15
Rev Ind Organ (2012) 40:191–205 DOI 10.1007/s11151-011-9318-4 Construction Procurement Auctions: Do Entrant Bidders Employ More Aggressive Strategies than Incumbent Bidders? Sheng Li · Peter Philips Published online: 4 August 2011 © Springer Science+Business Media, LLC. 2011 Abstract Asymmetric auction theory predicts that weak bidders will bid more aggressively when facing strong bidders, while strong bidders will bid less aggres- sively when facing weak bidders. This paper finds empirical evidence to support this hypothesis regarding the behavior of weak bidders in construction auctions. Exam- ining a comprehensive data set of more than 7,500 Utah construction procurement auctions, we find that entrants bid more aggressively than do incumbents for most subcontractor types. Reflecting their inexperience and uncertainty regarding the true cost of projects, entrants’ bids are found to be more widely dispersed around the central tendency of bids. Keywords Aggressive · Bidder asymmetry · Dispersion · Entrants · Incumbents · Procurement auctions JEL Classification D44 · D82 · L74 1 Introduction Subcontracting is a common business practice in procurement markets. In the construc- tion industry, two interesting phenomena are observed: first, there is greater variance S. Li (B ) Chinese Academy of Finance and Development, Central University of Finance and Economics, 39 South College Road, Beijing 100081, People’s Republic of China e-mail: [email protected] P. Philips Department of Economics, University of Utah, 1645 Campus Center Dr. Room 308, Salt Lake City, UT 84112-9300, USA e-mail: [email protected] 123

Construction Procurement Auctions: Do Entrant Bidders Employ More Aggressive Strategies than Incumbent Bidders?

Embed Size (px)

Citation preview

Rev Ind Organ (2012) 40:191–205DOI 10.1007/s11151-011-9318-4

Construction Procurement Auctions: Do EntrantBidders Employ More Aggressive Strategiesthan Incumbent Bidders?

Sheng Li · Peter Philips

Published online: 4 August 2011© Springer Science+Business Media, LLC. 2011

Abstract Asymmetric auction theory predicts that weak bidders will bid moreaggressively when facing strong bidders, while strong bidders will bid less aggres-sively when facing weak bidders. This paper finds empirical evidence to support thishypothesis regarding the behavior of weak bidders in construction auctions. Exam-ining a comprehensive data set of more than 7,500 Utah construction procurementauctions, we find that entrants bid more aggressively than do incumbents for mostsubcontractor types. Reflecting their inexperience and uncertainty regarding the truecost of projects, entrants’ bids are found to be more widely dispersed around the centraltendency of bids.

Keywords Aggressive · Bidder asymmetry · Dispersion · Entrants · Incumbents ·Procurement auctions

JEL Classification D44 · D82 · L74

1 Introduction

Subcontracting is a common business practice in procurement markets. In the construc-tion industry, two interesting phenomena are observed: first, there is greater variance

S. Li (B)Chinese Academy of Finance and Development, Central University of Finance and Economics,39 South College Road, Beijing 100081, People’s Republic of Chinae-mail: [email protected]

P. PhilipsDepartment of Economics, University of Utah, 1645 Campus Center Dr. Room 308, Salt Lake City,UT 84112-9300, USAe-mail: [email protected]

123

192 S. Li, P. Philips

in auctions with entrant subcontractors than there is among experienced firms. Second,the likelihood of winning a bid is higher for entrant subcontractors than it is for expe-rienced subcontractors. These two phenomena may result from the specific biddingbehavior of entrant subcontractors.

Regarding the first phenomenon, several reasons may explain the higher varianceamong entrant subcontractors’ bids. For instance, it might be caused by an increasingnumber of bidders, entrant subcontractors’ lack of experience in estimating projectcosts, etc. Regarding the second phenomenon, the intuitive explanation is that entrantsubcontractors bid more aggressively than do experienced subcontractors, in an effortto gain a foothold in the industry. In this paper, we collected data from Utah construc-tion subcontractor auctions and used these data to investigate these two phenomenaby comparing the bidding styles of entrants and incumbents, which is an importantsubject in asymmetric auction studies.

In our analysis, we placed the hypothesized bidding behavior by entrants in eco-nomic and institutional contexts by controlling for the subcontractor’s history, basedon the experience of his/her firm, as measured by its age, the ratio of that subcontrac-tor’s wins-to-bids up to the time of the current auction, the current work backlog ofboth the bidder and its competitors in the auction, and the influence of the seasonalwork cycle.

Our central findings are that, relative to incumbent subcontractors, entrant subcon-tractors as a group have greater dispersion in their bids, a result that possibly can beattributed to entrants’ inexperience and uncertainty about a project’s true cost. In addi-tion, entrants bid more aggressively than do incumbent subcontractors for most typesof subcontracting projects. However, when facing entrant competitors in constructionprocurement auctions, incumbent subcontractors have less dispersion in their bids forsome types of subcontracting work.

