Multi-attribute auctions with different types of attributes: Enacting properties in multi-attribute auctions

https://doi.org/10.1016/j.eswa.2014.02.023 Get rights and content

Highlights

  • The role which different attributes play in multi-attribute auctions is analyzed.

  • A multi-attribute auction mechanism based on such analysis is presented.

  • VMA2 allows the incorporation properties like fairness or trust into auctions.

  • VMA2 behavior is exemplified describing how an egalitarian auction can be done.

Abstract

Multi-attribute auctions allow agents to sell and purchase goods and services taking into account more attributes than just price (e.g. service time, tolerances, qualities, etc.). In this paper we analyze attributes involved during the auction process and propose to classify them between verifiable attributes, unverifiable attributes and auctioneer provided attributes. According to this classification we present VMA2, a new Vickrey-based reverse multi-attribute auction mechanism, which takes into account the different types of attributes involved in the auction and allows the auction customization in order to suit the auctioneer needs. On the one hand, the use of auctioneer provided attributes enables the inclusion of different auction concepts, such as social welfare, trust or robustness whilst, on the other hand, the use of verifiable attributes guarantee truthful bidding. The paper exemplifies the behavior of VMA2 describing how an egalitarian allocation can be achieved. The mechanism is then tested in a simulated manufacturing environment and compared with other existing auction allocation methods.

Introduction

Auctions are gaining relevance in today’s economy. They are increasingly being used in fields such as industry procurement (Strecker, 2004), supply chain management (Sadeh & Sun, 2003), electronic advertisement (Fu et al., 2012), service allocation (Tavana, Abtahi, & Khalili-Damghani, 2014) and electronic commerce (Shih, Lin, & Huang, 2007). For instance, using auctions, manufacturers can obtain feedstock from their providers at the best economic conditions, while popular web sites can sell their ad space to the marketing companies which offer the most profitable revenue. In an auction, the auctioneer (e.g. manufacturer) that wishes to buy goods (e.g. feedstock) announces what he wants to sell in a call for proposals. Interested agents, bid for selling the goods at the price they consider will obtain some payoff. Then, the auctioneer clears the auction, by determining a winner, and setting up the price of the good.

Auctioneers wish to optimize their revenue, and for that purpose auction mechanism designers have to provide mechanisms that ensure incentive compatibility, meaning that bidders are encouraged to provide bids with a price equal to the true value of the goods (Myerson, 1979). Moreover, companies want to optimize their revenue in the long and mid-term rather than in a single transaction. Revenue is directly related to the number of participants; it has been proven that the revenue of an auctioneer, when k bidders are participating in an auction, is at least as high than with a subset of k-1 bidders (Lematre et al., 2003, Murillo et al., 2008). Thus, keeping the participants to an efficient number is something that auction designers should consider in order to optimize auctioneer revenues. In doing so, global satisfaction and the preferences of the auction participants should be represented, for example, they can be aggregated as social welfare measures do (Chevaleyre et al., 2006). This aggregation can be conducted in different ways, as maximizing the sum of all of the revenues and payoffs (utilitarian), or following an egalitarian approach, pursuing a higher equity between agents’ revenues (Chevaleyre et al., 2006, Endriss et al., 2003).

Most auction research has concentrated on the incentive compatibility and social welfare of the mechanism, regarding the price of the goods. Other properties, such as quality of service and delivery time, have been considered as side constraints on the auction process, but not in the core procedure of clearing the auction (Zhao, Zhang, & Perrussel, 2011). For example, in certain auctions, all the bids which do not achieve a minimum quality level are filtered out and discarded before determining the winner. Multi-attribute auctions have been designed to deal with different attributes in the winner determination process. In this case, the auctioneer is posed with the problem of choosing from among a set of Pareto-optimal solutions. For example, in a multi-attribute auction in which an auctioneer wishes to externalize a task, the auctioneer can characterize the task according to different attributes, such as price and quality. The auctioneer desires the highest quality at a good price. He will receive different bids of different quality and price; it is reasonable to think that quality and price are often conflicting criteria, and several combinations are equally optimal. Thus, the auctioneer should make a decision and pick a combination of attributes. Depending on how this selection is done, truthful bidding based exclusively on price is no longer valid. For example, a bidder that is encouraged to provide the true value for the price, may provide a false quality, resulting in harmful consequences for the auctioneer.

