| Title | 海道宏明先生連続講義 CREPE Lecture Series by Prof. Hiroaki Kaido "Inference for Incomplete Economic Models: Robustness, Efficiency, and Testability" |
|---|---|
| Date | 2019年7月9日(火）13:00-14:45、7月10日（水）13:00-16:40 13:00-14:45 on July 9, 13:00-16:40 on July 10, 2019 |
| Venue | 東京大学学術交流棟（小島ホール）２階・小島コンファレンスルーム
[地図] Kojima Conference Room, Economics Research Annex (Kojima Hall), University of Tokyo [MAP] |
| Lecturer | 海道宏明（Hiroaki Kaido, Boston University) |
| Course description | An economic model is said to be incomplete if it makes set-valued predictions. Examples can be found in various areas of economics including but not limited to the models of consumer demand, labor supply, market entry, auction, and network formation. In many of these settings, the researcher's understanding on how an outcome gets selected is limited. This raises a number of questions surrounding inference based on data generated by incomplete models. In this lecture series, after reviewing recent developments in the econometrics literature, I focus on addressing the following questions. 1) Suppose we face competing hypotheses on structural parameters. When are data useful for testing them? 2) How much can we learn from data about the underlying structure while staying agnostic about the selection mechanism? What is the price to pay by being robust? 3) How can we compare statistical procedures in such settings? In short, they are about "Testability (or falsifiability)" of theories, "Robustness", and "Efficiency (or optimality)" of statistical procedures in environments where the researcher has access to data and an incomplete theory. The lectures will introduce tools for analyzing incomplete models, develop a framework for formalizing the concepts above, and discuss available results and future directions. Connections to other areas of economics and statistics, in particular decision theory and robust statistics, will also be discussed. |
| Information | 英語での講義となります（Lecture in English） |