This stock sentiment report will be analysing Novartis AG’s stock for the week commencing 20th September 2021, using StockGeist.ai API data. The report has been written by Finance Focused. The StockGeist platform determines the market sentiment indicators towards a stock in real-time by applying natural language processing to the latest news updates and social media posts. All times given are in GMT+0.
Novartis researches, develops, manufactures and markets medicines, diagnostic tests and other devices which assist in improving patient outcomes. Originating/based in Switzerland, and now global, Novartis has seen significant success since branching out into the pharmaceutical industry. Presently focusing on advanced therapy platforms and data science, a technological outlook suggests a modern USP that could, in the long-term, improve Novartis’ currently volatile horizontal stock price.
Beginning with an overview of NVS stock’s 2020 performance, -10% in operating net cash flow seems bleak; despite growing operating income, payments related to legal matters and lower divestment proceeds overshadowed any positives for the company’s cash flow. If legal matters persist, and divestment proceeds continue to diminish, Novartis’ measly 2.4% growth in the past 5 years will likely not be enough to prevent its price dropping in the next few years – to reflect an uncertain and unsettling company future.
Although, net income improving 86%, and 84% EPS accretion (though this could only be to calm investor nerves after NVS’ volatile underperformance), contradict negative sentiment from the fundamental FCF analysis above. This explains volatile price movements in NVS stock over 2020 (and 2019). However, such growth may be unable to continue as the TTM NWC has recently dropped negative to -$3,386,000,000; an unsustainable long-term figure suggesting Novartis cannot afford to continue creating new growth options – and relying on its current medicines, though performing well, will provide little to no growth.
Now, work is ongoing searching through new options for lung cancer patients, furthering a buy indication as this could provide a profitable growth option should Novartis develop a successful advanced medicine in the lung cancer field. More specifically, most recently, since the 27th September, NVS stock has rebounded +0.47% from opening at $82.50. This is likely due to the most recent media surrounding Novartis: new data reinforced efficacy and convenience of its Cosentyx® 300mg autoinjector in adults with psoriasis, and Novartis’ medicines for malaria venture reported positive results for their phase 2b study focusing on children with malaria.
We assessed the impact of incoming emotional and information messages on the price of Novartis AG. The assessment was done by gathering sentiment data using StockGeist’s REST API, proposing and testing a model using data from the week commencing 13th Sept and then validating the outcomes using data from the week commencing 20th Sept.
Initially, we made stacked area plots of the incoming messages to get a deeper insight into the stock’s situation. From Fig 1, we can see that there are not enough emotional messages to create a model around as too many of the hourly emotional values are zeros. Therefore, the analysis will only focus on incoming information messages.
The proposed trading strategy using sentiment data involved two signals. An initial entry signal if the normalised exponential moving average (EMA) is above 0.9, and then an evaluation is made based on the sentiment of the current hourly sentiment data. We chose the EMA over the API 24hr MA because the EMA gave clearer spikes in a short term window. If over 70% of the incoming information messages were positive or negative, a buy or short signal was triggered.
Finally, changes in the stock prices for the model were evaluated and parameters were optimised to get the best % changes after a buy or short signal. If the buy/sell signal occurred outside of market hours, then we found the next valid timestamp.
The analysis will only cover the results of the validation stage, but the same can be applied to the modelling and testing stage results. Below are two tables with the stock information for when a buy/sell signal was triggered along with the percentage changes in the price of the stock after 30, 60 and 90 minutes. The following will break down the events hour by hour in the validation stage where the signals described in the methodology section were triggered.
First, we see two opposing out of market signals triggered on Sept 21st. Since they both would then result in buy and sell signals when the market opens, and this scenario wasn’t considered during the methodology, these two signals were discounted. Naturally, a lot of events can occur outside of market hours. Therefore, further analysis of the overall sentiment between the signal and the market opening would be more proper, rather than choosing one signal over the other.
Next, on Sept 23th a spike in the data before market hours was recorded. Since over 70% of the total information messages were negative, it suggested a sell. However, the trade didn’t succeed as every 30 minutes after the signal the percentage change remained positive. On the same day, we get another high influx of messages at 4 pm. The sentiment towards the stock was positive, suggesting a buy, but this time, the stock price kept fluctuating in the next 90 minutes.
The day after, on Sept 24th, a final increase in information messages was recorded. On this occasion, the sentiment of the articles was positive, suggesting a buy. What followed was a further decrease in the stock price.
After proposing a trading strategy using data from the platforms REST API, the performance of the trading strategy turned out to be very weak. The predictions based on the sentiment data were incorrect most of the time, and the changes in prices were insignificant when compared to the moving average percentage change of the stock price. Therefore, there is not enough evidence to suggest that using the tone of daily fluctuation of information messages affects the price of the chosen stock.
The accuracy test at first can be viewed as negative, however, the fluctuations in the stock messages are small, from 1-12 articles per hour. In other stock analysis reports such as Qualcomm or NIO Inc, we were dealing with daily fluctuations between 0-10 messages per hour and a sudden sharp spike to 80 – 120 messages at one point during the week which we don’t see in Novartis AG messages during the two weeks that were analysed. It is somewhat reassuring to see the conclusion of the accuracy test, as an increase in only 4-5 messages overall in a recorded spike doesn’t seem to affect the stock price, which is a reasonable outcome.
Overall, using daily fluctuations in sentiment data could not result in a profitable trading strategy since these fluctuations are only between 1-10 articles which are small compared to the moving average of information messages.
Despite the negative factors discussed earlier in the ‘Historical Analysis’ section such as unsustainable NWC, recent sentiment does have positive aspects, including a forward P/E ratio of 14.93 (indicating undervaluation). Further, the launch of new drugs, an example being multiple sclerosis drug Mayzent, is expected to boost sales. Because of multiple new drug launches, risk is mitigated; not to mention that both recent launches are expected successes (the other is a breast cancer drug named Piqray). This provides evidence for a current strong buy rating. Moreover, the overall professional analysts’ average price target is $105.79 (courtesy of stocknews.com), 27.9% above NVS’ current stock price.