This paper provides more evidence to support the prediction of asymmetric auc-tion theory that weak bidders will bid more aggressively when facing strong bidders.Entrant and incumbent subcontractors’ uncertainty about project costs and specifica-tion are also investigated in this paper. From a practical standpoint, this phenomenonmeans that the owners or general contractors who accept bids from contractors orsubcontractors benefit from increased competitive pressure during the auction, but runan increased risk that the lowest bid may be below the true cost of the project.

2 Literature

In many procurement auctions, bidders’ heterogeneity of size (Laffont et al. 1995)or capacity (Jofre-Bonet and Pesendorfer 2000) introduces asymmetric features intoauctions. Asymmetric auction theory assumes that bidders’ valuations are drawn fromdifferent distributions of bidders. Maskin and Riley (2000) seminal work shows thatin a static, first-price, sealed-bid auction, the weaker bidder reduces his bid furtherbelow his valuation than does the stronger bidder. They also show that a weaker bid-der, whose value is drawn from a weaker distribution, bids more aggressively (i.e.,closer to the actual value) when facing a stronger bidder, while stronger bidders bidless aggressively when facing weaker bidders.

123

Entrant Bidders in Construction Procurement Auctions 193

Several experimental and empirical studies have examined asymmetry betweenentrant and incumbent bidders. In a study of first-price auctions that was designed byGüth et al. (2005), weak bidders were found to bid more aggressively than did strongbidders. Examining road construction procurement auction data from the OklahomaDepartment of Transportation, De Silva et al. (2003) found that entrant contractorsbid more aggressively on highway projects. They also found a greater dispersion inentrants’ cost estimates. More recently, Estache and Iimi (2010) study that examinedpublic procurement auctions for infrastructure projects in various developing coun-tries also found that entrant bidders that faced incumbents bid more aggressively onroad projects.

The remainder of this paper is organized as follows: Section 3 describes Utah con-struction procurement auctions and the data we collected on these auctions. Hypothe-ses and models are presented in Sect. 4. Regression results are reported and discussedin Sect. 5. Section 6 offers conclusions, and practical implications are discussed inSect. 7.

3 Construction Procurement Auctions in Utah

Our sample of construction auctions comes from the Builders Bid Service of Utah(BBS), which is a non-profit bid depository. BBS collects subcontractor bids, sortedby subcontractor type, and assembles a simultaneous set of sealed-bid openings andpresents these to all general contractors, which then bid to the project owner immedi-ately thereafter. This one-stop shopping system allows all general contractors to seeall subcontractor prices prior to submitting a bid to the project owner. Note that thisanalysis uses only data regarding subcontractors’ bids to the general contractors.

Both De Silva et al. (2003) and Estache and Iimi (2010) define a contractor as an“entrant” at that point in time when that firm makes its initial bid in their data. In con-trast, we conceptualize entrants as inexperienced “rookies” that are seeking a footholdin the industry over time. Consequently, we classify subcontractors as entrants in allauctions in which they participated during the first year in which they appear in ourdata.

As shown in Table 1, we have gathered data for 7,514 auctions involving 47,399bids spanning the years 1969 to 2003.1,2 The subcontractors in our sample includethose who specialize in four areas: ceramic tile, masonry, painting, and roofing. Inthe BBS data, there are only two types of auctions: auctions with only incumbentsparticipating, and auctions with both incumbents and entrants (“mixed auctions”). Noauctions involving only entrants are included in our data. Of the auctions in our sample,

1 All bid records have been adjusted for inflation.2 For most construction projects, surety bonding is used to protect the project owner and developer. Newsubcontractors may not have the required performance history to qualify for the bonding certificate, andthose that do will have limited bonding capacity. This limits new subcontractors’ access to all constructionprojects. To compare the new firms and experienced firms on a fair standard, we limit the data set to withinthe 95th percentile of incumbents’ mean bids for all mixed auctions. For instance, the 95th percentile ofincumbents’ mean bid for all mixed tile auctions is $153,760.9, so we rule out all tile projects with a higherincumbents’ mean than $153,760.9.

123

194 S. Li, P. Philips

36% involved a mixture of incumbent and entrant bidders, while 64% involved onlyincumbent bidders.

The mean variance of bids in incumbent-only auctions (0.042) is lower than inmixed auctions (0.055). This indicates that less-experienced entrants experienced rel-atively greater uncertainty regarding the true cost of projects.3 As a group, entrantswon 18.66% of their auctions, while incumbents’ win ratio was 15.59%.4 In mixedauctions, incumbents’ win ratios dropped even further to 11.49%. Entrants’ higherwin ratios provide preliminary evidence that they bid more aggressively, while incum-bents’ lower win ratios in mixed auctions sheds light on the possibility that incumbentsbid more conservatively when facing entrants.

4 Hypotheses and the Empirical Model

First we focus on the bidding strategies of entrant subcontractors. Unlike incum-bents, who have greater knowledge regarding both the limits of their capabilities andthe meaning of various specification requirements, entrants lack estimating and con-struction experience and, therefore, face greater uncertainty regarding both their owncapabilities and the full implications of the requirements that are established by thespecifications of a contract. That entrants’ bids are more widely dispersed around thecentral tendency of bids reflects greater collective uncertainty regarding the value ofthe project.

