講座題目：Expected Returns and Foundation Models of Language

講座時間：2022年12月7日（星期三）上午10:30-11:30

講座方式：騰訊會議ID：916-518-287

密碼：221207

講座鏈接：https://meeting.tencent.com/dm/6OM94M1NaSl3

主 講 人：

修大成，現(xiàn)任芝加哥大學(xué)布斯商學(xué)院計量經(jīng)濟(jì)學(xué)和統(tǒng)計學(xué)教授。修大成教授的研究興趣包括：設(shè)計統(tǒng)計方法并將其應(yīng)用于金融數(shù)據(jù)，來研究數(shù)據(jù)中所反映的經(jīng)濟(jì)學(xué)含義。早期的研究涉及風(fēng)險測量和投資組合管理，包括高頻數(shù)據(jù)和衍生產(chǎn)品的計量經(jīng)濟(jì)學(xué)模型。目前的研究主要集中在設(shè)計機器學(xué)習(xí)方法來解決資產(chǎn)定價領(lǐng)域的大數(shù)據(jù)問題。他已經(jīng)在Econometrica，Journal of Political Economy，Journal of Finance，Review of Financial Studies，Journal of the American Statistical Association，以及 Annals of Statistics等期刊上發(fā)表了研究成果。修教授擔(dān)任Journal of Financial Econometrics的共同主編，是Review of Financial Studies，Management Science，Journal of Econometrics，Journal of Business & Economic Statistics，Journal of Applied Econometrics，the Econometrics Journal以及Journal of Empirical Finance等期刊的副主編。因其研究，他獲得了多項榮譽，包括金融計量經(jīng)濟(jì)協(xié)會會士、Journal of Econometrics會士、Swiss Finance Institute Outstanding Paper Award、AQR Insight Award和歐洲金融協(xié)會年會最佳會議論文等。

講座簡介：

We extract contextualized representations of news text to predict returns using the state-of-the-art foundation models in natural language processing. The contextualized representation of news reflects its content more accurately than the bag-of-words representation prevalent in the literature. In particular, the latter approach is more prone to errors when negation words appear in news articles. Moreover, we provide polyglot evidence on news-induced return predictability in 16 international equity markets with news written in 13 different languages. Information in newswires is assimilated into prices with an inefficient delay that is broadly consistent with limits-to-arbitrage, yet can be exploited in a real-time trading strategy. Furthermore, a trading strategy that exploits fresh news in the form of news alerts leads to even higher Sharpe ratios.