慶応義塾大学産業研究所プロジェクト
代表者：大津泰介
(1) 社会科学における因果推論と構造分析の手法開発と応用 (New Methodology for Causal Inference and Structural Analysis in Social Science)
社会科学での実証分析においては，実験計画法の発想を取り入れた因果推論と，最適化行動に基づいた理論モデルによる構造分析という二つの大きな潮流がある．このプロジェクトでは，それぞれの手法における現在の課題に取り組み，新しい分析手法を提案し応用することを目指す．具体的には，より大規模かつ複雑なデータのもとでの因果の識別と推測，より現実的な仮定のもとでの構造パラメータの識別と推測を扱う．
In social science, there are two major approaches for empirical studies: causal inference by adapting notions of randomized experiments, and structural analysis by utilizing theoretical models based on agents’ optimization behaviors. In this project, I plan to develop novel methodologies for both approaches on applied studies. In particular, this project develops identification and inference methods for causal objects under larger and more complex data environments, and for structural parameters under more flexible assumptions.
(2) Big Data時代の方法論開発と応用 (New Methodology for Big Data Analysis in Social Science and Its Applications)
近年の情報技術の飛躍的な発展に伴い，科学の諸分野において多様な形態を持つ膨大なデータがえられるようになり，データ分析の方法論においても様々な取り組みがなされつつある．このプロジェクトでは，このようなBig Data時代の到来を踏まえ，帰納と演繹の両側面から新しいデータ分析の方法論を開発し，現実のデータに応用することを目指したい．具体的には下記の問題について取り組む予定である．(1) 社会科学における因果分析と政策評価．(2)社会科学における機械学習の手法を用いたBig Dataの分析．(3)ネットワーク・データの分析．(4)産業のゲーム理論的構造モデルと分析．(5)オークションの実証分析．(6)動学的構造モデルの分析.(1)-(3)では帰納的な手法を，(4)-(6)では演繹的な手法を採用し，前者は労働・公共分野，後者は産業分野への応用を目的としている．
With the dramatic development of information technology in recent years, enormous amount of data in various forms has become available in diverse fields of science, and a great variety of efforts are being made to develop methodologies for such big data analyses. Prominent examples in social science include network data (e.g. Facebook and Twitter), consumer behavior data (e.g. Amazon and eBay), and field experimental data for socio-economic behaviors, among many others. In this project, based on the arrival of such Big Data era, I plan to develop new methodologies for data analysis from both inductive and deductive perspectives, and apply them to real data. Specifically, I plan to investigate the following problems: (i) Causal analysis and policy evaluation in social science, (ii) Big data analysis by machine learning methods in social science, (iii) Analysis on network data, (iv) Game theoretic model and structural analysis for industries, (v) Empirical analysis on auction data, and (vi) Analysis on dynamic structural models. i)-iii) employ inductive approaches, and I plan to apply for labor or public fields. iv)-vi) adopt deductive approaches, and I plan to consider applications for analyses of industries.
Published Papers
1. Conditional GMM estimation for gravity models (with Masaya Nishihata), Economics Bulletin (2020), 40, 1106-1111.
2. Kolmogorov-Smirnov type test for generated variables (with Go Taniguchi), Economics Letters (2020), 195, 109401.
3. Sample sensitivity for two-step and continuous updating GMM (with Rikuto Onishi) Economics Letters (2021), 198, 109685.
4. Inference on incomplete information games with multi-dimensional actions (with Hideyuki Tomiyama), Economics Letters (2022), 215, 110440.
5. Inference on conditional moment restriction models with generated variables (with Ryo Kimoto), Economics Letters (2022), 215, 110454.
6. Empirical likelihood inference for Oaxaca-Blinder decomposition (with Shiori Tanaka) forthcoming in Economics Letters.
7. Empirical likelihood inference for monotone index model (with Mengshan Xu and Keisuke Takahata), forthcoming in Japanese Journal of Statistics and Data Science.
8. On large market asymptotics for spatial price competition models (with Keita Sunada), forthcoming in Economics Letters.
Working Papers
1. Finite-population inference via GMM estimator (with Haruo Kakehi), revise and resubmit for Journal of Business & Economic Statistics.
2. Optimal testing in a class of nonregular models (with Yuya Shimizu)
Works in Progress
• Isotonic nonparametric instrumental variable regression (with Mengshan Xu and Kazuhiko Shinoda)
• MPEC approach for empirical likelihood (with Kohei Hayashida)
• Simulated maximum likelihood for high-dimensional models (with Yusuke Inami)
• Empirical likelihood for geographically weighted regression (with Yuki Shibatsuji)
• Testing number of regimes in regime switching models with endogenous latent factor (with Naoya Nagasaka)
• Nonparametric estimation for auction models under shape constraints (with Kai Maeda)
• Density ratio approach for DID models (with Noriaki Okamoto)
• Nonparametric identification of production functions (with Ryuki Kobayashi)
• Semiparametric estimation of rank ordered dependent variable models (with Kei Kitagawa)
参加者 (過去の慶応義塾の計量経済学演習での共同研究者も含む)
• 泉隆一郎（Wesleyan University, Assistant Professor）
• 稲見勇輔（慶應義塾大学修士課程）
• 太田悠太（慶應義塾大学助教）
• 大西陸仁（University of Texas-Austin, PhD student）
• 岡本憲曉（慶應義塾大学博士課程）
• 筧悠夫（University of Wisconsin-Madison, PhD student）
• 北川慶（慶應義塾大学修士課程）
• 木本遼（Pennsylvania State University, PhD student）
• 小林流基（慶應義塾大学助教）
• 篠田和彦（名古屋大学講師）
• 柴辻優樹（慶應義塾大学博士課程）
• 清水祐弥（University of Wisconsin-Madison, PhD student）
• 杉浦航（University of Houston, PhD student）
• 砂田啓太（University of Rochester, PhD student）
• 高畑圭佑（Synspective）
• 高梨耕作（理研）
• 田辺潤一郎（慶應義塾大学博士課程）
• 谷口豪（三菱総研）
• 高野昌也（Monash University, Post-doctoral fellow）
• 田中詩織（Pennsylvania State University, PhD student）
• 富山英之（Pennsylvania State University, PhD student）
• 長坂直哉（Indiana University, PhD student）
• 西畑壮哉（三菱UFJリサーチ&コンサルティング）
• 林田光平（UCSD, PhD student）
• 前田快（University of York, PhD student）
• マクリン謙一郎（Temple University, Assistant Professor）
• 頼慶泰（University of Mannheim, Assistant Professor）