantaged
bidders with a second information advantage in the form of better market under-
standing by restricting the market understanding of the cost-advantaged bidder.
Our three treatment factors are represented by the cost disadvantage ratio (d), cost
estimate precision as a coefficient of variation (CV), and market understanding of
the cost estimate precisions (CV info). Details on each of the treatment factors are
discussed below.
2 This mirrors the project cost methodology of Wooten et al. (2020) and—as suggested by Rothkopf et al.
(2003)—the bidding strategy should be independent of the magnitude of the project cost.
15405915, 2022, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/deci.12498 by National Cheng Kung University, Wiley Online Library on [25/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
496 Overcoming Cost Disadvantages in Procurement Auctions
Cost Disadvantage
Cost disadvantage (d) represents the difference in the cost of production between
the high- and low-cost seller. We mirror Rothkopf et al. (2003) exactly, with the
disadvantage given as a ratio. The values depicted in the original normative model
include d = 1.05, 1.1, 1.5, and 2.0—corresponding to cost disadvantages of 5%,
10%, 50%, and 100%. We use two of these (1.05 and 1.1) in our experiments and
exclude the others. Not only are the two highest values generally unrealistic, but
there is little chance a seller saddled with such high costs could compete. In our nu-
merical simulation, even a cost disadvantage of 10% proved too large to overcome
with just improved precision. Predictions from the normative model also bear this
out (Rothkopf et al., 2003), so we test only the two smaller levels.
model we benchmark against includes full visibility by the sellers into the auction
conditions. Our second information lever to overcome a cost disadvantage comes
in the form of better market understanding. Market understanding is defined by
which cost estimate precisions are known—either own (the seller knows the CV
of their own estimate but not their competitor’s CV) or both (the seller knows
both CVs). Because our benchmark starts with full market understanding for each
seller, we reduce the market understanding of the cost-advantaged seller in some
scenarios in order to give the cost-disadvantaged seller help. We then examine the
effect of this benefit. All of our scenarios are shown above in Table 1.
4 Even though we remove a handful of irrational bidders and outlier bids to increase validity of the dataset,
the results are robust to leaving them in—we obtain nearly identical results without any exclusions.
15405915, 2022, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/deci.12498 by National Cheng Kung University, Wiley Online Library on [25/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wooten et al. 499
Cost-disadvantaged Bidders
Our first research question seeks to confirm theoretical predictions for FPSB pro-
curement auctions featuring two bidders with different cost structures. Will bidders
behave as theory would expect, with cost-advantaged sellers earning more profit
thanks to higher markup percentages than their disadvantaged rivals? We begin by
presenting the summarized results from the experiments and overall mean com-
parisons to show how markups differ for our bidders.
Table 3 shows % Markup and % Profit for each scenario. Also included are
the normative model and numerical simulation predictions for four sets of parame-
ters where the only asymmetries present are cost and precision. These predictions
are placed directly above the experiment scenario to which they correspond (sce-
narios 1–4). We did not calculate a numerical solution for the treatments focusing
on the impact of market understanding, as such scenarios would require a different
model and assumptions.
The immediate takeaway from our four benchmark scenarios is that lab par-
ticipants bid significantly less (as indicated by percent markup) than theory would
predict.5 This is not unexpected in the lab setting.
Despite the lower magnitudes, the spread between the markups and relative
comparisons across treatments are directionally correct and related to theoreti-
cal predictions, with disadvantaged bidders that seem more aggressive with their
markups than the advantaged bidders. The difference in bidders (e.g., 8.2% vs.
12.0%) is not statistically significant when evaluated with a Wilcoxon rank sum
test. However, there are more factors that need to be accounted for. We know that
every round is not conducted in isolation. Participants bid in 20 consecutive rounds
(after the training rounds), so learning and behavioral effects may exist.
Figure 3 shows the average markup by round for our benchmark scenarios.
First, it is easy to again see how bidders in the lab are more aggressive than theory
would predict, with actual markups (dashed lines) below the solid lines. Second,
bidders appear to become less aggressive over time, with markup percentages that
move closer to the normative benchmarks as rounds progress. One way to track
that move is by fitting a linear trendline to the markup averages by round. Those
slopes are as follows:
On average, bidders become less aggressive with their markups over time.
This resembles a main result from an analysis of Utah construction procurement
5As outlined in Aloysius et al. (2016), direct comparisons between measures that do not depend on ex
post cost realizations are straightforward. Because markups do not depend on actual cost, we start with
comparison t-tests.
Table 3: Summary of the observed experiment outcomes (and theoretical predictions).
500
Cost
Note: Wilcoxon rank sum test shown comparing participant-level averages between seller types (D vs. A).