It is also important to note that a project may be more valuable to the entrant firm thatseeks a foothold in construction than it is to established subcontractors. New entrantsneed to win work in order to establish a favorable reputation, build their bondingcapacity, secure a flow of revenue to sustain their new business, and attract qualifiedemployees. Thus, relative to established incumbents, entrants may be more inclined toreduce their markup by bidding more closely to their estimated cost of construction.This constitutes a more aggressive bidding strategy.

Thus, we can formulate two hypotheses regarding entrants’ bidding behavior: First,we propose a hypothesis that reflects the uncertainty that is inherent in subcontracting:Bids by entrants, who grapple with greater uncertainty, will have more variance thanincumbents’ bids. Second, we propose a hypothesis that reflects the aggressiveness ofsubcontractors in their bidding: Entrants will bid lower than will incumbents.5

3 However, this greater project variance may simply be an artifact of the increased mean number of biddersin mixed auctions (9.66), as compared to incumbent-only auctions (7.0).4 In the bids we study, in which subcontractors bid to general contractors, the low bid typically wins.However, general contractors will occasionally reject the subcontractor with the lowest bid if there areconcerns regarding the prospects of the subcontractor performing adequately. Still, there is pressure on thegeneral contractor to accept the lowest subcontractor bid, given the prospect that the general contractor’scompetitors also will select the lowest bid for subcontracting work.5 There is some ambiguity in distinguishing these hypotheses. Uncertainty might lead to aggressive entrantbidding if both entrants and incumbents’ bids have roughly the same means. Greater variance amongentrants’ bids would imply a greater likelihood that an entrant’s bid would be the lowest. Alternatively,aggressiveness could be the reason for greater dispersion of entrant bids if there is a longer tail of lowentrants’ bids, as compared to incumbents’ bids.

123

Entrant Bidders in Construction Procurement Auctions 195

We test these phenomena separately by defining two measurements meant to cap-ture these interrelated, but distinguishable behaviors: To operationalize uncertainty,DISPERSION for each auction is defined as follows:

DISPERSION = |bid − incumbent’s mean|incumbent’s mean

DISPERSION measures how far each subcontractor’s bid is away from the centraltendency of experienced subcontractors’ bids in the auction. In studies of general con-tractor auctions, engineers’ or architects’ estimates are often used to approximate thetrue value of a project (Bajari and Ye 2003; De Silva et al. 2003; Estache and Iimi2010). Lacking such estimates for subcontractor auctions, we use the mean of bids byexperienced subcontractors to estimate the true value of a project.6

Next, we construct RELATIVE_INC as a standardized measurement to capturehow aggressively subcontractors bid. RELATIVE_INC for each auction is defined asfollows:

RELATIVE_INC = bid

incumbent’s mean

Lower values of RELATIVE_INC indicate that a more aggressive bidder is submittinga lower bid, relative to the mean of incumbents’ bids.

Thus, our two primary hypotheses about entrants are:

H1.1 Entrants’ bid dispersion will be greater than that of incumbents.H2.1 Entrants will bid more aggressively than do incumbents.

Next, we need to determine whether or not the presence of entrants in auc-tions causes different bidding behavior among incumbents. Facing competition fromentrants, incumbents may bid more cautiously, willingly losing some bids in order toavoid the prospect of bidding too low and, as a result, failing to perform, damagingtheir established reputations, and imperiling their bonding capacity. Following Hongand Shum (2002), under a common-value setting, established contractors bid lessaggressively to avoid the winner’s curse. Thus, in general, experienced contractors bidmore conservatively in order to protect their established businesses.

When they know that they are competing against entrants, incumbents may taketheir bidding more seriously and, as a result, may more carefully evaluate a project,which, in turn, leads to a narrower dispersion around the central tendency. This biddingstrategy helps incumbents avoid the trap of the winner’s curse in which incumbentsbid so low that they lose money after winning the project.

We also put forward two secondary hypotheses about incumbents:

H1.2 Facing entrants, incumbents’ bids will be less dispersed.H2.2 Facing entrants, incumbents will bid more conservatively.

6 Because incumbents have survived for at least a year—in most cases, many years—their experience andlonger-term survival make the central tendency of their bids a reasonable estimate of the true value of theproject.

123

196 S. Li, P. Philips

To test both H1.1 and H1.2, the following Model 1 is constructed:

LN_DISPERSIONit = α · ENTRANTit + β · INC_MIXEDit

+χ · Z ′it + δ · W ′

it + φ · Y ′it + εit

To test both H2.1 and H2.2, a comparable Model 2 is constructed:

LN_RELATIVE_INCit = α · ENTRANTit + β · INC_MIXEDit

+χ · Z ′it + δ · W ′

it + φ · Y ′it + εit

In both models, we adopt the logarithmic form of the dependent variable. The focusvariables are ENTRANTit and INC_MIXEDit. ENTRANTit equals 1 if the bidder i isclassified as an entrant in auction t; otherwise it equals 0. INC_MIXEDit is defined as1 when an incumbent is bidding in a mixed auction; otherwise it equals 0.

