Procurement Procedure Reform as Seen in the Three Year Experiment on Random Cut Method in the Form of Designated Competitive Bidding
Faculty Fellow, RIETI
Hokkaido's experimental introduction of random cut method in the form of designated competitive bidding: background and outcome
Looking back on a Procurement Procedure reform undertaken by Hokkaido in the past three years, I would like to discuss the benefits of taking an institutional design engineering approach that incorporates an experimental economics methodology into institutional design (A thesis, upon which this article is based, is to be published in the August edition of "Kosei Torihiki" (Fair Trade)).
The Hokkaido Prefectural Government embarked on a sweeping reform of its procurement procedure in 2000. The move was prompted by the revelation in October 1999 of, and subsequent on-site investigation by the Fair Trade Commission into allegations of bid-rigging over an agricultural civil engineering works contract offered by the Kamikawa Subprefectural Government in Hokkaido. In April 2000, the prefectural government set up a tender administration committee and announced an action plan to reform the procurement procedure. In line with the action plan, a series of reform initiatives have been undertaken in the subsequent three years. The prefectural government has expanded the scope of general competitive bidding and begun to announce its ceiling price for a construction project prior to holding the bidding. The measure that attracted the greatest attention, however, was the introduction of the so-called "random cut method" in the form of designated competitive bidding. According to the guidelines on its implementation, this new system is defined as follows:
Under the random cut method in the form of designated competitive bidding, a preliminary selection based on certain criteria is carried out to choose candidate bidders of a number exceeding the actual number of participants in the bidding. Then, "random" number selection is used to "cut" some of the candidates - those exceeding the planned number - and name the remaining as officially selected bidders.
This arrangement has been designed with the intention of preventing contract-commissioning parties from arbitrarily selecting bidders. The new rule, however, was met with substantial criticisms. During a hearing with members of the local construction industry, many criticized the system, saying that: (1) it is difficult to compete with anonymous rivals whose identity would not be disclosed until the bidding date, and (2) contractors are having difficulty in making an annual business plan because they cannot predict the number of contracts for which they can participate in the bidding, as a result of their random determination.
Also, strong complaints have been filed by contractors who, despite being selected as a candidate, had been given no chance to bid as they were excluded by the random selection each time. In response to this, the prefectural government reduced the number of candidate bidders from 1.5 times the actual number of bidders (as designated in the initial plan) to 1.2 times from 2000 onward.
The three-year trial period for the random cut method in the form of designated competitive bidding has now ended, and in 2003, the Hokkaido Prefectural Government set up a tender supervising committee to begin review work with a view to possibly abolishing it.
Model analysis of random cut method in the form of designated competitive bidding
As mentioned above, the introduction of the random cut method in the form of designated competitive bidding was intended to prevent the arbitrary selection of bidders by contract-commissioning entities. However, sufficient discussion has yet to take place as to what changes this reform has brought to contractors' bidding behavior and the patterns of their bid-rigging practices.
In an attempt to review how this particular institutional change affected the overall procurement procedure, I have been studying this random cut method in the form of designated competitive bidding since 2001, taking both theoretical and experimental approaches. In my research, I used the simplest possible model so as to capture the true nature of this procurement procedure. The model is an expanded version of a model developed by RIETI Faculty Fellow Saijo Tatsuyoshi to study bid-rigging practices.
The model, as it is applied to my study, assumes that there are three incumbent local contractors, which have been routinely winning contracts under the conventional bidding system, plus one external contractor who has been added under the new bidding system. One of these four contractors is to be excluded by random selection. In this way, the probability of the external contractor being excluded is positive. If the external contractor is to be always selected as an official bidder, the model comes down to one of general competitive bidding. The model assumes that the three local contractors would resort to a bid-rigging scheme, as they are able to consult and agree on a side-payment arrangement among themselves. It was assumed so because even if the three companies do not resort to such a bid-rigging scheme, the model would also come down to competitive bidding. Thus, the model has been designed in such a way as to generate a management environment peculiar to the situation under the random cut method in the form of designated competitive bidding. Other assumptions of the model include that a ceiling price for bidding prices be announced prior to bidding taking place, and that construction costs are the same for all the participating contractors, and made known to all.
From the viewpoint of economic theory, the timing of the selection of official bidders and bidding are extremely important factors in the random cut method in the form of designated competitive bidding. When there is the possibility of bid-rigging, and if the selection of official bidders is to be carried out by the drawing of lots prior to bidding, (1) the external contractor is not among those selected, and the situation is the same as that under the conventional selective tendering system, (2) the external contractor is selected, and the situation is the same as that under competitive bidding. However, if the drawing of lots to select official bidders is to be conducted at the same time as bidding, or at a point between holding bidding and awarding a contract, a situation peculiar to the random cut method in the form of designated competitive bidding would emerge. Thus, pure strategy equilibrium does not exist. The reasons for this are as follows:
First of all, it must be noted that none of the contractors know the result of the draw. Under this situation, the optimum choice for both the three incumbent local contractors and the new external contractor is to win the contract by offering a bidding price slightly lower than the other side, which would lead to fierce price competition (Bertrand competition). If both the incumbent and external contractors anticipate such a scenario, their bidding prices would fall to the level barely exceeding the actual construction costs. However, if the three local contractors expect the external contractor's bidding price to be this low, then they can anticipate higher expected profits by offering a bidding price slightly lower than the ceiling price, betting on the possibility of the external contractor being excluded in the drawing of lots. But then, if the external contractor anticipates this scenario, the contractor would be better off setting a bidding price slightly lower than the ceiling price. In this kind of situation, no price is ideal, because every price is dominated by another, meaning that for any possible bidding-price a better price can always be found, and so pure strategy equilibrium does not exist.
