Sentiment Indicators

Does the combination of an artificial intelligence (AI)-generated financial news sentiment with a complex financial stress metric generate good stock market timing signals? In their April 2024 paper entitled “Mixing Financial Stress with GPT-4 News Sentiment Analysis for Optimal Risk-On/Risk-Off Decisions”, Baptiste Lefort, Eric Benhamou, Jean-Jacques Ohana, David Saltiel, Beatrice Guez and Thomas Jacquot devise and test a risk-on/risk-off strategy for stock market timing. The strategy combines:

• Stress Index (SI): based on VIX, TED spread, Credit Default Swap index and realized volatilities across major equity, bond and commodity markets, all normalized and then aggregated by category. Overall SI is the average of category results, rescaled to fall between 0 and 1.
• News sentiment signal: 10-day moving average of ChatGPT 4 assessments of the sentiment (1 for positive or 0 for negative) in Bloomberg daily market summaries.

They consider six strategies:

1. Benchmark (or Long Only) – buy and hold the index, with constant volatility scaling to match the final (retrospective) volatility of an active strategy.
2. VIX – weight the stock index according to VIX, with times of stress indicated by VIX above its 80th percentile.
3. SI – weight the stock index according to the value of SI as described above.
4. News – weight the stock index according to the ChatGPT 4 news sentiment signal.
5. SI News – weight the stock index according to the product of SI and News.
6. Dynamic SI News – because SI News either significantly outperforms or underperforms SI alone during subperiods, each month weight the stock index according to either SI alone or SI News, whichever has the better Sharpe ratio over the past 250 trading days at the end of the prior month.

They test the strategy on the S&P 500 Index, the NASDAQ 100 Index and an equal-weighted combination of these two indexes plus the Nikkei 225, Euro Stoxx 50 and Emerging Markets indexes. They assume trading frictions of 0.2% of value traded. Using daily values of all specified inputs during January 2005 through December 2023, they find that: Keep Reading

“CNN Fear and Greed Index as Return Predictor” reports findings from a draft study that the CNN Fear and Greed Index (F&G) may be useful for U.S. stock index timing. The authors of that paper generously provided their hand-collected sample of daily CNN F&G levels for 4/7/21 through 3/8/24. We partly validate and extend that sample using daily values from the Timeline view of Fear & Greed Index as of 8/12/24. We then relate daily CNN F&G and daily changes in CNN F&G to daily returns for SPDR S&P 500 ETF Trust (SPY). Using the validated/extended sample of daily CNN F&G and contemporaneous daily dividend-adjusted prices for SPY during 4/7/21 through 8/12/24, we find that: Keep Reading

Is the CNN Fear and Greed Index useful for predicting asset returns? In the July 2024 draft of their paper entitled “The CNN Fear and Greed Index as a Predictor of Us Equity Index Returns”, flagged by a subscriber, Hugh Farrell and Fergal O’Connor use regressions of hand-collected data to investigate whether the index reliably predicts returns on S&P 500, Nasdaq Composite and Russell 3000 stock indexes and gold. The CNN Fear and Greed Index is the simple average of seven factors (market momentum, stock price strength, stock price breadth, put-to-call options ratio, VIX to measure market volatility, safe haven demand and junk bond demand), each scaled to a range of to 100. The value 1 (100) indicates extreme fear (greed). Using daily CNN Fear and Greed Index levels from a GitHub repository during January 2011 through mid-September 2020 and from the Wayback Machine during early April 2021 through early March 2024 (intervening data are unavailable), and contemporaneous daily stock index levels and gold price, they find that:

Keep Reading

A subscriber noted and asked: “Michigan (at one point) claimed that the inflation expectations part of their survey of consumers was predictive. That was from a paper long ago. I wonder if it is still true.” To investigate, we relate monthly “Expected Changes in Prices” (expected annual inflation) from the monthly University of Michigan Survey of Consumers and actual U.S. inflation data based on the monthly non-seasonally adjusted consumer price index (U.S. city average, All items). The University of Michigan releases final survey data near the end of the measured month. We consider two relationships:

• Expected annual inflation versus one-year hence actual annual inflation.
• Monthly change in expected annual inflation versus monthly change in actual annual inflation.