The control variables can be divided into three sets: Zit controls for bidder char-acteristics, which include the age of the firm (FIRM_AGE), the firm’s historicalratio of auctions won to total auctions entered (HIST_WIN_RATIO), and the nat-ural log of the value of the current backlog of projects (LN_BACKLOG); Wit, whichcontrols for the aggregate value of backlogged work among rivals in the auction(LN_RIVALS_BACKLOG); and Yit, which controls for seasonal and annual varia-tions in bidding behavior with a monthly dummy variable (MONTH) and a yearlydummy variable (BID_YEAR). Summary statistics, including original and log formsfor both dependent variables, are provided in Table 2.

Next, we discuss our expectations for the control variables that are used in thesemodels:

FIRM_AGE. A subcontractor’s construction experience is measured in years byFIRM_AGE, beginning with 1 when the subcontractor first bids in the BBS data. Wehypothesize that more experienced subcontractors more accurately estimate projectcosts and, therefore, bid closer to the central tendency of bids. FIRM_AGE, whichreflects business longevity, also measures the subcontractor’s reputational and eco-nomic circumstances. Because it has a more established and more secure economicfoothold, an older firm is less likely to bid aggressively, relative to younger incumbentfirms and entrants. We, therefore, expect a positive sign for FIRM_AGE on LN_REL-ATIVE_INC.

HIST_WIN_RATIO. Some subcontractors may have individual bidding stylesthat are more aggressive than others, due either to variation in their risk preferencesor differences in firm cost structures. Risk-taking subcontractors and firms with lowercost structures are more likely to have won past auctions.

We use HIST_WIN_RATIO to estimate differences among firms in risk preferencesand cost structures. This is defined as the ratio of past auctions that have been wonby a subcontractor to the total past auctions in which that subcontractor participated.Similar to De Silva et al. (2003), we expect past winners to be currently more aggres-sive bidders, and, therefore, the sign of HIST_WIN_RATIO on LN_RELATIVE_INCis expected to be negative. However, we are agnostic regarding the effect of historicalwins on the dispersion of bids.

123

Entrant Bidders in Construction Procurement Auctions 197

LN_BACKLOG. The current value of the auctions won by the subcontractor overthe previous year is used as a proxy for whether or not the firm is facing capacity con-straints. Here we assume that all projects take one year to complete and all projectswon in a particular year are fully counted as a firm’s backlog through the end of thatyear, regardless of their time span. However, both of these assumptions are rough, atbest. According to this definition, 31% of subcontractors had a backlog of zero at thetime of the auction. Because we use a natural log form, LN_BACKLOG, we adjustfor zero backlogged work with ln(BACKLOG+1).

Jofre-Bonet and Pesendorfer (2000) analyze California highway and street con-struction auctions and find that contractors’ behavior is affected by capacity con-straints. In addition, De Silva et al. (2003) and Estache and Iimi (2010) find that firmswith greater backlogs bid less aggressively. Theoretically, the busier a contractor is, thelower is the opportunity cost to that contractor of losing an additional auction, whichmay induce less aggressive bidding. Thus, we expect a positive sign for LN_BACK-LOG on LN_RELATIVE_INC. Once again, however, we are agnostic regarding theway in which work backlogs may affect bid dispersion.

LN_RIVALS_BACKLOG. Project owners that solicit construction bids typicallyprovide plan specifications and often call public meetings to describe projects to pro-spective contractors and subcontractors. These procedures inform potential biddersof their potential rival contractors. RIVALS_BACKLOG is constructed as the sum-mation of all eventual rivals’ backlogged work in any one auction. Because we usea natural log form, LN_RIVALS_BACKLOG, we adjust for zero backlogged workwith ln(RIVALS_BACKLOG + 1).

Presuming that busy rivals will bid less aggressively, a subcontractor that facesalready occupied competitors may seek to reduce the cost of estimating a project. Theresulting more casual estimation should lead to a wider dispersion around the centraltendency of bids. Thus, we expect the sign of LN_RIVALS_BACKLOG on LN_DIS-PERSION to be positive. We are agnostic regarding whether or not subcontractorsfacing busy rivals will bid more aggressively.

5 Results

Using an OLS regression, we test our hypotheses with LN_DISPERSION as the depen-dent variable in Model 1 and LN_RELATIVE_INC as the dependent variable in Model2. Our primary hypotheses that entrant subcontractors are less certain of the true cost ofthe project, yet bid more aggressively, are confirmed by the results shown for models1 and 2 in Table 3, respectively. Relative to more experienced incumbent subcontrac-tors, we find a statistically significant and substantial increase in entrants’ uncertaintyregarding the cost of a project, along with a statistically significant and meaningfullymore aggressive bidding among entrants.

However, no evidence is found to support hypothesis H1.2 that incumbents bidcloser to the central tendency when entrants are present in auctions. The result of Model2 goes against our hypothesis that incumbents are more conservative in their bidding.When facing competition from an entrant, our result shows that incumbents actuallybid slightly more aggressively, rather than more conservatively, as we had expected.