Our work concerns manufacturing and service environments, in which such attributes different than price are also important. For example, in one of the most frequent problems we are posed on, a company who owns a service distribution network (e.g. a telecommunications company or a power distribution company) must maintain the network in a wide operational area (e.g. at a country level). To this end, it outsources the network maintenance operations to third party entities distributed along the the operational area (resource providers) with different skills and costs. When there is an incident in the network, the company needs to assign the task maintenance to one of its providers, according to the tasks features, and also to the emergency level (i.e. an incidence that affects a health care organization has higher priority than a residential home). Moreover, to ensure that the company has providers with different characteristics (response time, skills, geographical distribution, etc.) and competitive prices, the company strategy includes keeping as much resources providers interested in his contracts as possible. For that purpose, the company may need to favor egalitarian allocations, in order to avoid certain providers leaving the market and preventing oligopolistic situations (Murillo, López, Muñoz, & Busquets, 2012). On another context, for example, when attending health care organization incidences, the company may need to favor reliable allocations more than efficient ones (Ramchurn et al., 2009). Therefore, there is a need of studying the different attributes that should be taken into account in an auction, and understanding their role in the allocation properties.

To preserve truthful bidding in scenarios where money is not the only key issue (e.g. pay-per-click advertising), new auction mechanisms have been designed. An example of them are position auctions which are used in advertisement scenarios, such as Google Ads (Varian, 2007), where auctions occur in a repetitive way among a set of bidders. Due to repetition, auctioneers are able to qualify bidders according to some attributes. When clearing the auction, the auctioneer unifies the qualification attributes and the received bids in order to determine the winner. By doing so, this kind of auctioneer guarantees incentive compatibility. Nonetheless, the attributes that this mechanism deals with are not manipulable by bidders, as they are provided by the auctioneer itself.

In this paper we are going one step forward, by distinguishing different types of attributes regarding the goods to be sold (verifiable and unverifiable attributes), in addition to attributes qualifying the buyers. Verifiable attributes are the ones that can be checked upon the reception of the goods, as for example, its quality. Conversely, unverifiable attributes cannot be checked due to its subjectiveness, such as the price of the object (a good can be expensive or not depending on the utility provided to the buyer). The third kind of attributes, auctioneer provided attributes, concerns information about the actual goods, but also regarding the bidders (e.g. the reliability of a bidder).

Therefore, the first contribution of the paper is the classification of attributes according to the ownership (bidder or auctioneer provided) and their types (verifiable or unverifiable). Second, and using attributes of this categorization, we design a new multi-attribute auction mechanism that extends previous multi-attribute auctions, taking ideas from position auctions. Our mechanism lets the auctioneer obtain some desirable properties as incentive compatibility, so that the bidders are encouraged to bid with the true properties they can provide (duration, quality of service), at the best price. Third, the consideration of attributes regarding the bidders’ behavior, enables the configuration of different social welfare approaches (utilitarian, egalitarian) and opens up a range of possibilities for the incorporation of other auction properties, such as robustness. To illustrate and test the auction mechanism, we simulate an industrial environment where different agents auction services, which must be carried out by external service providers, while trying to obtain a service at a reasonable price, time and quality.

This paper is organized as follows: first we introduce some basic concepts regarding auctions. In Section 3 we present a brief state of the art of multi-attribute auctions. In Section 4 we analyze and classify the different attributes that can take part in a multi-attribute auctions. Based on this classification, in Section 5 we present a multi-attribute auction mechanism to enact different auction properties. Then, we empirically test the mechanism in a simulator to illustrate its working, and compare it with some baseline allocation methods. Finally, in Section 7 we present the conclusions of our work and point to possible future lines of research.

Section snippets

Auctions background

An auction is a method for buying and selling goods using a bid system in which the winning bid obtains the auctioned goods (Krishna, 2002). In manufacturing domains, as we are used to, auctions often follow a reverse schema; the auctioneer wishes to buy a task to be done, and the bidders sell their working capacity at a given price. For the sake of clarity, we will follow this approach in this paper.

According to the reverse auction approach, the main steps of an auction are as follows:

  • 1.

    Call for

Related work

Multi-attribute auctions have been used in the electronic advertisement markets (Athey & Ellison, 2011), for instance Varian (2007) proposes to include an extra attribute provided by the auctioneer itself, not by the bidder. This approach is similar to trust-based approaches as the one presented by Ramchurn et al. (2009) where, to enhance robust allocations, agents describe other agents’ reliability using a trust parameter, and include this parameter in the winner determination phase. In the

Attribute typologies in multi-attribute auctions

In multi-attribute auctions, the item which is sold is defined by a set of attributes, in addition to the price, that it is determined in the auction process. Multi-attribute auctions may involve many attributes regarding the item, as for example, the size and quality requested by the buyers, and the price offered by the sellers. Moreover, in some uni-attribute auctions the auction is cleared by taking into account other attributes besides the ones offered by the bidder. For instance,

VMA2: multi-attribute auction based on verifiable, unverifiable, and auctioneer-provided attributes

Using the classification of attributes provided in the previous section, we present the VMA2 multi-attribute auction mechanism. VMA2 follows a Vickrey-based approach to guarantee the incentive compatibility of the mechanism (Pla, López, & Murillo, 2013).

Let ( a 1 , , a n ) be the bidders, and a 0 the auctioneer, who wants to buy an item it. This time, the auctioneer requires, along with the price, additional attributes from the item, for example, price and quality. Each agent provides one bid b i

Experimentation

To illustrate and test the behavior of the presented mechanism, we simulated the processes and the services required by a company during a certain time period under different scenarios.

Conclusions

When using auctions for resource allocation in manufacturing and service problems, other attributes than price should be taken into account. In such domains, having a cheap price for a resource to deploy a task could be interesting, but so is assuring that tasks will be delivered on time, or at the required quality. Therefore, the different steps of an auction, bidding, winner determination, and payment, should be determined according to several attributes, in what is called multi-attribute

References (44)

  • M.-J. Bellosta et al.

    Preference-based english reverse auctions

    Artificial Intelligence

    (2011)
  • D.H. Shih et al.

    MoRVAM: A reverse Vickrey auction system for mobile commerce

    Expert Systems with Applications

    (2007)
  • H.R. Varian

    Position auctions

    International Journal of Industrial Organization

    (2007)
  • Athey, S., & Ellison, G. (2011). Position auctions with consumer search. Quarterly Journal of Economics 126 (3),...
  • Bartal, Y., Gonen, R., & Nisan, N. (2002). Incentive compatible multi unit combinatorial auctions. In TARK 03 (pp....
  • M.-J. Bellosta et al.

    A unified framework for multiple criteria auction mechanisms

    Web Intelligence and Agent Systems

    (2008)
  • Bichler, M., Kaukal, M., & Segev, A. (1999). Multi-attribute auctions for electronic procurement. In Proceedings of the...
  • Y.-K. Che

    Design competition through multidimensional auctions

    The RAND Journal of Economics

    (1993)
  • Y. Chevaleyre et al.

    Issues in multiagent resource allocation

    Informatica

    (2006)
  • E. David et al.

    Protocols and strategies for automated multi-attribute auctions

  • U. Endriss et al.

    Resource allocation in egalitarian agent societies

  • Fu, H., Jordan, P. R., Mahdian, M., Nadav, U., Talgam-Cohen, I., & Vassilvitskii, S. (2012). Ad auctions with data. In...
  • J.L. Gastwirth

    The estimation of the lorenz curve and gini index

    The Review of Economics and Statistics

    (1972)
  • B.P. Harrenstein et al.

    A qualitative vickrey auction

  • K. Hoesch-Klohe et al.

    Towards green business process management

  • V. Krishna

    Auction theory

    (2002)
  • Lee, J., & Szymanski, B. (2005). A novel auction mechanism for selling time-sensitive e-services. In IEEE conference on...
  • D. Lehman et al.

    The winner determination problem

    (2006)
  • Lematre, M., Verfaillie, G., Fargier, H., Lang, J., Bataille, N., & m. Lachiver, J. (2003). Equitable allocation of...
  • MacKie-Mason, J. K., & Varian, H. R. (1994). Generalized vickrey...
  • T. Mahr et al.

    Multi-attribute vickrey auctions when utility functions are unknown

  • J. Murillo et al.

    Fairness in recurrent auctions with competing markets and supply fluctuations

    Computational Intelligence

    (2012)
  • Cited by (28)

    • A multi-attribute double auction and bargaining model for emergency material procurement

      2022, International Journal of Production Economics
      Citation Excerpt :

      In the mechanism improvement, Sadaoui and Shil (2016) proposed a multi-round, first-score and semi-sealed multi-attribute reverse auction mechanism considering the hard constraints and qualitative preferences of the buyer's actual needs. Pla et al. (2014) divided the attributes of goods or services into verifiable attributes, unverifiable attributes and attributes provided by the auctioneer. They proposed a reverse multi-attribute auction mechanism based on Vickrey, and analyzed the role of different types of attributes in the auction.

    • Evaluating the Potential Contribution of Multi-Attribute Auctions to Achieve Agri-Environmental Targets and Efficient Payment Design

      2020, Ecological Economics
      Citation Excerpt :

      Among others, key features are the type of scoring rule (i.e. that defines a winning bid) and the use of a multi-attribute format (Hansen, 1988; Che, 1993). Asker and Cantillon (2008) and Pla et al. (2016) show that a multi-attribute auction allows farmers (bidders) to bid on multiple attributes besides the price (e.g. type and quality of the auctioned good, measures of input reduction, actions to reach expected outcomes, etc.), and the public agency (auctioneer) to deal with several attributes in the winners' selection processes (Schilizzi, 2017). Although auctions have received considerable interest, there are only a few studies - in particular in the field of agri-environmental schemes (AESs) - that analyse the opportunity to design multi-attribute tenders within a fixed budget mechanism1 (Glebe, 2008, 2013; Iho et al., 2014).

    View all citing articles on Scopus
    View full text