登入選單

返回Google圖書搜尋

Two-Stage Learning of a Systemic Risk Factor

Wenzhi Wang

出版	SSRN, 2019
URL	http://books.google.com.hk/books?id=dEj6zgEACAAJ&hl=&source=gbs_api

註釋

Motivated by multi-dimensional sources of systemic risk in the economy and the low frequency nature of the observed empirical proxies for some of these, we propose a two-stage learning procedure to construct one better high-frequency (i.e., daily) systemic risk factor. In the first stage, we use a Kalman-Filtering approach to synthesize the information about systemic risk contained in 19 different monthly proxies for systemic risk and produce one better monthly Bayesian risk factor. The low frequency (i.e., monthly) Bayesian factor can be used to predict the cross-section of stock returns out of sample. In particular, a strategy that goes long the quintile portfolio with the highest exposure to the Bayesian factor and short the quintile portfolio with the lowest exposure to the Bayesian factor yields a Fama-French-Carhart alpha of 1.7% per month (20.4% annualized). This provides a new empirical evidence to support three mainstream models: the ICAPM, the Intermediary Asset Pricing Model, and the Disaster Risk Model. On top of the first stage, the second stage is to impute the monthly Bayesian factor and achieve our final goal of generating a high-frequency risk factor. We use the Natural Language Process technique Word Embedding to read the headlines and abstracts of all daily financial articles from the New York Times and aggregate all daily semantic and syntactic information into one 100-dimension word vector. This daily 100-dimension word vector is then used as a regressor to mimic and track the monthly Bayesian factor. The mimicking regression could then produce daily mimicking values of the monthly risk factor, which become our daily systemic risk factor. This daily text mimicking risk factor is validated in several ways including by showing how well it captures the 2008 crisis. We also find that the text mimicking risk factor is priced in the cross-section of stock returns at the daily level and can predict large swings in the VIX using a quantile regression approach, which sheds some light on the puzzling relation between the macro-economy and stock market volatility.