Sentiment Indicators

Are the latest forms of artificial intelligence (AI) better at forecasting stock market returns than humans? In his February 2023 preliminary paper entitled “Surveying Generative AI’s Economic Expectations”, Leland Bybee summarizes results of monthly and quarterly forecasts by a large language model (ChatGPT-3.5) of U.S. stock market returns and 13 economic variables based on samples of Wall Street Journal (WSJ) news articles. He uses the S&P 500 Index as a proxy for the U.S. stock market. He asks ChatGPT to provide reasons for responses. He compares accuracy of ChatGPT forecasts to those from: (1) surveys of humans, including the Survey of Professional Forecasters, the American Association of Individual Investors (AAII) and the Duke CFO Survey; and, (2) a wide range of fundamental and economic predictors tested in past research. Using monthly samples of 300 randomly selected WSJ news articles, results of human surveys and various fundamental/economic data during 1984 through 2021, he finds that:

Keep Reading

What kind of return patterns are typical of beginnings and ends of equity bull and bear markets? In his April 2023 paper entitled “Investor Overreaction: Evidence From Bull and Bear Markets”, Valeriy Zakamulin examines return patterns of U.S. stock market bull and bear states as a way to decide when investors tend to overreact. He uses the S&P 500 Index as a proxy for the U.S. stock market. He applies a pattern recognition algorithm to: (1) identify index peaks and troughs; and, (2) ensure that a full bear-bull cycle lasts at least 16 months and bear or bull states last at least 5 months, unless the index rises or falls by more than 20%. He then standardizes the duration of each market state to 10 intervals and assumes that the bull or bear return evolves quadratically with state age. Because the available sample is relatively small, he applies bootstrapping to enhance reliability of findings. Using monthly S&P 500 Index returns (excluding dividends) during January 1926 through December 2022, he finds that: Keep Reading

Can advanced natural language processing models such as ChatGPT extract sentiment from firm news headlines that usefully predicts associated next-day stock returns? In their April 2023 paper entitled “Can ChatGPT Forecast Stock Price Movements? Return Predictability and Large Language Models”, Alejandro Lopez-Lira and Yuehua Tang test the ability of ChatGPT to predict next-day returns of individual stocks via analysis of relevant article and press release headlines from RavenPack, a leading provider of news sentiment data. They pre-process the headlines to ensure unique content and high relevance to a specific firm. They next instruct ChatGPT to designate whether each headline is good, bad or irrelevant for firm stock price, as follows:

“Forget all your previous instructions. Pretend you are a financial expert. You are
a financial expert with stock recommendation experience. Answer “YES” if good
news [+1], “NO” if bad news [-1} , or “UNKNOWN” if uncertain in the first line [0]. Then
elaborate with one short and concise sentence on the next line. Is this headline
good or bad for the stock price of _company_name_ in the _term_ term?”

They then compute a ChatGPT score for each stock in the news (averaging if there is more than one headline for a firm) and relate all stock scores to stock returns lagged by one day. They further compare predictive powers of ChatGPT sentiment scores to those provided by RavenPack. Using daily returns for a broad sample of U.S. common stocks and daily news headlines from RavenPack during October 2021 (post-training period for ChatGPT) through December 2022, they find that: Keep Reading

Can surveys of various expert and inexpert groups usefully predict stock market returns? In their March 2023 paper entitled “How Accurate Are Survey Forecasts on the Market?”, Songrun He, Jiaen Li and Guofu Zhou assess abilities of the following three surveys to predict S&P 500 Index returns:

• The Livingston (LIV) survey of professional forecasters from industry and academia, conducted each June and December.
• The Chief Financial Officer (CFO) survey directed quarterly to about 4,500 U.S. firms.
• A combination (NX) survey that combines inputs from sources such as the UBS/Gallup survey, the Conference Board survey and the Michigan Survey of Consumers to represent the typical U.S. household.

For comparison, they also look at two other predictors, one based on a set of economic variables and the other based on aggregate short interest for U.S. stocks. Their benchmark forecast is a simple random walk tethered to the historical mean return. They test forecast accuracies statistically and gauge the economic value of each forecast based on out-of-sample certainty equivalence gain and Sharpe ratio for a portfolio that times the S&P 500 Index based on the forecast (versus buying and holding the index). Using data for the selected surveys, the set of economic variables, aggregate short interest for U.S. stocks and the S&P 500 Index as available (various start dates) through December 2020, they find that: Keep Reading

Does combining a sentiment indicator with a trend following indicator improve performance of a stocks-bonds timing strategy? In his October 2022 paper entitled “The Complementarity of Trend Following and Relative Sentiment”, Raymond Micaletti investigates effects of combining the following trend following (TF) and relative sentiment (RS) indicators:

• TF – at the end of each month switch to a broad U.S. stock market index (an aggregate bond index) when the prior-close stock market index crosses above (below) its 10-month simple moving average (SMA) strategy. This strategy is the best of six similar SMA strategies.
• RS – each week update the equity allocation from 0% to 100% based on an equal-weighted combination of three prior-week inputs, two of which are driven by weekly Commitments of Traders reports and one of which is driven by monthly Sentix relative sentiment, with the balance of the portfolio in an aggregate bond index. Update the equity allocation only if it differs from the prior allocation by more than 10%.

