Animal Spirits

What does the U.S. stock market at industry/firm levels say about investor expectations during and after the 2019 coronavirus (COVID-19) pandemic? In the April 2020 update of their paper entitled “Feverish Stock Price Reactions to COVID-19”, Stefano Ramelli and Alexander Wagner examine and interpret industry/firm-level reactions to COVID-19 across three pandemic phases:

1. Incubation: January 2-17,
2. Outbreak: January 20-February 21,
3. Fever: February 24-March 20.

They estimate each stock’s abnormal return during these phases as its 1-factor (market) alpha minus its beta times the market excess return. They estimate alpha and beta via regression of daily excess stock returns on daily excess value-weighted market returns during 2019. They use the yield on 1-month U.S. Treasury bills (T-bill) as the risk-free rate for calculating excess return. Using daily dividend-adjusted stock prices for Russell 3000 stocks (excluding financial stocks for leverage-related analyses), market returns and T-bill yields during December 31, 2018 through March 20, 2020, they find that: Keep Reading

Are firms that engage the public via Twitter more expanding (via exposure) or shrinking (via adverse social media frenzy) their opportunity sets? In their January 2020 paper entitled “The Social Media Risk Premium”, Amin Hosseini, Gergana Jostova, Alexander Philipov and Robert Savickas investigate relationships between firm Twitter activity and stock return. Their data include firm Twitter presence, and level and nature of activity, as well as responses from followers. Using these Twitter data, accounting data and stock returns for all publicly held U.S. firms during 2007 through 2016 (33,445,318 tweets, generating 25,603,977 replies, 161,548,941 retweets and 265,738,508 likes), they find that: Keep Reading

Do investors systematically and exploitably underreact to deviations in firm fundamentals from recent averages? In their January 2020 paper entitled “Anchoring on Past Fundamentals”, Doron Avramov, Guy Kaplanski and Avanidhar Subrahmanyam investigate how deviations of quarterly firm accounting variables from averages over recent quarters relate to future returns across stocks. They first construct a stock performance deviation index (PDI) based on seven variables: (1) cash and short-term investments, (2) retained earnings, (3) operating income, (4) sales, (5) capital expenditures, (6) invested capital and (7) inventories. They then each month for each stock starting June 1977:

• Calculate the deviation for each variable as the difference between its most recent quarterly value and its average over the preceding three quarters, scaled by total assets.
• Rank each deviation (in percentiles) relative to deviations for the same variable for all stocks.
• Calculate PDI for a stock as the equally weighted average of percentile rankings across all seven variables.

They extend this approach to a more comprehensive fundamental-based deviation index (FDI) that considers deviations of all Compustat accounting variables plus 14 commonly used accounting ratios, with weights of deviation percentile rankings optimized via least absolute shrinkage and selection operator (LASSO) regression starting January 1979. For all variables, if the exact release date is unavailable, they assume a 60-day delay in release. For portfolio tests, they calculate returns to hedge portfolios that are long (short) stocks in the top (bottom) tenth, or decile, of PDIs or FDIs, with holding intervals ranging from one to 24 months. Using monthly data needed to construct PDI, FDI and 30 style, technical, fundamental and liquidity control variables across a broad sample of reasonably liquid U.S. common stocks with positive book values and prices over $5 during January 1976 through October 2017, they find that: Keep Reading

Is retail trading a reliable driver of U.S. stock momentum? In his November 2019 paper entitled “Retail Trading and Momentum Profitability”, Douglas Chung investigates interactions across stocks between current proportion of retail trading and future momentum returns. Specifically, for each month and for each of two recent stock samples, he:

• Sorts stocks into fifths (quintiles) by current proportion of retail trading.
• Within each proportion-of-retail-trading quintile:
  • Sorts stocks into sub-quintiles by return from 12 months ago to one month ago.
  • Calculates average next-month returns for an equal-weighted momentum portfolio that is long (short) the sub-quintile of stocks with the highest (lowest) past returns. He also considers other portfolio weighting schemes.
  • Measures alphas of these returns based on various widely accepted single-factor and multi-factor models of stock returns.

He next tests whether proportion of retail trading relates to a gambling motive (lottery trading) by constructing a stock lottery index from inverse of stock price, idiosyncratic volatility, idiosyncratic skewness and recent maximum daily return. In other words, he examines whether the lottery index value for a stock is a proxy for its proportion of retail trading. Using daily data for all NYSE retail orders during March 2004 through December 2014, for small NYSE trades of U.S. common stocks (a proxy for retail trading) during January 1993 through July 2000 and for lottery index inputs during 1940 through 2016, he finds that: Keep Reading

Which equity factors from among those included in the most widely accepted factor models are really important? In their October 2019 paper entitled “Winners from Winners: A Tale of Risk Factors”, Siddhartha Chib, Lingxiao Zhao, Dashan Huang and Guofu Zhou examine what set of equity factors from among the 12 used in four models with wide acceptance best explain behaviors of U.S. stocks. Their starting point is therefore the following market, fundamental and behavioral factors:

They compare 4,095 subsets (models) of these 12 factors models based on: Bayesian posterior probability; out-of-sample return forecasting performance; gross Sharpe ratios of the optimal mean variance factor portfolio; and, ability to explain various stock return anomalies. Using monthly data for the selected factors during January 1974 through December 2018, with the first 10 (last 12) months reserved for Bayesian prior training (out-of-sample testing), they find that: Keep Reading

Does the market react most strongly to the earliest quarterly earnings announcements? In their October 2019 paper entitled “Calendar Rotations: A New Approach for Studying the Impact of Timing using Earnings Announcements”, Suzie Noh, Eric So and Rodrigo Verdi study effects of the relative order of U.S. firm quarterly earnings announcements, which vary systematically for some firms according to the day of the week of the first day of a month. Specifically, they qualify firms by identifying those firms that exhibit systematic earnings announcement schedules (such as Friday of the fourth week after quarter ends, sometimes set in firm bylaws) for at least four consecutive same fiscal quarters. They then for each firm each fiscal quarter:

• Calculate EA Order, ranking of earnings announcement date divided by number of firms with the same fiscal quarter-end.
• Compute change in EA Order compared to the same fiscal quarter last year, indicating a calendar acceleration or delay in announcement. Positive (negative) change in EA Order indicates delay (acceleration)
• Examine effects of change in EA Order on media coverage (number of articles), stock trading volume and stock return from one trading day before to one trading day after earnings announcement.

Using sample of 76,622 firm-quarters during 2004 through 2017, they find that: Keep Reading

What are ways to mitigate biases that interfere with rational investment decision-making? In their September 2019 paper entitled “The Psychology of Financial Professionals and Their Clients”, Kent Baker, Greg Filbeck and Victor Ricciardi describe common psychological biases and suggest ways to overcome them. Based on their knowledge and experience, they conclude that: Keep Reading

Can investors identify stocks that incorporate news slowly enough to allow exploitation? In their August 2019 paper entitled “Tomorrow’s Fish and Chip Paper? Slowly Incorporated News and the Cross-Section of Stock Returns”, Ran Tao, Chris Brooks and Adrian Bell classify stocks incorporating news quickly (QI) or slowly (SI) into prices and investigate implications for associated future returns. Specifically, they each month:

1. Assign a sentiment score to each current-month news article about each stock based on words in the article.
2. Double-sort stocks by thirds based first on current-month abnormal (adjusted for size, industry value and industry momentum) returns and then on news sentiment scores, yielding nine groups.
3. Classify stocks that are: (a) high return/low sentiment (HRLS) or low return/high sentiment (LRHS) as SI; and, (b) high return/high sentiment (HRHS) or low return/low sentiment (LRLS) as QI.
4. Measure average next-month returns of equally-weighted SI and QI portfolios that are, respectively: (a) long LRHS stocks and short HRLS stocks; and, (b) long HRHS stocks and short LRLS stocks.
5. Measure average next-month return of an equally weighted portfolio that is long the SI portfolio and short the QI portfolio (Slow-Minus-Quick, SMQ).

They then examine whether limited investor attention or differences in news complexity and informativeness better explain results. Using firm-level news data, firm characteristics and associated stock returns for a broad sample of U.S. common stocks during 1979 through 2016, they find that: Keep Reading

How does the market react when firms announce adoption of blockchain technology? In the May 2019 draft of their paper entitled “Bitcoin Speculation or Value Creation? Corporate Blockchain Investments and Stock Market Reactions”, Don Autore, Nicholas Clarke and Danling Jiang study stock price reactions to initial public announcements of investments in blockchain technology by listed U.S. firms. Their key metric is buy-and-hold abnormal return (BHAR) relative to each of five benchmarks: (1) portfolios of stocks matched on size and book-to-market (BM); (2) portfolios of stocks matched on market beta; 3) a broad value-weighted market index; (4) iShares Global Financials ETF (IXG); and, (5) iShares Global Tech ETF (IXN). Their announcement event windows is five trading days before initial public announcement of an investment in blockchain technology (-5) to 65 trading days after (65). Using dates of initial public announcements of investments in blockchain technology and contemporaneous daily returns for 207 stocks listed on NYSE and NASDAQ during October 2008 through March 2018, they find that:

Keep Reading

Who do investors trust more, expert advisors or algorithms? In her March 2019 paper entitled “Algorithmic Decision-Making: The Death of Second Opinions?”, Nizan Packin employs a survey conducted on Amazon Mechanical Turk to assess automation bias when making significant investment decisions. Each of four groups of respondents received one of the following four questions (response scale 1 to 5):

1. “You decide to invest 15% of your savings in the stock market. You find a reputable stockbroker, who makes investment recommendations. How confident are you that you got the best recommendation possible for your investment?”
2. “You decide to invest 60% of your savings in the stock market. You find a reputable stockbroker, who makes investment recommendations. How confident are you that you got the best recommendation possible for your investment?”
3. “You decide to invest 15% of your savings in the stock market. You find a reputable online automated investment advisor, who makes investment recommendations. How confident are you that you got the best recommendation possible for your investment?”
4. “You decide to invest 60% of your savings in the stock market. You find a reputable online automated investment advisor, who makes investment recommendations. How confident are you that you got the best recommendation possible for your investment?”

A followup question asked about level of comfort trusting again the same expert (human or algorithmic) after learning that the initial recommendation resulted in a significant loss. Analyses included controls for respondent age, gender, socioeconomic status, having some college education, race and political ideology (liberal/conservative). Based on 800 total responses to specified survey questions, she finds that: Keep Reading