Search Theory in Economics

这个项目扩展了搜索理论的前沿，以阐明三个经济问题。第一个问题是，我们应该如何期望大学的拒绝率随着能力的不同而变化，并询问它们是否是衡量优秀的有用指标。初步研究表明，在最好的大学里，拒读率可能与质量指标相反。第二个应用程序探讨了最优清算的经济学，例如在20世纪90年代或2008年著名的雷曼兄弟破产之前的清盘信托公司。早期的研究表明，在利率较低的情况下，最优的销售策略的进展速度要慢得多，这证实了艾伦•格林斯潘（Alan Greenspan）提出的一个观点。第三部分对网络搜索的经济学进行了理论分析，比如发生在亚马逊等网站上的搜索；早期的研究发现包括，卖家不一定会发现，打造一个信息量最大的搜索引擎是最有利可图的。本课题在搜索理论和有向搜索模型中分别进行了两个动态规划研究。第一个项目将拍卖理论的见解整合到一个定向搜索问题中，该问题是由期刊提交和大学申请过程驱动的。随着期刊或学院排名的提高，其标准也会提高，申请者也会提高。但拒绝率一定会上升并不明显，事实上，根据完整的信息，早期的结果表明，拒绝率不会上升。这两个动态规划练习探讨了两个固有的非平稳搜索模型。第一个问题是，当面对愿意以特定价格购买部分资产的周期性买家到来时，人们如何最优地平仓一项资产中的大量头寸。在许多关键结果中，最重要的是“可选性收益递减”，即价值函数在位置上是凹的，但边际价值是凸的。第二个动态规划练习探讨如何在许多排序选项中进行最佳搜索。这种情况出现在网络搜索中，这个模型易于处理，足以产生许多新颖可用的比较静态。

This project extends the frontier of search theory to shed light on three economic problems. The first asks how we should expect rejection rates at colleges to vary with caliber, and inquires whether they are informative measures of excellence. Preliminary research suggests that rejection rates might be reverse quality measures at the very best colleges. The second application explores the economics of optimal liquidation, such as transpired with the Resolution Trust Corporation in the 1990s or prior to the famous bankruptcy of Lehman-Brothers in 2008. Early work reveals, among other things, how an optimal selling strategy proceeds far more slowly at lower interest rates, formalizing an insight publicized by Alan Greenspan. The third develops a theoretical analysis of the economics of web search, such as occurs in web sites like Amazon; among early findings are that the seller need not find it most profitable to make a maximally informative search engine. The project pursues two dynamic programming projects in search theory, and one directed search model. The first project incorporates insights from auction theory into a directed search problem that is motivated by the journal submission and college application processes. As the rank of the journal or college rises, its standards rise and its applicants improve. But it is not obvious that rejection rates must rise and in fact, with complete information, early results suggest that they do not. The two dynamic programming exercises explore two inherently nonstationary search models. The first is how does one optimally unwind a large position in an asset when facing periodic arrivals of buyers willing to buy portion at a specified price. Among many key results, most important is the "decreasing returns to optionality" that the value function is concave in the position - but the marginal value is convex. The second dynamic programming exercise explores how one optimally searches among many ranked options. This arises in web search, and this model is tractable enough to produce many novel usable comparative statics.