As a separate (investor-oriented) test, we relate monthly change in expected annual inflation to next-month total returns for SPDR S&P 500 ETF Trust (SPY) and iShares 20+ Year Treasury Bond ETF (TLT). Using monthly survey/inflation data since January 1978 (limited by survey data) and monthly SPY and TLT total returns since July 2002 (limited by TLT), all through April 2024, we find that: Keep Reading

What does the body of research say about implications of new social media for financial markets? In their April 2024 paper entitled “Social Media and Finance”, Anthony Cookson, William Mullins and Marina Niessner survey research on social media in finance, distinguishing between research using social media to explore investor behaviors and research on the effects of social media on financial markets. Based on the body of relevant research, they conclude that: Keep Reading

Does combining widely used measures of equity market stress with news sentiment as interpreted by large language models such as ChatGPT support a robust risk-on/risk-off market timing strategy? In their April 2024 paper entitled “Stress Index Strategy Enhanced with Financial News Sentiment Analysis for the Equity Markets”, Baptiste Lefort, Eric Benhamou, Jean-Jacques Ohana, David Saltiel, Beatrice Guez and Thomas Jacquot test a risk-on/risk-off strategy for equity markets that combines:

• A conventional stress index (SI) signal derived from VIX, the TED spread, a credit default swap (CDS) index and volatilities of major equity, bond and commodity markets. They standardize each measure, aggregate measures by asset class, average results across asset classes and normalize the average to fall between 0 and 1.
• A ChatGPT 4 assessment of market sentiment from Bloomberg Daily Market Wraps over the past 10 days to determine whether it is above (risk-on) or below (risk-off) historical average.

They consider six strategies and apply them to the S&P 500 Index alone, the NASDAQ Index alone or an equal-weighted basket of S&P 500, NASDAQ, Nikkei, Euro Stoxx and Emerging Markets indexes:

1. Buy-and-Hold the index or basket of indexes (benchmark).
2. VIX: Risk-off when VIX is above its 80th percentile (about 26).
3. SI: Weight stocks according to the SI signal alone.
4. News: Hold stocks according to the Bloomberg Daily Market Wraps sentiment signal alone.
5. SI+News: Weight stocks according to the product of SI and News signals.
6. Dynamic SI+News: each month weight stocks using either the SI+News signal or the SI signal, whichever has the higher Sharpe ratio over the last 250 trading days.

For comparison, they retrospectively scale long-only benchmarks to have the same volatility as the best-performing active strategy. For all 18 strategy tests, they assume frictions of 0.02% on portfolio turnover. Using the specified SI inputs and daily stock index returns since January 2005, and Bloomberg Daily Market Wraps since 2010, all through December 2023, they find that:

Keep Reading

Can ChatGPT extract market sentiment from financial news that is useful for timing equity markets? In their April 2024 paper entitled “Sentiment Analysis of Bloomberg Markets Wrap Using ChatGPT: Application to the NASDAQ”, Baptiste Lefort, Eric Benhamou, Jean-Jacques Ohana, David Saltiel, Beatrice Guez and Thomas Jacquot use ChatGPT to assess whether daily Bloomberg Global Markets Wrap, Market Talks and Morning Reports anticipate NASDAQ returns. Specifically, they each day:

• Ask ChatGPT to identify important news themes and characterize them as headlines.
• Ask ChatGPT to assess whether each headline is positive, negative or neutral for future stock prices.
• Compute a sentiment score that combines sentiments for all daily headlines.
• Compute a daily cumulative sentiment score (C) for the last 20 trading days.
• Compute a daily detrended cumulative sentiment score (DC) by comparing C to its value over the last 20 trading days (extending the overall lookback interval to 40 days).
• If C is positive (negative), take a long (short) position in the NASDAQ index with a 2-day lag to ensure executability and a debit of 0.2% trading frictions for position changes. Repeat this evolution for DC.

