报告题目：Do Asset Prices Help Predicting Inflation? Evidence Based on Individual Stock Prices and Machine Learning Algorithm
报告人：黄乃静 中央财经大学
报告时间：2022年12月14日(周三) 16:00-17:30
腾讯会议ID：337 514 023
内容摘要
Asset prices reflect the discounted expected value of future earnings, which suggests that they should be good predictors of inflation. Empirical evidence, however, is mixed and puzzling. In this paper, we extract information from the individual stock prices, a prototype of asset price, combining with various machine learning methods, to revisit the debate on the role of asset prices in predicting inflation. Based on the data of the U.S., U.K. and Japan, and multiple sample periods experiments, our results suggest: Firstly, using information from individual stock prices significantly improves the predictive accuracy of inflation forecasts than univariate benchmark models (ARIMA, RW, UCSV) in general. The improvement is even more remarkable in medium-to-long term forecasting, and in high inflation volatility periods. Secondly, compared with major aggregate asset price indices as predictors, the individual stock prices have greatly improved inflation forecasting performance. Thirdly, we further compare the predictive power of the widely used data set in inflation forecasting, namely the FRED-MD, with individual stock prices, which consolidate our major result. Lastly, shrinkage estimators, such as LASSO family models and Ridge Regression model, are more successful relative to other non-linear machine learning algorithms in general.
主讲人简介
黄乃静，中央财经大学经济学院长聘副教授，博士生导师，国民经济系主任。2015年取得美国波士顿学院经济学博士学位。以金融计量经济学和货币经济学为主要研究方向，先后在《管理科学学报》《经济学动态》《系统工程理论与实践》《中国软科学》Emerging Market Finance and Trade, Applied Economic Letters等国内外学术刊物发表多篇论文。获得2017年厦门大学金融工程与量化金融学术会议最佳论文奖、2014年美国国家科学基金会旅行奖励（US National Science Foundation Travel Grant）等学术奖项。主持国家自然科学基金，并参与多项国家自科重点、国家社科重大基金。
