Australasian Journal of Construction Economics and Building
the results of subseries analysis in detecting changes in bidding behaviour over time. The third zart details the test res{lts of the effects of xidding feedxac} inforyation on the xidders?? bidding competitiveness.
Exploratory Analysis
Table 1 shows the sample size for both the partial and full information feedback groups. For the partial information feedback group (i.e., Bidders A1 to A5), a total of 261 bids were obtained with the other 14 representing outliers or no-bid decisions. The full information feedback group (i.e., Bidders B1 to B5) provided 249 bids with 26 bids representing outliers and no-bid decisions. The removal of outliers was based on criterion set forth by the Hong Kong SAR government which considers all bids that are 25% above the lowest bid to be non- serious bids (Skitmore 2002). An one-sample Kolmogorov-Smirnov (K-S) test on the two data samples reveals a p-value less than 0.05, indicating a violation of the assumption of normal distribution. Thus, non-parametric tests were used for subsequent testing of hypotheses.
Bid
Feedback Conditions Partial Information Full Information
N 261 249
Percent 94.9% 90.5%
No-bid/Outlier N 14 26
Percent 5.1% 9.5%
N 275 275
Total
Percent 100.0% 100.0%
Table 1 Sample size for partial and full information feedback groups
For the exploratory analysis, scatter plots were used to display the spread of data for the partial and full information feedback groups as shown in Figures 1(a) and 1(b), respectively. LOWESS curves fitted to the scatter plots allow for bidding trends to be observed from the dataset. LOWESS stands for \The LOWESS fit line is based on local polynomial least squares fit to a set of data points. The fit is then \to provide a more accurate trend representation of the dataset (Hardle,1990). The robustified LOWESS curve is more resistant to effects of noise and/or marginal outliers in any particular dataset. It can be seen that there is a steep decreasing trend in MUCR (i.e., more competitive bids) for the partial information feedback group from Round 1 to 2. This can be
considered as a learning curve adjustment. In this, with the winning bids revealed to all bidders, the bidders were able to assess their bids relative to the winning bids, and to review their performance and adjust their bidding strategies accordingly. However, from Round 3 onwards, the curve begins to stabilise, zossixlu rezresenting the xidders?? oztiy{y xidding trend. It should be noted that the profit and loss statements provided to the bidders were not available until the end of Round 2 since the hypothetical projects have minimum project duration of two rounds. However, the feedback on profit/loss on winning bids did not trigger an immediate response in bidding trend. This observation is explained in Fu et al. (2004) where contractors becoming experienced through recurrent bidding and having obtained the optimal level of bidding strategy. Pawlowsky (2001) suggests that the steady state phase in bidding trend is because of the balance between the market forces and the knowledge of sustaining optimal bidding strategy. He advocates that a behavioural regularity is a symbol of best practices that were developed from a survival need to stay in the market and to make a profit. However, as Fu et al. (2004) pointed out, when the market forces change, learning occurs again and bidders will find alternative strategies to adapt to the new environment. This may explain the slight decreasing trend in MUCR for the last few rounds of the experiment for the partial information feedback group.
Soo, A and Oo, B L (2010) ‘The effect of information feedback in construction bidding’, Australasian Journal of Construction Economics and Building, 10 (1/2) 65‐75
70
Australasian Journal of Construction Economics and Building
(a) Partial information feedback group (b) Full information feedback group
Figure 1 MUCR scatter plots for partial and full information feedback groups
For the full information feedback group, there is a continually decreasing trend in the MUCR, leading to generally lower bids. The LOWESS curve shows a fairly \consistent bidding trend among the five bidders. It appears that the full bidding feedback
information throughout the experiment did not trigger an immediate response in bidding behaviour among the bidders. This could be due to the fact that the bidders had made full use of the bidding feedback information (identity and bids from all competing bidders) and were able to bid consistently and competitively. Another possible explanation for the rather consistent trend is that the extensive xidding feedxac} inforyation retards the xidders?? responses (learning) to new situations (Kagel and Levin 2002). The full information feedback group in the experiment have sufficient information to formulate their bidding strategies, and thus there is less reliance on experience and learning gained through recurrent bidding. Comparing the two groups, it can be observed from the LOWESS curves that the full information feedback group bid more consistently and with lower overall bids compared to the partial bidding feedback information group. These are denoted by the LOWESS curves for partial information feedback group being above MUCR of 1 and the full information feedback group having the curve mostly below 1. The latter suggests that Bidders B1 to B5 in full information feedback group had bid aggressively in general with tiny or even negative mark-ups (see Figure 1(b)). The amount of \feedback group also represents a lower bidding consistency, which is probably indicative of the adjustment in bidding strategies among Bidders A1 to A5.