The combined strategy (TFRS) is a 50-50 mix of TF and RS. He applies frictions of 0.04% to account for costs of both stock and bond index allocation changes. For interpretation of results, he focuses on nine times the equity index suffers a drawdown of at least 10% from an all-time high. Using daily U.S. equity market total returns and U.S. Treasury bill yields (for Sharpe ratio calculations) from the Kenneth French data library, daily levels of Bloomberg Barclays U.S. Aggregate Bond Total Return Index, weekly Commitments of Traders reports and the monthly Sentix economic outlook survey of institutional and individual investors during November 1994 through August 2022, he finds that: Keep Reading

Does the sentiment expressed by major newspapers about the economy usefully predict stock market returns? To investigate, we employ the Daily News Sentiment Index, constructed from “economics-related news articles from 24 major U.S. newspapers [across] all major regions of the country…with at least 200 words… The Daily News Sentiment Index is constructed as a trailing weighted-average of time series, with weights that decline geometrically with the length of time since article publication.” Update frequency is weekly, suggesting use of a 7-day simple moving average (SMA7). We calculate the SMA7 on Sundays and relate this series to weekly S&P 500 Index returns calculated from closes on the following Mondays. Using the specified weekly data series during January 1980 (limited by the sentiment series) through July 2022, we find that: Keep Reading

A subscriber suggested testing SentimenTrader’s Dumb Money Confidence model “that incorporates more than a dozen indicators that have a track record of cycling to extremes, and equating with ebbs and flows in sentiment among broad categories of investors.” To investigate, we transcribe monthly values of Dumb Money Confidence from the chart at the link and relate this series to monthly SPDR S&P 500 ETF Trust (SPY) total returns, calculated from the open on the first trading day after a Dumb Money Confidence date to the open on the first trading day after the next Dumb Money Confidence date. Using the specified data from the end of December 1998 (limited by the Dumb Money Confidence series) through the end of July 2022, we find that: Keep Reading

Do the Best Brands, as published annually by Interbrand based on net present value of predicted incremental earnings due to brand, offer superior investment performance due to pricing power and superior operating practices? In their June 2022 paper entitled “Is Buffett Right? Brand Values and Long-run Stock Returns”, Hamid Boustanifar and Young Dae Kang examine the investment performance of Best Brands. Best Brands companies must be global, have publicly available financial data, be visible and have the expectation of positive long-term profitability above the cost of capital). Up to 2007 (subsequently), Interbrand published Best Brand lists in July or August (late September or October). The authors each year reform a Best Brands portfolio limited to U.S. firms the first day of the month after publication, thereby excluding immediate announcement effects on stock prices. For stocks encompassing multiple brands (e.g., Google and YouTube for Alphabet), they map brands to stocks by summing brand values. Using firm characteristics, accounting data and stock prices for a broad sample of U.S. stocks during 2000 (the first Best Brands list) through 2020, they find that:

Keep Reading

A reader requested evaluation of the Fosback Index and its Ned Davis variant. The creators of these indicators argue that a high (low) ratio of cash equivalents to assets among equity mutual funds indicates strong (weak) potential demand for stocks. The Investment Company Institute (ICI) surveys mutual fund managers monthly (with a lag of about a month) to measure the aggregate equity mutual fund liquidity ratio (LR). Only past year-end values of LR are readily available. Norman Fosback adjusts raw LR based on current interest rates, reasoning that mutual fund managers have more (less) incentive to hold cash when interest rates are high (low). We adjust the effect of interest rates via linear regression of annual LR against year-end yield of the 3-month U.S. Treasury bill (T-bill). We then define the difference between raw and adjusted values as Excess LR and relate this variable to annual returns of the Fidelity Fund (FFIDX) as a proxy for U.S. stock market total performance. Using year-end values of aggregate equity mutual fund LR from the 2021 Investment Company Fact Book, Table 15, year-end T-bill yield and annual returns for FFIDX during December 1984 through December 2021 ( 36 years), 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