123

198 S. Li, P. Philips

Our results also indicate that more experienced subcontractors are more cautiousbidders. The estimated coefficients for FIRM_AGE show that better established, moreexperienced subcontractors bid less aggressively than do entrants, with bids from theformer falling closer to the central tendency of incumbent bids. In Model 2 we findthat firms with higher HIST_WIN_RATIOs bid more aggressively, a result that reflectseither a greater appetite for risk or a lower cost structure. Although we had been agnos-tic regarding the effect of past wins on current bid dispersion, we find that an increasein the historical win ratio results in a corresponding increase in bid dispersion. It maybe that taking a casual approach to the estimation of project results is a successfulbidding strategy that results in many wins.

As expected, Model 2 shows that subcontractors with more work in backlog bidless aggressively. Negative and significant in Model 1, the estimated coefficient forLN_BACKLOG indicates that busy subcontractors who do not need new work asmuch as others hew closer to the consensus regarding project costs and eschewaggressive bidding. When competition is reduced because rivals already are busy(higher LN_RIVALS_BACKLOG), we find that subcontractors bid less carefully (highLN_DISPERSION). However, the results do not indicate that subcontractors bid moreaggressively or less aggressively under these less-competitive conditions.

Model 1 shows that there is a seasonal bid dispersion pattern, with bid dispersionnarrowing as the peak season approaches (i.e., from December-January, the reference,through August).7 This corresponds to seasonal increases in the value of projects,which rise from a mean of $134,972 in January to highs of $170,407 in May and$172,459 in June, only to fall back to $150,082 in December. As the value of projectsincrease coming into the busy season, we speculate that subcontractors invest morein estimation in order to reduce lost-markup risk, in which a winning bid results in adamaging underestimate of true costs. These results are consistent with Bajari and Ye(2003) analysis, which argues that contractors may choose to bid below cost duringslack times to avoid losing skilled workers to other, aggressively bidding contractors.

Table 2 shows that the peak months for auctions are February, March, April, andMay, with bid openings decreasing in June and falling fairly steadily thereafter, untilDecember and January. In Model 2, with December-January as the reference, MONTHdummy variables show that subcontractors bid less aggressively from May to June,since many firms already have plenty of work for the summer and lack the capacityor the desire to take on new projects at that time of year.

Bidding behavior may vary by subcontractor type for various reasons, such as dif-ferences in entry barriers. In the BBS data, the entry barrier increases sequentially fromtile to masonry to painting to roofing. To check whether or not entrant and incumbentsubcontractors in different fields exhibit different dispersion and aggression behavior,we deploy the Chow test on our models.

The Chow test compares the coefficients of all independent variables for the pairsof pooled datasets and four subcontractors. The test results, reported in Table 4, showthat all OLS multiple regression functions differ across subcontractors groups. No

7 Considering the cold outdoor temperatures, both January and December traditionally are considered tobe the slack season in Utah’s construction industry.

123

Entrant Bidders in Construction Procurement Auctions 199

individual subcontractor dataset shares a similar structure with the pooled dataset.This suggests the need for further study by each subcontractor type.

Regression results by subcontractor types (tile, masonry, painting, and roofing)are reported separately in Tables 5 and 6. As shown in Table 5, the coefficients forENTRANT are consistently positive and statistically significant for all four subcon-tractor types; this is a result that indicates greater entrant bid dispersion relative toincumbents. Hypothesis H1.1 is again supported by this result, which may be inter-preted to mean that entrants across all four fields of subcontracting are more uncer-tain than are incumbents with regard to their own capabilities and/or the demands ofcontract specifications.

Regarding hypothesis H1.2, only two subcontractor fields show significant results.In masonry and painting, incumbents that face entrants brought their bids closer tothe incumbents’ mean. However, bids from tile and roofing subcontractors did notconcentrate toward the incumbent mean. The industry-disaggregated samples showedno conflicting results for three categories of control variables, as shown in Table 5.

As for hypothesis H2.1, entrant subcontractors in tile, masonry, and roofing bidmore aggressively than did incumbents, as shown in Table 6. Bids by subcontrac-tors in painting did not follow that trend. However, no significant result is found forINC_MIXED, which refutes the outcome we observed in the pooled data set. Theindustry-disaggregated samples do not show that incumbents bid more conservativelywhen they face entrants.8 When we examine the control variables in Table 6, the coeffi-cients of FIRM_AGE, HIST_WIN_RATIO, and LOG_BACKLOG show a consistentpattern, as we observed in the pooled dataset in Table 3.

Bajari and Ye (2003) note that competitors’ capacity/work backlog should be treatedas an important factor in constructing an auction model. However, no significant resultis found in their study or the later study by De Silva et al. (2003). Using the BBS datafor masonry and painting, we find rivals’ backlogs affect bidders’ behavior. Whentheir rivals are known to be busy, masonry and painting subcontractors take the risk ofbidding more aggressively. In other words, when they sense that they are likely to winan auction because their rivals are already busy with other projects, these two types ofsubcontractors bid more aggressively to ensure that they will win.

6 Conclusions

Our primary finding that entrant subcontractors bid more aggressively is supportedfor each subcontractor type—except painting. This finding is consistent with the

8 We disaggregate by four contractor types in an unreported quantile regression. For roofing only, we finda significant coefficient for INC_MIXED at the 0.9 percentile of LN_RELATIVE_INC, of 0.030. Otherthings being equal, this implies that the 90% percent of LN_RELATIVE_INC for a roofing incumbentfacing an entrant is 0.030 above that of a roofing incumbent not facing an entrant.