Changes in Bidding Trend
The subseries analysis in Skitmore and Runeson (2006) was adopted here to test the hypothesis on changes in bidding trend over time. The disjointedness observed in the LOWESS curves over the eight bidding rounds was used as a basis to divide all bids for each information feedback group into subseries. A Friedman test was then applied to the multiple related subseries of bids for checking statistically significant differences in bidding trends.
Soo, A and Oo, B L (2010) ‘The effect of information feedback in construction bidding’, Australasian Journal of Construction Economics and Building, 10 (1/2) 65‐75
71
Australasian Journal of Construction Economics and Building
From the LOWESS curves in Figures 1(a) and 1(b), it can be seen that there are three disjointed trends for both the partial and full information feedback groups, indicating three subseries that comprised bids from Round 1 to 2, 3 to 4 and 5 to 8. The Round 2 disjoint co{ld zossixlu xe d{e to the release of zrofit and loss stateyents to the xidders?? (triggering a behavioural response), however the Round 5 disjoint common to both information feedback conditions may be purely coincidental. The mean ranks of each subseries were then calculated, and the Chi-square values found in Friedman test are reported in Table 2. It should be noted that the mean rank is different to an arithmetic mean, and is an average of the MUCR rank with respect to the round subseries. A higher mean rank for a particular subseries indicates that the subseries has a higher MUCR overall. Consistent with the decreasing trends in LOWESS curves, the results show that the mean rank for Round 1 to 2 subseries is higher than the subsequent two subseries in both information feedback groups. However, it is clear that there is no statistically significant difference in MUCR for the three related subseries for the partial information feedback group ( p = ??~4??3v~ ??eszite the “}in}s” in the respective LOWESS curve, the results indicate that the variations in bids over time in response to the given set of partial bidding feedback information are statistically insignificant. This seemingly stable trend indicated by the respective mean rank can partly be explained xeca{se the xidders did not atteyzt “aggressive” yanoe{vres s{ch as xidding at or xelo| the unbiased cost estimate (MUCR remained above 1.0 at all times).
In considering the full information feedback group, the results show that there are significant differences in MUCR across the three related subseries at the conventional 0.05 cut-off level. There is evidence to support that the bidders do modify their pricing behaviour based on the given set of full bidding feedback information. The respective mean rank indicates a rather steep decreasing trend in MUCR over the bidding periods, suggesting considerable
changes in the bid price in terms of both frequency and magnitude. This observation is similar to that of Neugebauer and Selten (2006). They found that bidders adjusted their bids yore freq{entlu |hen theu received clear inforyation on their coyzetitors?? xids~ Here, the hypothesis H1 |hich states that “inexperienced bidders do change their pricing behaviour
systematically in response to a given fek gn eiddinj needeico inngrpikign” is considered supported. It appears that learning does occur for both the partial and full information feedback groups but at different rates, with increasing accuracy and consistency in bid prices over the respective three subseries (i.e., decreasing differences in MUCR).
