Monte Carlo Simulation - Applications in Econometrics and Economic Modeling

本文主要研究两个问题：(1)第一次价格拍卖中的非对称投标人和投标人联盟；(2)动态潜变量和非均衡模型。这些主题的共同特点是在分析上难以解决，因此需要数值技术进行研究。硬件和软件方法使这些问题的计算方法变得可行。第一个主题很重要的一个原因是，联邦政府在大宗商品的销售和采购中广泛使用第一价格拍卖。然而，人们对在这些拍卖中结成联盟的可行性知之甚少。如果所有投标人都是相同的，并且如果投标人联盟是可能的，那么他们的投标人的出价来源与非串通投标人的出价来源根本不同。因此，分析投标人串通的第一步是理解非合作行为的异类投标人的行为。将获得关于第一价格拍卖模型的解决方案，并利用结果来理解和描述投标人在何种情况下可以有利可图地串通。第二个主题很重要的一个原因是，潜变量技术可以用于估计美国货币需求的模型，其中许多重要变量本质上是潜在的(或未观察到的)和动态的。这些变量的例子是预期和永久收入，通常还有衡量货币本身的指标。研究人员已经开发出一种对未观察到的变量进行数字模拟的技术，允许在不删除潜在变量的情况下进行推断。

This research focuses on two topics: (1) Asymmetric bidders and coalitions of bidders at first price auctions and, (2) Dynamic latent variables and disequilibrium models. The topics share the characteristic of being analytically intractable and, therefore, requiring numerical techniques for investigation. Hardware and software methods have made a computational approach to these problems feasible. One reason the first topic is important is that the federal government uses first price auctions extensively for both the sale and procurement of commodities. Nevertheless, very little is understood about the feasibility of coalitions forming at these auctions. If all bidders are identical and if coalitions of bidders are possible, then their bids stem from a source that is fundamentally different from the source for the bids of non-colluding bidders. Consequently, a first step in analyzing bidder collusion is to understand the behavior of heterogeneous bidders who act non-cooperatively. Solutions will be obtained for first price auction models and the results utilized to understand and characterize the circumstances under which bidders can profitably collude. One reason the second topic is important is because the latent variable technique can be used in estimating models of the U.S. money demand where a number of important variables are inherently latent (or unobserved) and dynamic. Examples of these variables are expected and permanent income and often the measure of money, itself. The investigators have developed a technique for numerically simulating the unobserved variables that permits inferences to be drawn without removing the latent variables.