However, at the 0.75 and 0.5 percentiles, the coefficients are not significant, but for the 0.25 and 0.1percentiles, both coefficients are -0.012 and -0.017, respectively. This pattern provides evidence showingthat incumbents that face entrants bid most aggressively at the 0.1 percentile of lowest bids, with biddingaggressiveness declining between the 0.1 to the 0.25 percentiles. However, in the case of roofing subcon-tractors, the coefficient reverses itself to positive at the 0.9 percentile. These results explain the negativesign that we obtained for the regression on the pooled data set in Table 3.

123

200 S. Li, P. Philips

asymmetric auction theory, put forth in Maskin and Riley (2000), which posits that aweak bidder that faces a strong bidder will bid more aggressively. Our results also areconsistent with De Silva et al. (2003) and Estache and Iimi (2010) construction studies,which find that entrants bid more aggressively than do incumbents. Our finding thatentrants’ bids are more dispersed probably reflects, at least in part, a greater uncertaintyamong entrants regarding a project’s requirements and their own capabilities.

Evidence from both masonry and painting subcontractors supports our hypothesisthat incumbents’ bids will be less dispersed when they must compete against entrantsin an auction. Perhaps this can be attributed to incumbents’ strategies for dealing withcompetition from entrants, which leads incumbents carefully to evaluate a project andstay close to the central tendency in order to avoid losing money from winning anauction with a lower-than-cost bid.

We contextualized entrant-incumbent bidding behavior within a broader model thatcontrols for experience, bidding style, the capacity constraints of a firm and its rivals,and seasonal patterns in bidding. The results show that increased experience leadsto less aggressive bidding and less uncertainty regarding the value of a project. Wealso found that bidders who were aggressive in the past continued their aggressivebidding into the present, a result that reflects either an unmeasured cost advantage orperhaps greater risk tolerance. Finally, we found that bidders were more aggressivewhen their rivals were busy with larger backlogs of work. Seasonal cycles, along witha subcontractor’s backlog of work, also altered bidding strategies—with increasedaggressiveness exhibited in slack seasonal periods.

7 Implications for Construction Procurement Auctions

Within the context of auctions for subcontracting work, our study emphasizes both therisks and benefits to general contractors and project owners of having inexperiencedsubcontractors participate in auctions. Needing a foothold in the industry, entrant sub-contractors become risk-takers, sacrificing the markup that provides a safety marginbetween the cost of a project and the accepted bid price. Entrants assume greater risk,even though they are less certain of the true cost of a project than are their moreexperienced rivals. Greater risk-taking by entrants who also face increased uncer-tainty increases the probability that a low bid from an entrant subcontractor will beinsufficient to complete a project satisfactorily.

As such, general contractors that receive bids from entrant subcontractors face adilemma: Accepting the lowest bid from an entrant subcontractor raises the risk ofwork failures and interruptions occurring, but rejecting the lowest bid raises the risk oflosing in the general-contractor auction held by the project owner, since rival generalcontractors may include the entrant-subcontractor’s low bid in their bids.

Thus, while the participation of entrants may enhance competitive pressures at thesubcontractor level, overall competitive pressure in these multilayered auctions maylead otherwise cautious and experienced general contractors to accept riskier bids fromentrant-subcontractor. Project owners, in turn, may never see the higher, but less risky,bids of more experienced subcontractors if, in an effort to offset the prices of theircompetitors, all the general contractors in an auction select the same low bid from therisk-taking rookie subcontractor.

123

Entrant Bidders in Construction Procurement Auctions 201

It is also important to note that seasonal patterns shape construction bidding, withslack times generally leading to more aggressive bidding. Thus, the purchasers of con-struction services through auctions should anticipate that 1) the participation of entrantsubcontractors and 2) auctions held in the slack season are more likely to yield lowerprices. While inexperience, reduced investment in cost-estimation, and/or compelledrisk-taking during slack times may lower auction prices, these factors also may bringwith them a greater risk of project interruptions, quality concerns, or a contractor’sfailure to perform.

Acknowledgements We are grateful to our editor, Lawrence White, and two anonymous referees fortheir insightful comments, which have greatly improved this paper. Any remaining errors are those of theauthors.

Appendix

See Tables 1, 2, 3, 4, 5, and 6.