Round Subseries
Mean Rank
N
Chi-Square
df
p
Partial information feedback group
1 - 2 3 - 4 5 - 8
2.133 1.975 1.892
60
1.816
2
0.403
Full information feedback group
1 - 2 3 - 4 5 - 8
2.542 1.822 1.636
59
27.526
2
0.000
Table 2 Friedman test results for partial and full information feedback groups
The Effects of Information Feedback Conditions on Bidding Competitiveness
Table 3 shows the means of MUCR for the partial and full information feedback groups. It can be seen that the full information feedback group is more competitive on average with a lower mean MUCR compared to those in partial information feedback group. In addition, the standard deviation for the full information feedback group is lower, indicating a higher degree of consistency in their bidding attempts.
Soo, A and Oo, B L (2010) ‘The effect of information feedback in construction bidding’, Australasian Journal of Construction Economics and Building, 10 (1/2) 65‐75
72
Australasian Journal of Construction Economics and Building
N
Partial Information Full Information
261 249
Mean MUCR 1.0197 1.0057
Std. Deviation
0.0361 0.0284
Table 3 Comparison of means of MUCR
To validate hypothesis H2, a Mann-Whitney U test was performed to test the difference in bidding competitiveness in terms of MUCR between the two groups. The results shows that the mean MUCR for full information feedback group is lower than the partial information feedback group at p < 0.05 (U = 19356.6; Z = -7.899; p = 0.000). Hypothesis H2 is thus supported where inexperienced bidders with full bidding feedback information are more competitive than those with partial bidding feedback information. This suggests that, in construction bidding, a full information feedback condition would lead to lower average bids. The conjecture to explain the findings is that comprehensive bidding feedback information opens up an \decisions, and thus become more competitive and consistent in their bidding attempts. This conjecture is further supported by a paperback survey of the subjects with full bidding feedback information. The survey was allotted in every bidding round and the subjects were asked to indicate the usage of bidding data. For the eight bidding rounds there is a response rate of 90% whom had all indicated that they had utilised the bidding feedback information in formulating their bidding strategies and bid prices.
In examining the means of MUCR further, it is clear that the percentage mark-ups for both the partial and full information feedback groups are relatively low. While the partial information feedback group had a mean percentage mark-up of 1.97, a mean percentage mark-up as low as 0.57 was recorded for the full information feedback group. The general impression here is that the bidders had submitted bids with tiny or even negative mark-ups in the experiment, and that the resultant bids would likely result in a loss. In this case, the majority of the bidders in both groups recorded a loss at the end of experiment, with several instances of ‘s{icidal?? xids that res{lted in big losses. Although the phenomenon of submitting suicidal bids is not new in literature (e.g. Fellows and Langford 1980; Dyer and ??agel 199??v, it |o{ld seey that the xidders?? xidding xehavio{r in the exzeriyent is affected by loss aversion to different degrees. However, the observed bidding trend shall not nullify the hypothesis testing results since the method of assigning final cost for the hypothetical projects is purely random.
Conclusions
This research examines the effects of information feedback conditions in construction bidding through an experimental setting, where two groups of inexperienced bidders were supplied with full and partial bidding feedback information, respectively. The results show
that the variations in bids over time in response to a given set of full bidding feedback information are statistically significant, but not for bids from bidders with access to partial bidding feedback information. Hypothesis H1 is thus considered partially supported. The test results for hypothesis H2 show that inexperienced bidders with full bidding data from previous bidding rounds are more competitive than those with partial bidding feedback information. The mean mark-{z coyzetitiveness ratio (i~e~, xidder??s xid/{nxiased cost estimate) for the full information feedback group is 0.014 lower than the partial information feedback group. This suggests that full information feedback condition would lead to lower average bids in construction bidding. The findings add to both our theoretical and empirical understanding of construction bidding: an understanding of the process of changes in the
price of building work, and how the process can be manipulated through the release of bidding feedback information.