Brackets show t-test significance for averages versus the theoretical predictions (for D and A).
Theory 1 predictions from normative model (and simulation); Theory 2–4 from numerical simulation.
Significance levels:
∗
<.10,
∗∗
<.05, and
∗∗∗
<.01.
Overcoming Cost Disadvantages in Procurement Auctions
15405915, 2022, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/deci.12498 by National Cheng Kung University, Wiley Online Library on [25/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
15405915, 2022, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/deci.12498 by National Cheng Kung University, Wiley Online Library on [25/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wooten et al. 501
auctions in which bidders with more experience bid less aggressively and with less
dispersion than inexperienced bidders (Li & Philips, 2012). From here on out, we
will control for Bid Round to capture this learning effect with experience. We will
also include Cash Balance as another situational control to account for how much
money a participant has won or lost at any point in time, in order to account for
both skill level and the one part of the experimental condition that changes round-
to-round.
In conclusion, comparing the lab results with the normative model predic-
tions highlights that bidders in the lab are more aggressive with their bids (Table 3
and Figure 3). We also observe that bidders learn over time, with average markups
that move closer to predicted values with experience (Figure 3). These results raise
several questions: What behaviors lead to these outcomes and what are the effect
of these markups on profit? These questions are answered in our next two sections.
between- and within-subject variances that relate % Profit for each individual bid
to our experimental conditions. If information advantages help cost-disadvantaged
bidders earn more profit, we should see positive coefficients on Precision and Mar-
ket Understanding. Only precision improvements turn out to be positive. Results
are presented in Table 4.
We start by estimating a baseline model (4–1) with just the cost asymme-
try and control variables. Across all sessions, we observe that the profit (per bid)
decreases with a cost disadvantage (of either 5% or 10%)—as expected—and in-
creases with both the Bid Round and Cash Balance.
Next, we include our treatment effects for precision and market under-
standing (4–2) as the simplest measure of whether additional information helps
cost disadvantaged bidders. The reference category (intercept) is defined to be a
cost-advantaged bidder with low precision (10% CV) and no market understand-
ing. The coefficients on High Precision (1.25, p = .004) and Perfect Precision
15405915, 2022, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/deci.12498 by National Cheng Kung University, Wiley Online Library on [25/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wooten et al. 503
(1.52, p = .001) are positive and significant, whereas the one on Market Under-
standing (–2.03, p = .000) is negative and significant. This suggests that having
better estimating ability leads to higher profits but that knowing your competitor’s
estimating precision (in addition to your own) is detrimental. The coefficients on
cost in this simple model are not significant. However, as dyadic experiments, the
bidding situation matters: to control for an opponent’s knowledge, we add addi-
tional treatment controls and check the robustness of our results.
When controlling for both bidders’ treatments (4–3), a consistent result
emerges. We again observe that being at a cost disadvantage is bad for profit
(–3.31, p = .000; –3.71, p = .000), that having better cost estimating ability in-
creases profit (1.26, p = .002; 1.54, p = .000), and that knowing your competi-
tion’s estimating precision lowers profit (–1.11, p = .007). The magnitude of the
coefficients gives the direct impact on profit margin, so improving cost estimating
precision does not at first appear to be enough to overcome either level of cost
disadvantage. However, the bidding situation must be accounted for. Doing that,
improved precision can lead to higher profits.6 That supports our theoretical find-
ings (Figure 1), where a CV smaller than 3% resulted in higher expected profits
for the disadvantaged bidder, and our average observed profit outcomes (Table 3).
Also, it seems counterintuitive that possessing more information about a compet-
ing bidder would reduce performance, but that result is similar to recent findings in
CV procurement auctions (Wooten at al., 2020). Thus, although more precise cost
estimates may help offset a cost disadvantage, knowledge about the cost estimating
abilities of a competitor may not. From a practical standpoint, a cost-disadvantaged
firm can become more competitive by improving their cost estimation process rel-
ative to their competitors. Time spent learning competitor estimation abilities may
not be worth the effort.
Bidder Behavior
We demonstrated that profit (our ultimate outcome) is impacted by our treatments
in the prior section. Here, we look at how bidder decisions (in the form of markups)
are affected. First, we examine how individuals modify their strategies depending
on their bidding situation. Second, we examine departures from expected bidding
strategies and attempt to explain why bidders might be deviating from theoretical
predictions.
Wooten et al. (2020) highlight that a bidder’s previous auction result in-
fluences their next bid. We test for the same effect in our procurement auction,
and see similar reactions (Figure 4). Bidders who profit in the previous round in-
crease their next bid, bidders who lose money in the previous round (winner’s
curse) increase their next bid even more, and those who do not win in the previous
round are generally more aggressive, marking up their estimate less in the next
round.