Table 1 Summary statistics of Utah construction auctions

Variable All samples Tile Masonry Painting Roofing

Number of auctions 7,514 1,573 2,074 2,185 1,682

Number of firms 2,034 256 794 707 277

Number of bids 47,399 9,385 16,374 14,053 7,587

Number of bids by entrants inmixed auctions

4,024 551 1,464 1,499 510

Number of wins by entrants inmixed auctions

751 102 267 256 126

Win ratio of entrants in mixedauctions (%)

18.66 18.51 18.24 17.08 24.71

Number of bids by incumbents 43375 8,834 14,910 12,554 7,077

Number of wins by incumbents 6763 1,471 1,807 1,929 1,556

Win ratio of incumbents in allauctions (%)

15.59 16.65 12.12 15.37 21.99

Number of auctions having onlyincumbentsa

4,792 1,129 1,216 1,175 1,272

Mean variance of auctions havingonly incumbents

0.042 0.032 0.032 0.076 0.028

Number of mixed auctions 2722 444 858 1010 410

Mean variance of mixed auctions 0.055 0.033 0.037 0.089 0.039

Number of bids by incumbents inauction having only incumbents

26,221 6,130 8,242 6,630 5,219

Number of bids by incumbents inmixed auction

17,154 2,704 6,668 5,924 1,858

Number of wins by incumbentsin mixed auction

1,971 342 591 754 284

Win ratio of incumbents in mixedauctions (%)

11.49 12.65 8.86 12.73 15.29

a Each auction’s variance is the mean of [(bid/incumbents’mean) − 1]2

123

202 S. Li, P. Philips

Table 2 Summary statistics of regression variables

Variable Pooled mean Tile mean Masonry mean Painting mean Roofing mean

Dispersion 0.153a 0.125 0.133 0.214 0.117

[0.164] [0.133] [0.132] [0.209] [0.133]

ln_dispersion −2.441 −2.616 −2.522 −2.058 −2.758

[1.426] [1.341] [1.324] [1.441] [1.555]

relative_inc 1.000 1.000 0.996 1.006 0.998

[0.224] [0.183] [0.187] [0.299] [0.177]

ln_relative_inc −0.023 −0.015 −0.021 −0.034 −0.017

[0.216] [0.176] [0.182] [0.282] [0.180]

entrant 0.085 0.059 0.089 0.107 0.067

[0.279] [0.235] [0.285] [0.309] [0.250]

inc_mixed 0.362 0.288 0.407 0.422 0.245

[0.481] [0.453] [0.491] [0.494] [0.430]

firm_age 10.602 11.781 9.909 9.606 12.488

[8.077] [8.200] [7.652] [7.785] [8.812]

hist_win_ratio 0.139 0.156 0.107 0.134 0.197

[0.120] [0.113] [0.111] [0.123] [0.116]

ln_backlog 8.528 9.832 7.540 7.589 10.783

[5.820] [4.679] [6.454] [5.510] [5.244]

ln_rivals_backlog 13.852 13.711 14.493 12.889 14.424

[2.114] [1.106] [2.255] [2.116] [5.244]

month1 0.061 0.057 0.063 0.062 0.058

[0.239] [0.232] [0.243] [0.241] [0.233]

month2 0.104 0.094 0.111 0.104 0.100

[0.305] [0.291] [0.314] [0.305] [0.300]

month3 0.112 0.109 0.117 0.115 0.102

[0.316] [0.311] [0.321] [0.320] [0.302]

month4 0.113 0.106 0.114 0.112 0.123

[0.317] [0.308] [0.317] [0.316] [0.329]

month5 0.105 0.100 0.106 0.102 0.118

[0.307] [0.300] [0.308] [0.303] [0.322]

month6 0.088 0.098 0.086 0.083 0.093

[0.284] [0.297] [0.280] [0.276] [0.290]

month7 0.063 0.068 0.062 0.063 0.058

[0.243] [0.252] [0.240] [0.243] [0.234]

month8 0.084 0.092 0.082 0.085 0.078

[0.278] [0.290] [0.274] [0.278] [0.268]

month9 0.073 0.076 0.073 0.075 0.069

[0.261] [0.264] [0.260] [0.263] [0.253]

month10 0.075 0.082 0.070 0.080 0.072

[0.264] [0.274] [0.254] [0.271] [0.258]

123

Entrant Bidders in Construction Procurement Auctions 203

Table 2 continued

Variable Pooled mean Tile mean Masonry mean Painting mean Roofing mean

month11 0.056 0.054 0.056 0.056 0.059

[0.231] [0.227] [0.231] [0.230] [0.236]

month12 0.064 0.064 0.062 0.063 0.071

[0.245] [0.245] [0.241] [0.243] [0.257]

Standard deviations are bracketedYearly dummy variables are not reported due to space limitationsa In BBS data, a few bidders made extremely low or high bids in auctions, such as a bid is only 3% of, or10 times, the incumbents’ mean. By ruling out these records using a 5% criteria, we find that these outliersdo not affect the results of our later regressions

Table 3 Regression results ofpooled data

Variables (1) (2)ln_dispersion ln_relative_inc

entrant 0.399*** −0.044***

[0.025] [0.005]

inc_mixed −0.005 −0.005**

[0.016] [0.002]

firm_age −0.007*** 0.002***

[0.001] [0.0002]

hist_win_ratio 0.679*** −0.392***

[0.055] [0.010]

ln_backlog −0.018*** 0.001***

[0.001] [0.0002]

ln_rivals_backlog 0.038*** −0.001

[0.009] [0.0005]

month2 −0.002 0.006

[0.025] [0.004]

month3 −0.064* 0.006

[0.027] [0.004]

month4 −0.048 0.007

[0.0265] [0.004]

month5 −0.054** 0.008*

[0.027] [0.004]

month6 −0.083*** 0.009**

[0.030] [0.004]

month7 −0.071* 0.008

[0.032] [0.005]

month8 −0.073* 0.005

[0.030] [0.004]

month9 −0.012 0.003

[0.031] [0.005]