They separately examine cumulative performances of the long, short and overall returns of the C and DC variations of this strategy, focusing on Sharpe, Sortino and Calmar ratios as key performance metrics. Their benchmark is buying and holding the NASDAQ index. Using the specified daily financial news sources and daily NASDAQ index returns during 2010 through 2023, they find that:

Keep Reading

Are sentiments conveyed in Seeking Alpha articles useful for stock picking? In their January 2023 paper entitled “Seeking Alpha: More Sophisticated Than Meets the Eye”, Duo Selina Pei, Abhinav Anand and Xing Huan apply two-pass natural language processing to test the informativeness of articles from Seeking Alpha incremental to publicly available earnings data. Specifically, they each month:

• Associate articles with one or more specific stocks.
• Extract positive and negative sentiment at both phrase and aggregate levels for each article/stock.
• Calculate a standardized net sentiment for each article/stock based on the difference between positive and negative mentions, emphasizing event sentiment over general sentiment.
• Rank articles/stocks based on standardized net sentiment over the last month. Reform equal-weighted portfolios of articles/stocks by ranked tenths (deciles). Calculate both immediate [-1,+1] and 90-day future [+2,+90] average gross raw returns and average gross abnormal returns adjusted for size, book-to-market and momentum.
• Sort stocks into 20 groups based on monthly standardized net sentiments up to two days before portfolio selection, excluding stocks with few articles or neutral sentiment. Reform an equal-weighted hedge portfolio that is long stocks with the highest sentiments and short stocks with the lowest (on average, 105 long and 86 short positions).

Using 350,095 articles published on Seeking Alpha since its inception in 2004 through the beginning of October 2018, daily returns of matched stocks and their options and associated earnings surprise data as available, they find that: Keep Reading

Can artificial intelligence platforms such as ChatGPT be good stock day traders? In his March 2024 paper entitled “Can ChatGPT Generate Stock Tickers to Buy and Sell for Day Trading?”, Sangheum Cho tests whether ChatGPT 3.5 turbo supports profitable day trading. He instructs ChatGPT to pretend to be a professional day trader who picks from among U.S. listed stocks 100 to buy and 100 to sell for short-term returns based on daily Bloomberg and the Wall Street Journal news blurbs on Twitter. Each day, prior to the market open, he:

• Uses the Refinitiv Eikon News Monitor to collect the selected tweets from the past 24 hours. He removes hyperlinks and duplicate tweets.
• Segments the tweets into batches to accommodate ChatGPT processing limitations.
• For each batch, asks ChatGPT to generate 100 BUY and 100 SELL signals, with 30 iterations for each batch to amplify signals by suppressing spurious selections. He then constructs equal-weighted long and short portfolios of stocks with signals.
• For each stock with signals:
  • Sums BUY and SELL signals across batches/iterations to calculate SUM_BUY and SUM_SELL signals. He constructs signal count-weighted long and short portfolios from these summed signals.
  • Subtracts SUM_SELL from  SUM_BUY to calculate NET_BUY and NET_SELL signals. He constructs signal magnitude-weighted long and short portfolios from these netted signals.

For each portfolio, he excludes stocks with zero daily volume, missing daily prices or incomplete trading histories for the previous five trading days. He measures returns from the market open to the market close. Using 222,659 tweets (only 16,359 of which are firm-specific) and daily opening and closing prices for U.S. listed common stocks during December 2022 through December 2023 (271 trading days), he finds that:

Keep Reading

“Using the Money Anxiety Index for ETF Selection” examines whether the proprietary Money Anxiety Index (MAI) can select long and short portfolios of ETFs that beat the S&P 500 Index (ignoring dividends). Test outputs are 5-year, 3-year and 1-year cumulative returns. A deeper look at performance may be helpful. We extend the test period by eight months and focus on the full period. We consider SPDR S&P 500 ETF Trust (SPY) with dividends as a benchmark. We also consider Invesco QQQ Trust (QQQ), which has much affinity with the MAI-selected ETFs, as a benchmark. We compute monthly return statistics, along with compound annual growth rates (CAGR) and maximum drawdowns (MaxDD). Using beginning-of-month, dividend-adjusted prices for the 10 ETFs in the MAI-generated portfolios and the two benchmarks from the beginning of May 2018 through the beginning of March 2024, we find that: Keep Reading