PROCUREMENT AUCTION

 

Submitted by

Mythri A (mythri@csa.iisc.ernet.in)

Sandhya G (sandhya@csa.iisc.ernet.in)

 

Auction

 

Auction can be defined as a market institution with explicit rules determining resource allocation and payments depending on the bids offered from the market participants. In an auction we will have a set of traders who would like to sell the product and set of traders who would like to buy the product. One set of traders will place the bids and other set can choose from the bids and payment is made depending on the rules.

 

Auctions can be classified into various types depending on the rules followed and number of buyers and sellers involved and number of items involved. There can be single unit auctions, multiple units of same product called multi unit auctions or auctions for multiple products called combinatorial auctions.

 

Auctions reduce the complexity of negotiations because there are fixed set of rules to be followed; they are ideal for the implementation in computer. Online auctions have an advantage that many traders can participate. Internet may make auctions more secure as bidders may come from all over the world and the algorithm can be designed not to reveal buyers and suppliers private information. Rapid advances in computational capability make it possible to conduct sophisticated allocation rules quickly and at low cost. Auctions can be classified into various types depending on number of buyers and sellers and the rules used in the auction. There can be also one seller and a set of buyers or vice versa. The techniques used to deal them are the same.

 

To participate in the auction, each supplier (buyer) is required to submit a bid to an auctioneer (also called a market intermediary or an agent). Based on these bids, the quantities to purchase (sell) from, and the payment to each supplier (buyer), are determined by the auctioneer through pre-specified rules known as a mechanism. A mechanism is incentive compatible (or induces truth-telling) if each supplier's dominant strategy is to submit her true production cost as her bid. Incentive compatibility is important because it is usually a requirement for an auction to be efficient. An efficient mechanism is one that guarantees an allocation of production quantities that minimizes the total system cost.

 

There is a recent emphasis on using auctions in the supply chain as an effective and efficient means of achieving lower acquisition costs, lower barriers for new suppliers to enter a market, and consequently better market efficiency. Exploiting recent advances in information technology, such auctions can be carried out through the Internet, referred to as online auctions. Online auctions allow geographically diverse buyers and sellers to exchange goods, services, and information, and to dynamically determine prices that reflect the demand and supply at a certain point of time so that efficient matches of supply and demand can be realized.

 

Procurement Auction

 

In Procurement Auctions there will be a single buyer who wants to procure some product and a set of sellers who are interested in selling the product. The buyer might require a single product or multiple products. The auction, which we use, will support the requirement of buyer for multiple units.

 

There are three dominant types of mechanisms for multi-unit auctions, known as Pay as-You-Bid, uniform-price, and VCG auctions (e.g. second-price auctions in the single-unit setting, also known as Vickrey auctions) (Klemperer 1999). The Pay-as-you-Bid auction is self-explanatory. In a uniform auction a uniform price is paid for each unit purchased. The price can be either the first rejected or the last accepted bid. In economics literature it is well known that neither Pay-as-You-Bid nor uniform-price auctions is incentive compatible or efficient, whereas the family of mechanisms known as VCG auctions, attributed to Vickrey (1961), Clarke (1971), and Groves (1973), are both incentive compatible and efficient. In a VCG auction, the buyers' payment to a supplier is based not only on the bids submitted, but also on the contribution that the supplier makes to the system by participating in the auction.

 

Though VCG auctions are incentive compatible, VCG auctions are rarely used because of possible cheating by the auctioneer. In an online auction, however, the auctioneer can be a web site (virtual computer agent), which receives bids, and decides awards and payments based on some pre specified algorithm so that this problem can be minimized or eliminated. Furthermore, the Internet may make auctions more secure as bidders may come from all over the world and the algorithm can be designed not to reveal buyers and suppliers private information. Rapid advances in computational capability make it possible to conduct sophisticated allocation rules quickly and at low cost. With the recent advances in information technology, VCG auctions have the potential to become an important mechanism for online auctions.

 

Procurement Auction Model

 

Most auctions are price-driven. However, there are other costs associated with integrating a supply chain, e.g., costs associated with transportation, capacity management, inventories, etc. The model described takes into consideration the transportation costs involved in the shipment of the products.

 

In the model described in the paper[1] a buyer can have set of distribution sites. Each buyer has his own private values called consumption quantities. Consumption quantities at each site form the consumption vector. This model also considers the transportation costs from the source to destination. It tries to minimize the overall cost i.e. the production cost and the consumption cost together. Each supplier has a production cost, which can be described as a convex function of the quantities that she produces at her production locations. Such a cost can already include a reasonable profit margin of the industry. It also assumes that each supplier is a rational, self-interest player who is trying to maximize her own profit (i.e. the payment received minus her production cost). With general convex production costs, one cannot simply combine the production and transportation costs and treat them as a single cost function. In addition, the buyer pays the transportation costs associated with every shipment from a supplier location to a buyer location. We assume that the buyer knows the per-unit transportation cost along each arc of the network.

 

The payment to be made by the buyer is determined by the Vickrey strategy. The allocations are made with all the sellers. Once an allocation is made, each winner is removed from the scene and the allocation problem is solved again with the remaining sellers. The difference between the prices in the two cases is the contribution made by the winner to the market and is given as an incentive to him. The use of Vickrey strategy ensures incentive compatibility and allocative efficiency.