Experimental investigation of random cut method in the form of designated competitive bidding
When equilibrium does not exist, what happens? Economic theories alone cannot provide a clear-cut answer to this. So, I attempted to find some kind of law or rules through economic experiment, following the methodology of experimental economics. In this experiment, the final contract price and the patterns of bid-rigging practices under the random cut method in the form of designated competitive bidding were reviewed from various angles. For instance, cases in which the selection of official bidders takes place before bidding are compared to cases in which the selection is made after bidding, but before awarding a contract. Also comparisons were made between cases with and without an external contractor. Comparisons against cases under conventional designated competitive bidding and competitive bidding were made as well.
The experiment results show that final contract prices under the random cut method in the form of designated competitive bidding tend to be lower than those under the conventional designated competitive bidding but higher than those under competitive bidding. In the conventional designated competitive bidding, bidding prices concentrate at a level slightly lower than the ceiling price. Meanwhile, in competitive tendering, price competition was observed at a level close to the actual cost. Final contract prices under the random cut method in the form of designated competitive bidding fell somewhere between them.
Bid-rigging practices of a similar pattern were observed in the random cut method in the form of designated competitive bidding, regardless of the timing of the lot for selecting official bidders. Typically, one of the three local contractors would tender a bid equivalent to the ceiling price and the others at amounts above. Then, if the first one - the designated bidder among the three - successfully won the contract, the profits would be shared among the three of them. In some groups, however, no bid-rigging practices were observed and fierce price competition took place. Under a competitive bidding system, no bid-rigging practices were observed. What has been reconfirmed in this experiment is that the presence of an external contractor is effective in preventing bid-rigging prices as well as an excessive fall in contract prices.
Since no equilibrium exists under the random cut method in the form of designated competitive bidding as mentioned above, there is no telling for sure at which price contractors should tender a bid, or whether or not contractors should resort to a bid-rigging scheme in the first place. Such uncertainty is reflected in the diversity in experimental results under the random cut method in the form of designated competitive bidding. Realistically speaking, this means that neither contractors participating in bidding, nor a commissioning entity offering a contract can foresee the bidding result. The contract may be won at a price nearly as high as the ceiling price, or at a price that barely exceeds the costs. This could cause substantial confusion among people on the work front. Indeed, this is underpinned by a series of remarks that contractors from across Hokkaido made during the hearing on the system. With these experiment results in mind, I told the Hokkaido Shimbun newspaper in an interview (on Aug. 30, 2001) that the fully-fledged introduction of the random cut method in the form of designated competitive bidding should be put on hold.
Reform of tendering system and institutional design engineering
In response to the 2002 enactment of a law that prohibits the involvement of public officials in bid-rigging, local governments are expected to further proceed on the reform of procurement procedure. Some local governments, other than Hokkaido, are beginning to use a lottery for selecting official bidders as a means to prevent bid-rigging attempts between a commissioning entity and contractors. Actually, in some local governments, a ceiling price for a public works project or services is selected randomly from within a designated range. But I am quite skeptical about these moves. Let me put aside for a moment the problem concerning the distortion of efficient allocation of resources caused by the use of a lottery. What I find more problematic is that those local governments seem to think of using a lottery in selecting official bidders and/or a ceiling price as a way of demonstrating that no collusion exists between them and the contractors, while giving little consideration as to its impact on the overall procurement procedure. An institutional reform based on such a shortsighted point of view, however, may end up incurring huge social costs later on.
On the other hand, it is also true that economic theory remains very much underdeveloped on issues relating to procurement procedures and bid-rigging practices. In this regard, economists also deserve criticism for their negligence. The role of economists in further promoting the reform of procurement procedures is expected to become increasingly important. Economists must strive to develop a theoretical economic model that can capture the true nature of a procurement procedure, allowing the study and experimentation of various theories and institutions under varying conditions by using the methodology of experimental economics, and then, propose reform measures.
An institutional designing process - referred to as institutional design engineering by RIETI Faculty Fellow Saijo Tatsuyoshi - should be utilized more in the actual work front. In the case of Hokkaido's reform attempt with the random cut method in the form of designated competitive bidding, confusion could have been minimized, had the scheme been subjected to thorough study under the institutional design engineering approach. In the United States and Europe, various policy measures - including those concerning the creation of an electricity market, an emissions trading market, and the designing of a hospital-intern matching system - have been proposed, based on research results achieved by an institutional design engineering approach. I do hope that such an empirical approach will become widespread as a means to formulating Japan's economic and industrial policies as soon as possible.
June 10, 2003