Soo, A and Oo, B L (2010) ‘The effect of information feedback in construction bidding’, Australasian Journal of Construction Economics and Building, 10 (1/2) 65‐75
73
Australasian Journal of Construction Economics and Building
For construction clients, this study has demonstrated the need to consider the level of bidding feedback information in their formulation of procurement strategies. However, of the little empirical evidence in literature, further work is required to determine the point at which xidding feedxac} inforyation can xe regarded as ‘s{fficient?? to aid efficiency in construction bidding. It may be possible to replicate this study by having experienced bidders as subjects to oxtain eyzirical s{zzort on the extent to |hich contractors?? xid zricing is affected xu varying information feedback conditions, and thus to establish the external validity of the experiment. For further testing, an alternative would be to apply a statistical modelling technique, say, mixed effects model to provide estimation for bidding trends over time in response to different information feedback conditions
Acknowledgements
The authors would like to thank Raymond Loh for his assistance in experimental design as well as collecting the bidding data for the experiment.
References
Chan, SM, R{neson, G and S}ityore, M (199??v ‘Changes in zrofit as yar}et conditions change?? An historical st{du of a x{ilding firy??, Construction Management and Economics, 14(3), 253-264
??e Ne{fville, R, Hani, EN and Lesage, Y (1977v ‘Bidding yodels?? effects of xidders' ris} aversion??, Journal of the Construction Division, ASCE, 103, 57-70
??e Silva, ??G, ??{nne, ??, ??an}anayge, A and ??osyozo{lo{, G (??????8v ‘??he iyzact of z{xlic inforyation on xidding in high|au zroc{reyent a{ctions??, European Economic Review, 52 (1), 150-181
??e Silva, ??G, ??osyozo{lo{, G and Layarche, C (??????9v ‘??he effect of inforyation on the xidding and s{rvival of entrants in zroc{reyent a{ctions??, Journal of Public Economics, 93 (1-2), 56-72
??re|, ??S and S}ityore, M (199??v ‘Coyzetitiveness in xidding?? a cons{ltant's zerszective??, Construction Management and Economics, 10 (3), 227-247
??re|, ??S, and ??ello|s, R?? (199??v ‘??eedxac} in constr{ction contract xidding??, Construction Modernization and Education: Proc., CIB Int. Conf., Oct 21–24, WP. Chang, ed., China Architectural & Building Press, Beijing
??{f|enxerg, M and Gnee??u, U (????????v ‘??nforyation disclos{re in a{ctions?? an exzeriyent??, Journal of Economic Behavior and Organization, 48 (4), 431-444
??uer, ??, ??agel, JH and Levin, ?? (1989v ‘A Coyzarison of Naive and Exzerienced Bidders in Common Value Offer Auctions - A Laxoratoru Analusis??, Economic Journal, 99 (394), 108-15 Dyer, D and ??agel, JH (199??v ‘Bidding in coyyon val{e a{ctions?? ho| the coyyercial constr{ction ind{stru corrects for the |inners?? c{rse??, Management Science 42 (10), 1463– 1475
Engelbrecht-wiggans, R and ??ato}, E (??????8v ‘Regret and feedxac} inforyation in first-price sealed-xid a{ctions??, Management Science, 54 (4), 808-19
Eszonda, ?? (??????8v ‘??nforyation feedxac} in first zrice a{ctions??, Rand Journal of Economics, 39 (2), 491-508
??ello|s, R?? and Langford, ??A (198??v ‘??ecision theoru and tendering??, Building Technology and Management, Oct, 36-39
??lanagan, R and Noryan, G (1985v ‘Sealed xid a{ctions?? an azzlication to the x{ilding ind{stru??, Construction Management and Economics, 3 (2), 145- 61
??riedyan, L (195??v ‘A coyzetitive xidding strategu??, Operations Research, 4, 104-112