 

Notations 

N = total number of supplier production facilities

K = number of suppliers

M = total number of buyer locations

Nk = set of production facilities owned by supplier k

kn = index for the supplier that owns production facility n

qm = consumption at demand center m, qm > 0

xn = production quantity at production facility n

ynm = quantity shipped from production facility n to demand center m

zkm =  S Nk ynm, total quantity shipped to demand center m by supplier k

Ck(xk) = production cost function for supplier k (R|Nk|® R)

Fk(xk) = bidding function from supplier k (R|Nk| ® R)

Tnm = cost for shipping one unit from production facility n to demand center m

Boldfaced letters represent vectors or matrices.

 

Auction T

 

In Auction T the buyer submits a fixed consumption vector q to the auctioneer. Supplier k submits to the auctioneer a bid function Fk(xk) for supplying xk units, for which she incurs a production cost Ck(xk), xk R|Nk|. The suppliers may or may not see the consumption vector. The auctioneer will decide the quantities awarded to each of the suppliers, the amount transported from each supplier location to each of the buyer locations, and the payments made by the buyer to the sellers. This is not an iterative process and only a single round is involved. This auction can handle multiple units of the products. The bidding used is closed bidding. The sellers do not know what others have bid. They might or might not know the consumption vector of the buyer. Since the payment strategy used is Vickrey which is incentive compatible we can assume all the sellers would bid their true values.

 

Under Auction T, the auctioneer will minimize the sum of the accepted bids and the transportation costs, for a given consumption vector q,

 

                                                    as Min. Sk( Fk(xk) + Sn Sm tnm * ynm     

                                                    Subject to

S n ynm  = qm,   m= 1, . . .,M 

 

                                                            Sm.  ynm  = xn,  n = 1, . . .,N

 

                                                            ynm > 0,   m= 1, . . .,M,   n = 1, . . . , N. 

 

Let P(q) be the optimal value of the objective function for a given q. Define Q = {q : q > 0, P(q) < infinity  } and we restrict qÎ Q to ensure sufficient supply capacity. Since Fk(·) is closed, for any q ÎQ, an optimal solution exists.

 

The objective function states that we want to minimize the total payment to be made by the buyer. This includes both the transportation costs and the production costs. Fk(xk) denotes the bid function provided by the seller. It represents the amount to be paid if the supplier supplies xk amount of goods. ynm denotes the amount transferred from n to m and tnm represents the transportation cost of the same. So when we add both production and transportation costs we will get the required objective function.

 

The first constraint says that amount supplied from all the production sites to each consumption site should be equal to the amount required at the consumption site.

 

The second constraint states quantities transported from each site to all other sites should sum up to the quantity produced at that site.

 

The payment is based on Vickrey strategy, so the winners would pay the second highest price. To calculate this price we should solve n linear programming problems if the number of winners is n. For each winner we will solve the above problem with additional constraint that Xk=0, to get the price buyer should have paid if the seller was not present in the auction.

 

Let (xT , yT ) be an optimal solution, and P(q) be the optimal value of the objective function with the additional constraint xk = 0 (i.e., supplier k does not participate in the auction).

The buyer will pay supplier

                                                   Y T k (q) = P-k(q) - Pk(q) + Fk(xTk )

 

where P-k(q) - Pk(q) is the bonus payment made to supplier k, representing the value she adds to the system by participating in the auction. The buyer pays supplier k her bid Fk(xTk ) plus her contribution to the system.

 

This payment scheme belongs to the general truth-inducing VCG family described in Nisan and Ronen (2001). Consequently, rational suppliers will bid their costs, Fk(xk) = Ck(xk), irrespective of other suppliers bids. Therefore, Auction T is incentive compatible for all suppliers.

 

Other variations of the Procurement Model

 

Auction R

 

In this auction, the buyer submits a utility function (a proxy for her true consumption utility function) to the auctioneer, rather than a fixed consumption vector. The auctioneer treats this as if it were the profit (excluding acquisition costs) that the buyer makes by consuming q, a consumption vector, and considers it when making production and transportation decisions. The quantities that the buyer will be awarded and her payments are the outputs of the auction. Under the VCG payment structure, Auction R is incentive compatible for the suppliers, but not for the buyer. The auction mechanism meets the consumption vector q at minimum production and transportation cost and hence, is efficient for providing q. It also typically results in a consumption vector q that fails to optimize the total supply chain. Under Auction R, we show that the buyer’s payments are always less than or equal to those in Auction T for providing the same level of consumption.

 

Auction S

 

In this the buyer submits a fixed consumption vector to the auctioneer who will decide the production quantities at all supplier locations and the buyer’s payments to the suppliers by minimizing the total production cost in the system. Transportation decisions are subsequently made to match the demand and supply at the lowest total transportation cost. It is obvious that Auction T achieves a lower total supply chain cost than Auction S. However, the buyer may prefer Auction S under certain circumstances. It is seen that typical regular-overtime production cost structure can lead to higher payments in efficient auctions to distort buyer behavior.

 

References

 

Efficient Auction Mechanism For Supply Chain Management  Rachel R. Chen, Ganesh Janakiraman Robin Roundy, and Rachel Q. Zhang

 

Vickrey, W., 1961,  Counterspeculation, Auctions, and Competitive Sealed Tenders , Journal of Finance, 16, 8-37.