6As an example, if a cost-advantaged bidder (10% CV, own info) were bidding against a 5% cost-
disadvantaged bidder with perfect precision (0% CV, own info), then the coefficient impacts to profit margin
would be –3.35 (or –1.86 – 1.49) for the cost-advantaged bidder and -1.77 (or –3.31 + 1.54) for the cost-
disadvantaged bidder.
15405915, 2022, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/deci.12498 by National Cheng Kung University, Wiley Online Library on [25/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
504 Overcoming Cost Disadvantages in Procurement Auctions
however, virtually none of the bids are less than the estimated cost (Table 6),
because bidders recognize negative markups hurt profit in this case. Those two
adjustments lead to a contraction of the markup range as precision increases and
fewer instances of winner’s curse.
Market understanding (–3.20, p = .010) also makes bidders more aggressive
with their markups. In a CV lab setting, Wooten et al. (2020) show that know-
ing competitor information can lead to information overload and more aggres-
sive bids. This appears to be a similar to our PV auctions with uncertainty. In
effect, when cost-advantaged sellers know about the bidding precision of their
cost-disadvantaged rivals, they reduce their markup—and reduce their profits in
lockstep. Those bidders are better off not knowing about the cost estimating capa-
bilities of their opponents.
15405915, 2022, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/deci.12498 by National Cheng Kung University, Wiley Online Library on [25/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
506 Overcoming Cost Disadvantages in Procurement Auctions
0 0.8 0.9 – –
5 13.6 6.9 – –
10 20.5 9.8 18.2 3.6
two correspond to implicit measures that a bidder might intuit; the last measure is
more salient and correlated with our measure of interest. In all cases, a positive co-
efficient indicates marking up the next bid more when the opponent’s last bid was
higher. We observe positive, significant coefficients on both Prior Partner Bid and
Bid Round, indicating that bidders are learning over time and from their opponent’s
previous action. The derived Nash equilibrium markups (section THEORETICAL
PREDICTIONS) are conditional on your opponent bidding optimally. Obviously,
if they are bidding below Nash, your response will change. We capture that dy-
namic here. Our results are robust to various specifications of previous opponent
markups as well as excluding outliers, with mostly similar results throughout.
Table 8: Average procurement cost for buyer (as % above actual cost for
seller A).
Scenario 2–4 Scenario 5–7 Scenario 8–10
d = 1.05 d = 1.05 d = 1.1
CVd % Info = both Info = Seller D Info = Seller D
their disadvantage (essentially, better CVs only help the buyer while they are
equalizing bidder profits). Unfortunately, the reality seems to be a bit more
complicated than that, and no clear trends are apparent in our results. This is a
limitation of the study, as we designed it to focus on the bidders. We leave the
investigation of this phenomenon on the buyer’s side to future research.
Practically, it seems that buyers should seek to reduce cost asymmetries (e.g.,
sharing best practices and disseminating preferred subcontractors) and make mar-
ket information available (i.e., allow firms to understand each other’s estimation
precision). Both of those reduce procurement costs. Additionally, if buyers are
dealing with experienced bidders or those that display bidding tendencies that ap-
proach the theoretical predictions, then improving the precision of the disadvan-
taged bidder’s cost estimation capabilities may also help.
Summary
In real-world procurement auctions, buyers recognize that auctions can be asym-
metric. There are plenty of examples where buyers have attempted to help disad-
vantaged bidders (often for their own benefit). Subsidies are one approach to that
problem; set-asides are another. Many governments maintain set-aside thresholds
to award auctions to small businesses—in the United States, most federal agencies
try and spend at least 23% of their procurement budget on small firms; in Japan,
that target was 50.1% for small and medium firms in 2007 (Nakabayashi, 2013).
Such sizeable percentages indicate the effort being spent on methods to reduce pro-
curement auction asymmetries. Here, we present some additional tools that can be
used.
As the first investigation of this kind, there are some limitations to the study.
First, dealing with multiple asymmetries for which normative models cannot be
solved led us to benchmark our theoretical results for those cases through numer-
ical solutions. Second, as a practical limitation, we could only test so many levels
of the treatment variables in the initial study. Subsequent studies could focus on
particular focal areas (e.g., more granular precision intervals) in order to derive
additional insight.
15405915, 2022, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/deci.12498 by National Cheng Kung University, Wiley Online Library on [25/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
510 Overcoming Cost Disadvantages in Procurement Auctions
SUPPORTING INFORMATION
REFERENCES
Aloysius, J., Deck, C., Hao, L., & French, R. (2016). An experimental investi-
gation of procurement auctions with asymmetric sellers. Productions and
Operations Management, 25(10), 1763–1777.
