Do Media Sentiments Affect the US Stock Market?30 june 2022
Sentiment analysis as a method of quantitative estimation of opinions and emotions in texts is widely used not only in various marketing studies, but also in trading.
Nowadays, there are many indices to measure sentiment, such as the ZEW Expectations, the Michigan Consumer Sentiment Index (MCSI), the Conference Board US Consumer Confidence Index, the Bloomberg Consumer Comfort Index, the Speculative Sentiment Index (SSI), the SWFX Sentiment Index, and others.
The calculation of these indices is based on various surveys of the population regarding the economic situation, taking into account today and the near future.
In turn, Robert Shiller in his book "Irrational exuberance" on the behavior of prices in the stock market writes: "Nothing affects the stock market as often and as strongly as the news that interests it".
John Samma, in “Trading Against the Crowd”, refers to the impact of investor sentiment on the market:
“Experts' exaggerated optimism means that there are very few people left in the market who want to open new long positions, if we assume that investors are willing to listen to all this advice. On the contrary, a too high degree of "expert" pessimism speaks in favor of a few bears in the market, from which we can conclude that there is a high probability of a short squeeze and a close start to the rise. The best signals are obtained at excessive levels, as is the case with other indicators”.
In his book, he conducted “testing expert and individual investor opinion":
“I ran the trading system on each of the three sources of information: Investors Intelligence, Bullish Consensus, and the American Association of Independent Investors (AAII) survey.”
The test was conducted in conjunction with the data of the S&P 500 stock index.
We will test the Daily News Sentiment Index along with the DOW 30 and S&P 500 stock indices.
At the maximums of positive news background, the stock market unfolds downwards, at the minimums, where mass sentiment can be described as “everything is lost”, the market unfolds upwards.
The Daily News Sentiment index is described in the book “Measuring News Sentiment” by Shapiro, Adam Hale, Moritz Sudhof, and Daniel Wilson.
The Daily News Sentiment index is based on a lexical analysis of texts related to economics, and was built as a moving average weighted value of time series with weights, which decrease with increasing time since the article was published.
To calculate the news sentiment index, fixed regression effects are calculated for the articles:
Strategy (according to D. Samma):
We construct the EMA1-4W oscillator based on the data from the Daily News Sentiment index.
The EMA1-4W is constructed as the difference between one-week and four-week exponential moving averages.
To buy on tomorrow’s Open if:
- the EMA1-4W oscillator value is negative today;
- the EMA1-4W oscillator value today is 5% greater than its two last weeks’ value;
- this week’s close is greater than last week’s high.
To sell on tomorrow’s Open if:
- the EMA1-4W oscillator value is positive today;
- the EMA1-4W oscillator value today is 5% less than its two last weeks’ value;
- this week’s close is lower than last week’s low.
Closing a position
- 1, 2, 3, 4 weeks after opening;
- on the opposite signal.
Daily data of Daily News Sentiment Index
Period: 01/01/1980 – 06/06/2021
Daily quotes of indices (for calculating the correlation coefficient):
- DOW 30. Period: 04/08/2002 – 06/17/2021
- S&P 500. Period: 01/04/2006 – 06/17/2021
Weekly quotes of indices (for calculating the rate of return of the strategy according to D. Samma):
- DOW 30. Period: 02/08/1985 - 06/17/2021
- S&P 500. Period: 01/04/1980 - 06/17/2021
There are 8,708 values in total.
Market entries: 214.
Analysis of the obtained results
First, let's trace the relationship between changes in the Daily News Sentiment index and changes in the value of the DOW 30 and S&P 500 stock indices. Also consider the correlation in shifts: 5 days ago and forward.
The values of the correlation coefficient in the absence of shifts and with shifts do not reach the value of 0.1% modulo. This means that there is no linear relationship between changes in the Daily News Sentiment index and changes in the value of the DOW 30 and S&P 500 stock indices.
Next, let's test D. Samm's strategy for the Daily News Sentiment index.
We will evaluate the results according to the following criteria:
- The rate of return reflects the relative change in the quotations of financial instruments in percentage. A positive value of the rate of return indicates the profitability of the strategy, negative - about the loss.
The rate of return (R) of a financial instrument is calculated using the formula:
n - the number of transactions;
P (%) – the increment percentage of a financial instrument quote, at the time of fixing the position, is calculated as follows:
for buy positions
P (%) = (position closing price - position opening price) / position opening price * 100%
for sell positions
P (%) = (position opening price - position closing price) / position opening price * 100%
The results are presented in the diagrams:
The rate of return of the D.Samma strategy for the Daily News Sentiment index exceeds a significant value of 0.3% only for exiting the market on the opposite signal.
For the DOW 30 index, its value is 2.96%, for the S&P 500 index - 3.46%.
But here, too, there is a nuance. The average waiting time for an opposite signal to exit the market is 10 months.
There is no linear relationship between changes in the Daily News Sentiment index and changes in the value of the DOW 30 and S&P 500 stock indices.
The rate of return of the D.Samma strategy for the Daily News Sentiment index exceeds a significant value of 0.3% only for closing positions on the opposite signal. However, it will take an average of 10 months to wait for the formation of such a signal.
The effectiveness of the Daily News Sentiment index for the US stock market forecasting has not been identified.
Detailed results are presented in the appendix: