Animal Spirits

Do presumably expert early-stage startup investors, whether individuals (Angels) or institutions (Venture Capitalists) invest efficiently? In his June 2022 paper entitled “Predictably Bad Investments: Evidence from Venture Capitalists”, Diag Davenport applies machine learning methods based on information known at the time of investment to evaluate decisions of early-stage investors. He defines early-stage investments as equity deals within two years of incubator completion categorized in Pitchbook as deal types Series A, Series B, Seed Round or Angel (Individual). He define late-stage exit as initial public offering, merger/acquisition or funding categorized in Pitchbook as Series C or later. He uses his first five years of quantitative data and numerical transformations of the qualitative data (text) in training a model with XGBoost to predict future venture success. He then applies the model to the next three years of data to build a portfolio that substitutes conventional investments (such as the S&P 500 Index) for predictably bad ventures. Using venture financials and qualitative information about the CEO from Pitchbook for 16,054 startups accepted into top accelerator programs during 2009 through 2016 (2009-2013 for model training and 2014-2016 for testing), he finds that:

Keep Reading

Should investors expect materially different returns for stocks accepted or excluded by institutional investors based on firm environmental, social and corporate governance (ESG) policies and practices? In their April 2022 paper entitled “The Expected Returns of ESG Excluded Stocks. The Case of Exclusions from Norway’s Oil Fund”, Erika Berle, Wangwei He and Bernt Ødegaard analyze aggregate performance of stocks excluded by the Norwegian Government Pension Fund Global portfolio based on ESG-related conduct or products, used as a model by many institutional investors. They construct various equal-weighted (EW) and value-weighted (VW) portfolios of excluded stocks and measure returns and Fama-French 5-factor (market, size, book-to-market, profitability and investment) alphas of these portfolios. Using monthly returns in U.S. dollars and firm data for a sample of 186 excluded stocks, with some exclusions revoked, during 2005 through early 2022, they find that:

Keep Reading

A subscriber hypothesized that keeping old underwear is an early indicator of personal income at risk. Trends in underwear, as proxied by Hanesbrands Inc. (HBI), may therefore be a leading indicator of trends in the overall stock market. To test this hypothesis, we relate HBI returns to SPDR S&P 500 ETF Trust (SPY) returns at a monthly frequency. Using monthly dividend-adjusted prices for HBI and SPY during September 2006 (limited by HBI) through March 2022, we find that: Keep Reading

Do often frenzied investors/traders tend to overdo buying and selling, coming to their senses shortly thereafter? In other words, does the broad U.S. stock market tend to revert after short-term moves up or down? To check, we relate sequential past and future return intervals of 1, 2, 3, 5, 10, 15 and 21 trading days. Using daily closes of the S&P 500 Index over the period January 1928 through mid-March 2022, we find that: Keep Reading

Are firms offering products and services purported to mitigate climate change compelling investments? In the February 2022 revision of their paper entitled “Climate Solutions Investments”, Alexander Cheema-Fox, George Serafeim and Hui Wang analyze international reports, regional net zero frameworks, research papers and news to develop a list of 164 key words/phrases associated with climate change solution business areas. They apply these key words/phrases to firm descriptions to identify 632 actively traded pure plays in climate solutions. They then characterize geographies, accounting fundamentals and valuation ratios for this sample and construct monthly rebalanced value-weighted and equal-weighted climate solutions portfolios (CSP). Using monthly firm fundamentals and stock trading data for these 632 firms from the end of 2010 through October 2021, they find that: Keep Reading

Does variation in the number of reported cases of COVID-19 predict near-term asset returns? To investigate, we look for a test acknowledging that the available sample is short and very noisy. Specifically:

• To suppress noise, we use the 7-day moving average of U.S. COVID-19 cases.
• To avoid measurement overlap, we calculate weekly changes in this average and compare these changes to next-week returns for SPDR S&P 500 Trust (SPY) and iShares Barclays 20+ Year Treasury Bond (TLT).
• To assess reliability of any relationship, we look at rolling 13-week correlations between weekly changes in COVID-19 data and next-week asset returns. While 13 weeks is a short measurement interval for noisy data, consistency in outputs would offer some confidence that there is a reliable relationship.

Using weekly (Friday) COVID-19 case data from the Centers for Disease Control (CDC) and weekly (Friday close) dividend-adjusted SPY and TLT levels during late January 2020 (limited by COVID-19 data) through mid-September 2021, we find that: Keep Reading

Is intensity of public interest in a “bear market” useful for predicting stock market return? To investigate, we download monthly U.S. Google Trends search intensity data for “bear market” and relate this series to monthly S&P 500 Index returns. For comparison with the “investor fear gauge,” we also relate search data to monthly CBOE option-implied S&P 500 Index volatility (VIX) levels. Google Trends analyzes a percentage of Google web searches to estimate the number of searches done over a certain period. “Each data point is divided by the total searches of the geography and time range it represents to compare relative popularity… The resulting numbers are then scaled on a range of 0 to 100 based on a topic’s proportion to all searches on all topics.” Using the specified data as of 9/14/2021 for the period January 2004 (earliest available on Google Trends) through August 2021, we find that: Keep Reading

Do motives of financial market researchers justify strong skepticism of their findings? In his brief August 2021 paper entitled “Be Skeptical of Asset Management Research”, Campbell Harvey argues that economic incentives undermine belief in findings of both academic and practitioner financial market researchers. Based on his 35 years as an academic, advisor to asset management companies and editor of a top finance journal, he concludes that: Keep Reading

How frequently and permanently do individual U.S. investors sell stocks in a panic? In their August 2021 paper entitled “When Do Investors Freak Out?: Machine Learning Predictions of Panic Selling”, Daniel Elkind, Kathryn Kaminski, Andrew Lo, Kien Wei Siah and Chi Heem Wong examine frequency, timing and duration of panic selling. They define panic selling as a drop of at least 90% in account equity value within a month, of which at least 50% is due to trading. They also estimate the opportunity of cost of panic selling. Finally, they apply deep neural network software to predict a month in advance which individuals will panic sell based on recent market conditions and investor demographics/financial history. Using account equity value and trade data for 653,455 individual U.S. brokerage accounts belonging to 298,556 households during January 2003 through December 2015, they find that:

Keep Reading

Does technology amplify adverse herding among inexperienced investors? In their October 2020 paper entitled “Attention Induced Trading and Returns: Evidence from Robinhood Users”, Brad Barber, Xing Huang, Terrance Odean and Christopher Schwarz test the relationship between episodes of intense stock buying by retail (Robinhood) investors and future returns. Their source for buying intensity is the stock popularity feature of Robintrack from May 2, 2018 until discontinuation August 13, 2020 (with 11 dates missing and two hours missing for 16 other dates), during which the number of Robinhood user-stock positions grows from about 5 million to over 42 million. They define intense stock buying (herding event) as a dramatic daily increase in number of Robinhood users owning a particular stock in two ways:

1. Among stocks with at least 100 owners at the start of the day, select those in the top 0.5% of ratio of owners at the end of the day to owners at the beginning of the day.
2. Select stocks with at least 1,000 new owners and at least a 50% increase in owners during the day.

Using Robintrack data supporting these definitions and associated daily stock returns, open and close prices, closing bid-ask spreads and market capitalizations, they find that: Keep Reading