Banerjee, P. (2005). Common value auctions with asymmetric bidder information.
Economics Letters, 88, 47–53.
Campbell, C., & Levin, D. (2000). Can the seller benefit from an insider in common
value auctions? Journal of Economic Theory, 91, 106–120.
Capen, E., Clapp, R., & Campbell, W. (1971). Competitive bidding in high risk
situations. Journal of Petroleum Technology, 23, 641–653.
De Silva, D. G., Dunne, T., & Kosmopoulou, G. (2003). An empirical analysis
of entrant and incumbent bidding in road construction auctions. Journal of
Industrial Economics, 51, 295–316.
Dyer, D., & Kagel, J. H. (1996). Bidding in common value auctions: How the com-
mercial construction industry corrects for the winner’s curse. Management
Science, 42(10), 1463–1475.
Englebrecht-Wiggans, R., Milgrom, P. R., & Weber, R. J. (1983). Competitive bid-
ding and proprietary information. Journal of Mathematical Economics, 11,
161–169.
Estache, A., & Iimi, A. (2010). Bidder asymmetry in infrastructure procurement:
Are there any fringe bidders? Review of Industrial Organization, 36(2), 163–
187.
Fischbacher, U. (2007). z-Tree: Zurich toolbox for ready-made economic experi-
ments. Experimental Economics, 10, 171–178.
Flambard, V., & Perrigne, I. (2006). Asymmetry in procurement auctions: Evi-
dence in snow removal contracts. Economic Journal, 116, 1014–1036.
15405915, 2022, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/deci.12498 by National Cheng Kung University, Wiley Online Library on [25/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wooten et al. 511
Laffont, J. J., Ossard, H., & Vuong, Q. (1995). Econometrics of first price auctions.
Econometrica, 63(4), 953–980.
Landsberger, M., Rubenstein, J., Wolfstetter, E., & Zamir, S. (2001). First-price
auctions when the ranking of valuations is common knowledge. Review of
Economic Design, 6, 461–480.
Laskowski, M. C., & Slonin, R. L. (1999). An asymptotic solution for sealed
bid common-value auctions with bidders having asymmetric information.
Games of Economic Behavior, 28, 238–255.
Li, S., & Philips, P. (2012). Construction procurement auctions: Do entrant bid-
ders employ more aggressive strategies than incumbent bidders? Review of
Industrial Organization, 30(4), 191–205.
Maskin, E., & Riley, J. (2000). Asymmetric auctions. Review of Economic Studies,
67, 413–438.
McAfee, P., & McMillan, J. (1987). Auctions and bidding. Journal of Economic
Literature, 25(2), 699–738.
McAfee, R. P., & McMillan, J. (1989). Government procurement and international
trade. Journal of International Trade, 26, 291–308.
Milgrom, P. (1989). Auctions and bidding: A primer. Journal of Economic Per-
spectives, 3(3), 3–22.
Milgrom, P., & Weber, R. (1982). The value of information in a sealed-bid auction.
Journal of Mathematical Economics, 10, 105–114.
Myerson, R. B. (1981). Optimal auction design. Mathematics of Operations Re-
search, 6, 58–73.
Nakabayashi, J. (2013). Small business set-asides in procurement auctions: An
empirical analysis. Journal of Public Economics, 100, 28–44.
Pezanis-Christou, P. (2002). On the impact of low-balling: Experimental results in
asymmetric auctions. International Journal of Game Theory, 31, 69–89.
Povel, P., & Singh, R. (2006). Takeover contests with asymmetric bidders. Review
of Financial Studies, 19, 1399–1431.
Rothkopf, M. H. (1969). A model of rational competitive bidding. Management
Science, 15(7), 362–373.
Rothkopf, M. H., Harstad, R. M., & Fu, Y. (2003). Is subsidizing inefficient bidders
actually costly? Management Science, 49(1), 71–84.
Saini, V., & Suter, J. F. (2015). Capacity constraints and information revelation in
procurement auctions. Economic Inquiry, 53(2), 1236–1258.
Selten, R., Abbink, K., & Cox, R. (2005). Learning direction theory and the win-
ner’s curse. Experimental Economics, 8(1), 5–20.
Seydel, J. (2003). Evaluating and comparing bidding optimization effectiveness.
Journal of Construction Engineering and Management, 129(3), 285–292.
Wooten, J., Donohue, J., Fry, T., & Whitcomb, K. (2020). To thine own self be
true: Asymmetric information in procurement auctions. Production and Op-
erations Management, 29(7), 1679–1701.
15405915, 2022, 3, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/deci.12498 by National Cheng Kung University, Wiley Online Library on [25/04/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Wooten et al. 513
Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.
Alternative Proxies: