Simultaneous First-Price Auctions with Preferences over Combinations

拍卖之间的相互作用如何影响拍卖市场的经济绩效？拍卖-从经济上讲，任何涉及结构化、竞争性投标的机制-提供了一种简单、透明的机制，用以分配商品和服务，特别是在投标人对复杂或异质对象可能有特殊偏好的环境中。因此，拍卖在具有实际和政策意义的广泛应用中无处不在：eBay等在线平台、住房和汽车市场、公司控制市场以及公共机构采购基本商品和服务的市场，等等。这反过来又激发了对拍卖设计和绩效的广泛学术研究，产生了对参与拍卖市场的私人和公司以及对受托监督公共采购的公共代理人和机构都有意义的见解，虽然文献在了解拍卖市场方面取得了巨大进展，但迄今为止的实证研究主要集中在两大范式上：独立的单物品拍卖和多个相同物品的多单位拍卖。不幸的是，这些范例很少提到应用程序中广泛遇到的问题：决策者应该如何分配许多不同的对象，特别是当投标人将对象视为补充或替代品时？这个问题的一个常见解决方案是运行许多单独的密封竞标拍卖，竞标者同时竞争许多物品。基于这种设计的拍卖在公共采购应用中广泛存在，包括美国大多数州的公路维护合同市场、蒙特利尔的清雪合同市场、日本的回收服务市场、瑞典的清洁服务市场、美国外大陆架的石油和钻探权市场以及意大利的废物管理市场等。在这种情况下，众所周知，对多个对象的投标可能会产生“暴露问题”，这大大扭曲了投标人的战略激励，可能导致比标准单对象模型预测的更糟糕的收入和效率结果。然而，尽管在广泛的政策相关应用中普遍存在同步密封投标拍卖，但实际上对这种风险暴露问题的范围一无所知。在这项研究中，我们研究如何在应用程序，如高速公路承包拍卖之间的相互作用影响的收入和经济价值创造的采购市场。为此，我们首先开发了一个经验模型的投标同时密封投标拍卖，据我们所知，在文献中的第一个。至关重要的是，这个模型支持的程度，投标人认为拍卖对象的补充或替代品，一个关键参数的政策设计在多对象拍卖，目前几乎一无所知的估计。我们的模式适用于广泛的公共采购环境，包括上述环境。为了说明我们提出的方法，但是，我们专注于一个单一的应用程序，其数据是广泛可用的：密歇根州交通部（MDOT）拍卖的高速公路建设和维护合同。我们应用我们的模型来估计投标人的偏好和跨项目的协同作用在MDOT市场，初步结果表明，投标人认为小项目的补充和大项目的替代品。在这些估计的基础上，我们的目标是提供新的见解，对两个对象的重大政策利益：在何种程度上，目前的做法最大限度地提高价值，最大限度地减少公共部门的成本在MDOT市场，以及是否其他可能的多对象格式（如同时上升拍卖）可以改善在这个市场的结果。因此，我们的目标是提供新的见解，多物品拍卖的最佳设计，帮助解决一个重要的一类政策问题，在广泛的采购环境。

How does interaction between auctions affect economic performance of auction markets? Auctions - economically speaking, any mechanism involving structured, competitive bidding - provide a simple, transparent mechanism by which to allocate goods and services, particularly in environments where bidders have potentially idiosyncratic tastes over complex or heterogeneous objects. Consequently, auctions are ubiquitous in a wide range of applications of practical and policy interest: online platforms such as eBay, housing and auto markets, corporate control markets, and markets by which public agencies procure essential goods and services, among others too numerous to name. This in turn has motivated extensive academic research on auction design and performance, yielding insights relevant both for private individuals and companies participating in auction markets and for public agents and agencies entrusted with oversight of public procurement.While the literature has made tremendous progress in understanding auction markets, empirical research to date has focused on two broad paradigms: standalone single-object auctions and multi-unit auctions for many identical objects. Unfortunately, these paradigms say little about a problem widely encountered in applications: how should a policymaker allocate many different objects, especially when bidders view objects as complements or substitutes? One common solution to this problem is to run many separate sealed-bid auctions, with bidders competing for many objects simultaneously. Auctions based on this design are widely seen in public procurement applications, including markets for highway maintenance contracts in most US states, snow-clearing contracts in Montreal, recycling services in Japan, cleaning services in Sweden, oil and drilling rights in the US Outer Continental Shelf, and waste management in Italy among others. In such cases it is known that bidding for multiple objects can create "exposure problems" which substantially distort bidders' strategic incentives, potentially leading to much worse revenue and efficiency outcomes than would be predicted by the standard single-object model. Yet despite the prevalence of simultaneous sealed-bid auctions in a wide range of policy-relevant applications, virtually nothing is known about the scope of such exposure problems in practice. In this research, we study how interaction between auctions in applications such as highway contracting affects revenue and economic value created in procurement markets. Toward this end, we first develop an empirical model of bidding in simultaneous sealed-bid auctions, to our knowledge the first such in the literature. Crucially, this model supports estimation of the degree to which bidders view auctioned objects as complements or substitutes, a critical parameter for policy design in multi-object auctions about which virtually nothing is presently known. Our model is applicable in a wide range of public procurement settings, including those listed above. To illustrate our proposed methodology, however, we focus on a single application for which data is widely available: highway construction and maintenance contracts auctioned by the Michigan Department of Transportation (MDOT). We apply our model to estimate bidder preferences and cross-project synergies in the MDOT market, with preliminary results suggesting that bidders view small projects as complements and large projects as substitutes. Building on these estimates, we aim to provide new insight on two objects of substantial policy interest: to what degree current practices maximize value and minimize public-sector costs in the MDOT market, and whether other possible multi-object formats (e.g. simultaneous ascending auctions) could improve outcomes in this market. We thus aim to provide new insight on optimal design of multi-object auctions, helping resolve an important class of policy questions arising in a wide range of procurement contexts.