Soo, A and Oo, B L (2010) ‘The effect of information feedback in construction bidding’, Australasian Journal of Construction Economics and Building, 10 (1/2) 65‐75
74
Australasian Journal of Construction Economics and Building
??{, w??, ??re|, ??S and Lo, HP (??????4v ‘Start-up and steady-state learning in recurrent xidding??, Building Research and Information, 32 (6), 484-496.Hardle, W (1990) Applied Nonparametric Regression, Oxford University Press, Cambridge
??saac, RM and wal}er, JM (1985v ‘??nforyation and consziracu in sealed xid a{ctions??, Journal of Economic Behavior and Organization, 6 (2), 139-159
Kagel, JH and Levin, D (2002) Common Value Auctions and the Winner's Curse, Princeton University Press, New Jersey
Kerzner, H (2006) Project Management - a Systems Approach to Planning, Scheduling and Controlling, John Wiley & Sons, New Jersey
??ortane}, ??O, Soden, JV and Sodaro, ?? (1973v ‘Profit analuses and seq{ential xid zricing yodels??, Management Science Series A-Theory, 20 (3), 396-417
McCaffer, R (1976) Cgnkrickgrf’ Biddinj Behiligmr ind Tender Price Predickign, unpublished PhD thesis, Loughborough University of Technology
McCaffer, R and Pettitt, AN (197??v ‘Bidding xehavio{r in zro??ect yanageyent??, The Project Manager, 1 (5), 5–8
Milgroy, P (1989v ‘A{ctions and xidding - a zriyer??, Journal of Economic Perspectives, 3 (3), 3-22
Milgroy, P and wexer, RJ (198??v ‘??he val{e of inforyation in a sealed-xid a{ction??, Journal of Mathematical Economics, 10(1), 105-114
Ministry of Finance (2005) Contracts and Purchasing Procedures, Ministry of Finance, Singapore
Ne{gexa{er, ?? and Perote, J (??????8v ‘Bidding 'as if' ris} ne{tral in exzeriyental first zrice a{ctions |itho{t inforyation feedxac}??, Experimental Economics, 11 (2), 190-202
Ne{gexa{er, ?? and Selten, R (????????v ‘??ndivid{al xehavior of first-price auctions: The iyzortance of inforyation feedxac} in coyz{teri??ed exzeriyental yar}ets??, Games and Economic Behavior, 54 (1), 183-204
New South Wales Government (2005) Code of Practice for Procurement , New South Wales Government, Australia
Oc}enfels, A and Selten, R (??????5v ‘??yz{lse xalance eq{ilixri{y and feedxac} in first zrice a{ctions??, Games and Economic Behavior, 51 (1), 155-170
Oo, BL and Lo, HP (????1??v ??he effect of xidding s{ccess on contractors?? xidding xehavio{r, Journal of Construction, Engineering and Management ASCE, (under review)
Oo, BL and Soo, A (????1??v ‘Bidding exzeriyent?? testing the effects of learning and inforyation feedxac} in constr{ction xidding??, AUBEA 2010 Conference, Melbourne
Pawlowsky, P (2001) The treatment of organizational learning in management science. Handbook of Organizational Learning and Knowledge, Oxford University Press, New York Runeson, G (2000) Building Economics, Deakin University Press, Geelong
R{neson, G and S}ityore, M (1999v ‘??endering theoru revisited??, Construction Management and Economics, 17 (3), 285-296
S}ityore, M (????????v ‘??dentifuing non-coyzetitive xids in constr{ction contract a{ctions??, Omega, 30 (6), 443-449
S}ityore, M and R{neson, G (????????v ‘Bidding yodels?? testing the stationaritu ass{yztion??, Construction Management and Economics, 24 (8), 791-803
S}ityore, M and Syuth, HJ (??????7v ‘Pricing constr{ction |or}?? a yar}eting vie|zoint??, Construction Management and Economics, 25 (6), 619–630
S}ityore, RM R{neson, G and Chang, X (????????v ‘Constr{ction zrice foryation?? f{ll-cost zricing or neoclassical yicroeconoyic theoru???, Construction Management and Economics , 24 (7), 773–784
Soo, A and Oo, B L (2010) ‘The effect of information feedback in construction bidding’, Australasian Journal of Construction Economics and Building, 10 (1/2) 65‐75
75