Wolfe Research Vice Chairman and Head of QES Research, Yin Luo, hosted a webcast on the recent massive factor moves (+6 standard deviations losses in momentum and similar rallies in value). Topics of discussion included what happened, the likely causes of the momentum crash, further downside risks in the factor space, and more importantly, what to do from here.
Search Coverage List, Models & Reports
Search Results1-10 out of 194
On September 9, the price momentum factor in the US suffered from a -6x standard deviation loss (down -4.0%), matching the levels last seen during the March-May 2009 risk rally and far exceeding the summer 2007 quant crisis. On the same day, the book-to-market cyclical value signal rallied by 2.9% (a 5x sigma event). The massive rotation in the factor space is matched by much milder moves among the 11 sectors (from defensive sector to energy/financials). The momentum crash was not limited to the US – we observe similar sell-offs in most other regions, e.g., Europe, Canada, Australia, and LATAM.
This research continues our efforts of building stock selections models for the global public real estate market (including both REITs and property stocks). In Part I, we offered a thorough review of the investment universe (>800 REITs and property stocks globally), alternative data sources, and real estate stock selection factors.
The trade conflict between Washington and Beijing continued to escalate in August and may persist for a long time. The market has priced in a great deal of trade war risks. While the global equity market and Chinese currency continue to retreat with negative trade development, the magnitude of the market sell-off appears to be moderating, suggesting the market is increasingly resilient to negative sentiment.
Continuing our quest for alternative signals from unstructured textual data, in this research, we propose the Fifth Generation (5G) of our NLP/ML modeling framework. The focus of our 5G model suite is annual and interim corporate regulatory filings, sourced from the SEC’s EDGAR database for US companies and an alternative data vendor Mergent for international firms.
We combine our NLP signals, machine learning driven sentiment analysis (via Elastic Net and xgBoost), along with our deep learning algorithm (CNN) together in the new GINA, which supersedes the previous SPEC. The GINA model delivers strong and uncorrelated performance in the US markets, with a Sharpe ratio of 1.0x. The GINA model covers over 2,000 stocks in Europe, Asia (including Japan and China), LATAM, and Canada, with investment horizon beyond a year.
Seasonal effects (e.g., January effect, December tax-loss selling) are well documented in academic literature and our previous research. In this month’s Portfolio Compass, we focus on another interesting seasonal anomaly observed during the month of October. We find the performance of most common stock selection factors to be particularly strong in the month of October. The October effect is closely related to the “sell in May and go away” phenomenon. Investors obviously do not want to hold risky positions when they are away. As people take extended vacations in the summer months, they are likely to be more passive. As a result, the market tends to be not as efficient. As investors come back after Labor Day, market inefficiency has already accumulated to a point that makes factor investing particularly attractive.
Despite the sharp selloff in recent days, US equities are still up significantly from the December-2018 lows. It appears that the worries about the economy and the mounting economic, policy, trade and geopolitical uncertainties, as signaled by the strength in the bond market, are offset by the enthusiasm of a new round of accommodative monetary policies.
- 1 of 20
- next →