123

204 S. Li, P. Philips

Table 3 continued

Robust standard errors arebracketedYearly dummy variables are notreported due to space limitations*, **, and *** denote the 10%,5%, and 1% significance levels,respectively

Variables (1) (2)ln_dispersion ln_relative_inc

month10 −0.050 0.007

[0.032] [0.005]

month11 −0.039 0.006

[0.032] [0.005]

Constant −3.051*** 0.024***

[0.109] [0.009]

Observations 47,391 47,399

R-squared 0.026 0.045

Table 4 Chow test results

F-statistics are reported

Dependent variable lndispersion lnrelative_inc

Pooled Subcontractor

Pooled Tile 6.16 2.03

Pooled Masonry 7.50 2.03

Pooled Painting 27.96 3.92

Pooled Roofing 10.20 4.18

Table 5 Regression Results of Subcontractors’ Auctions with LN_DISPERSION

Subcontractor Tile Masonry Painting RoofingVariables ln_dispersion ln_dispersion ln_dispersion ln_dispersion

entrant 0.355*** 0.333*** 0.433*** 0.532***

[0.065] [0.041] [0.041] [0.072]

inc_mixed −0.058 −0.075*** −0.107*** −0.076

[0.035] [0.025] [0.030] [0.056]

firm_age −0.001 −0.005*** 0.001 −0.004

[0.002] [0.002] [0.002] [0.003]

hist_win_ratio 0.895*** 0.946*** 0.844*** 0.550***

[0.109] [0.075] [0.099] [0.182]

ln_backlog −0.016*** −0.007*** −0.014*** −0.021***

[0.003] [0.002] [0.002] [0.004]

ln_rivals_backlog 0.112*** 0.056*** 0.091*** 0.130***

[0.036] [0.014] [0.016] [0.029]

Constant −4.125*** −3.400*** −3.276*** −4.511***

[0.471] [0.172] [0.188] [0.384]

Observations 9,385 16,373 14,047 7,586

R-squared 0.041 0.035 0.040 0.064

Robust standard errors in bracketsMonthly and yearly dummy variables are not reported due to space limitations*, **, and *** denote the 10%, 5%, and 1% significance levels, respectively

123

Entrant Bidders in Construction Procurement Auctions 205

Table 6 Regression results of subcontractors’ auctions with LN_RELATIVE_INC

Subcontractor Tile Masonry Painting RoofingVariables ln_relative_inc ln_relative_inc ln_relative_inc ln_relative_inc

entrant −0.020* −0.078*** −0.014 −0.068***

[0.012] [0.007] [0.011] [0.013]

inc_mixed 0.002 −0.0009 −0.002 −0.002

[0.004] [0.003] [0.005] [0.005]

firm_age 0.003*** 0.002*** 0.003*** 0.001

[0.0003] [0.0002] [0.0004] [0.0004]

hist_win_ratio −0.370*** −0.380*** −0.548*** −0.265***

[0.017] [0.014] [0.021] [0.024]

ln_backlog 0.0005 0.001*** 0.0008 0.001**

[0.0005] [0.0002] [0.0005] [0.0005]

ln_rivals_backlog −0.00007 −0.002** −0.002* −0.001

[0.001] [0.0006] [0.001] [0.001]

Constant 0.023 0.033*** 0.043** 0.040**

[0.022] [0.012] [0.019] [0.020]

Observations 9,385 16,374 14,053 7,587

R-squared 0.059 0.064 0.052 0.030

Robust standard errors in bracketsMonthly and yearly dummy variables are not reported due to space limitations*, **, and *** denotes the 10%, 5%, and 1% significance levels, respectively

References

Bajari, P., & Ye, L. (2003). Deciding between competition and collusion. Review of Economics &Statistics, 85(4), 971–989.

De Silva, D., Dunne, T., & Kosmopoulou, G. (2003). An empirical analysis of entrant and incumbentbidding in road construction auctions. The Journal of Industrial Economics, 51(3), 295–316.

Estache, A., & Iimi, A. (2010). Bidder asymmetry in infrastructure procurement: Are there any fringebidders?. Review of Industrial Organization, 36(2), 163–187.

Güth, W., Ivanova-Stenzel, R., & Wolfstetter, E. (2005). Bidding behavior in asymmetric auctions: Anexperimental study. European Economic Review, 49(7), 1891–1913.

Hong, H., & Shum, M. (2002). Increasing competition and the winner’s curse: Evidence from procure-ment. Review of Economic Studies, 69(4), 871–898.

Jofre-Bonet, M., & Pesendorfer, M. (2000). Bidding behavior in a repeated procurement auction: Asummary. European Economic Review, 44, 1006–1020.

Laffont, J-J., Ossard, H., & Vuong, Q. (1995). Econometrics of first-price auctions. Econometrica,63(4), 953–980.

Maskin, E., & Riley, J. (2000). Asymmetric auctions. The Review of Economic Studies, 67(3), 413–438.

123