Essays on dynamic procurement

Abstract

This thesis studies the impact of uncertainty and its sequential resolution on the design of procurement mechanisms and, in particular, auctions. Procurement contracts constitute a considerable fraction of output in developed economies. These contracts are typically long-term agreements between a buyer (e.g., the government) and a seller (a supplier of a good or service). This long-term nature of procurement contracts makes renegotiation after the resolution of uncertainty more likely. As a consequence, analysing how these ex-post dynamics affect the ex-ante design of procurement mechanisms becomes very important. The first chapter of this thesis provides a survey and critical review of the theoretical literature on dynamic procurement. It provides a summary of the seminal contributions studying the optimal design of procurement mechanisms and reviews them in light of recent advancements in the theory of dynamic procurement. We describe the issue of procurement from a mechanism design standpoint, the dominant scoring auction, as well as the problems of adverse selection and corruption. The second part of this review describes different channels through which dynamic considerations impact this environment. In particular, we consider issues such as renegotiation and ex-post adaptions, the non-contractible nature of some relationships, repeated interactions, as well as the unique aspects relevant for large and complex projects. In the final part of the first chapter, we explore how the findings from the literature on dynamic procurement help us cast new light on the key static ideas from the seminal papers. The second chapter of this thesis explores the effect of cost-uncertainty on the optimal procurement mechanism. We make three main contributions to the literatures on mechanism design and procurement. First, contributing to the research on dynamic mechanism design, we derive the optimal mechanism for a multidimensional environment with sequential generation of private information. Unlike in the one-dimensional analysis of Eső and Szentes (2007), in our multidimensional dynamic model, the second stage private information does affect the agents’ rent. Second, contributing to the literature on procurement auctions, we show that, unlike in Che (1993), the scoring auction is no longer able to implement the optimal mechanism if contracts are renegotiated in response to the realisations of cost shocks. Finally, the paper introduces a stylised model of procurement with renegotiation that matches several features of observed procurement auctions. The third and final chapter of this thesis addresses a question that naturally emerges from the results of the second chapter. Scoring auctions with renegotiation are not the optimal procurement mechanism for the buyer. However, despite their sub-optimality, scoring auctions are the predominant mechanism used in practice. This chapter addresses the question of optimality within the class of scoring auctions: What is the optimal scoring rule? Furthermore, we quantify the gap between the overall optimal mechanism and the optimal scoring auction in terms of buyer payoffs. In addressing these questions, this chapter introduces a methodology that imposes further restrictions on a direct mechanism optimisation problem to find the optimum among a class of sub-optimal mechanisms. We further extend the result of the second chapter to a discrete type-space. This type-space facilitates an analytic solution to the optimisation problem for the restricted mechanism design problem. Finally, we introduce an auction with reserve-scores and show that it is the optimal scoring auction with renegotiation for discrete type-spaces.