报告主题：Trust and Contracts: Empirical Evidence

主讲嘉宾：姚加权教授

时      间：2021年4月19日下午14：30-17：30

地      点：腾讯会议705980640

摘要 :

Trust between parties should drive contract design: if partieswere suspicious about each others’ reaction to unplanned events, they mightagree to pay higher costs of negotiation exante to complete contracts. Using aunique sample of U.S. consulting contracts and a negative shock to trustbetween shareholders/managers (principals) and consultants (agents) staggeredacross space and over time, we find that lower trust increases contractcompleteness. Not only the complexity but also the verifiable states of theworld covered by contracts increase after trust drops. The results hold forseveral novel text-analysis-based measures of co-ntract completeness and do notari-se in falsification tests. At the clause level, we find that non-competeagreements, confid-entiality, indemnification, and termination ru-les are themost likely clauses added to contra-cts after a negative shock to trust andthese additions are not driven by new boilerplate contract templates. Theseclauses are those whose presence should be sensitive to the mutual trustbetween principals and agents.

嘉宾简介 :

姚加权，暨南大学金融学教授，博导。研究方向包括公司金融、劳动与金融、实证资产定价、数字经济、文本分析和机器学习，论文发表于Review of Financial Studies、ManagementScience、Journal of Financial and Quantitative Analysis和《管理科学学报》等国内外权威期刊，以及计算机顶会WWW2018和KDD2020。

报告主题：Selecting mutual funds from the stocks theyhold: a machine learning approach

主讲嘉宾：李斌教授

时      间：2021年4月19日下午14：30-17：30

地      点：腾讯会议705980640

摘要 :

We select mutual funds in real time by combining individual fundholdings and a large number (94) of stock characteristics to compute fund-levelexposures to characteristics on the ba-sis of the stocks they hold. The majorityof funds are exposed---both positively and negatively---to approximately 40-50characteristics. In addition, fund performance is non-linearly rela-ted to fundcharacteristics and their interactions. This feature proves important when we predictfund performance, as machine learning methods such as boosted regression trees(BRTs) significantly outperform standard linear frameworks. Our BRT-generatedforecasts enco-mpass the ones generated by the predictors of mutual fundperformance that have been proposed in the literature so far.

嘉宾简介 :

李斌，武汉大学经济与管理学院教授、博士生导师，研究方向为金融科技和投资管理等。论文发表于《Journal of Accounting Research》《ArtificialIntelligence》《Journal of Machine Learning Res-earch》《管理科学学报》《ICML》《IJCAI》等国内外权威期刊及顶会。担任金融研究中心主任、金融系副主任，兼任中国金融学年会理事、金融科技研究与教育五十人论坛成员等。

报告主题：Machine Learning in Forecasting ChineseStock Market Recessions

主讲嘉宾：马勇教授

时      间：2021年4月19日下午14：30-17：30

地      点：腾讯会议705980640

摘要 :

We investigate the predictive performance of various machinelearning classifiers in forecasting China's stock market recessions using abroad set of aggregate market-level factors. The em-pirical analysisdemonstrates that machine learning tools present the statistical predictiveperformance in comparison with the traditional logistic regression model widelyused in the literature, especially neural networks. The improvement inout-of-sample predictive ability translates large econ-omic benefits toinvestors after taking transaction costs into account. Moreover, our analysis revealsthat the most influential market-level features are based on price trends andmarket volatility.

嘉宾简介 :

马勇，湖南大学金融与统计学院教授，博士生导师，应用金融系主任。入选湖湘青年英才支持计划和湖南省121创新人才培养工程。在《Journal of Futures Markets》《 Quantitative Finance》《International Reviewof Finance》《管理科学学报》《中国管理科学》等国内外重要期刊发表论文30余篇。承担国家自然科学基金项目2项和省部级项目5项。