In an industry dominated by calculations and KPIs, sentiment analysis sticks out like a sore thumb in the investment sector. What is a metric based on emotions and opinions doing in a number-centric world? Yet despite this seeming incongruity, sentiment analysis in the asset monitoring world is quickly becoming ubiquitous.
This is largely because we live in a socially conscious society where a company’s business social index is almost as important as its product. As Abraham Lincoln put it, ‘With public sentiment, nothing can fail. Without it, nothing can succeed.’ In other words, sentiment matters. It’s no wonder then that sentiment analysis is projected to become a $4.3 billion industry by 2027.
So, what is sentiment analysis exactly? Is it just a bunch of new-fangled nonsense or can it add value to credit risk monitoring? More importantly, should a credit risk manager pay any heed to this metric? Let’s find out.
Sentiment analysis – A brief overview
What exactly is it?
Sentiment analysis has several interesting pseudonyms such as opinion extraction, subjective analysis, credit sentiment scoring, and emotion AI. Call it what you will but they all mean the same thing - the contextual analysis of data. Put simply, sentiment analysis is the process of determining if a sentence has neutral, negative, or positive vibes.
Over the past 2 to 3 decades, the world has become an increasingly connected place. In addition, the emergence of social media has given even the average Joe a platform to express his opinions freely – and express freely he does! Consequently, there is a wealth of information out there that companies can use to gauge the public’s sentiment towards them.
Similarly, financial institutes can use these values for the early identification and monitoring of risk-prone investments within their portfolio. For example, restaurants need good reviews to thrive. In fact, according to TripAdvisor, over 90% of customers choose a restaurant depending on their online ratings. Negative reviews can therefore sound the death knell for an eatery. So, it's a good idea for lenders with vested interests in such businesses to use sentiment analysis software to track online reviews. That way, they are aware of waning public sentiment towards their investment and can then make well-informed business decisions regarding it.
How sentiment analysis is done
One of the earliest instances of pundits using sentiment analysis dates back to 1824 when a public sentiment poll incorrectly predicted Andrew Jackson as the winner of the upcoming presidential election. The incorrect prediction was probably due to inadequate data and human error. Today, given the mountain of data at our disposal, AI-backed models are much more capable than human minds at sentiment analysis as they can scan large datasets in a matter of minutes.
These models use data collected from news outlets, press releases, various social media channels, company transcripts, and so on. It is important to note that the information gathered is publicly available material, so no privacy laws are being breached. The AI models scour the data to identify credit-impacting news. They then identify the credit sentiment behind this information to produce insights that asset managers can act upon if needed.
Can sentiment study predict a company’s fortunes?
That public sentiment can affect the financial sector is a well-documented occurrence. This is why stock markets usually perform well during Thanksgiving and Christmas. When the public mood is good, so is stock performance. Another prime example of this is the spike in market volatility seen every time a presidential election happens in the US. This is reflective of the jittery investor sentiment that is pervasive then.
So, the connection between sentiments and market performance is clear. But can
sentiment study actually predict how a company performs in the future? If done correctly, yes! And therein lies the utility of sentiment analysis for the credit risk monitoring industry. Negative sentiment values can often foretell a company’s downward spiral.
For example, our credit risk-monitoring software, TRaiCE, recently performed sentiment scores on a cross-section of companies from May 2020 to April 2021. Here are the results:
As you can see, companies such as Greensill and Hin Leong have a lot of negative sentiment scores during this time. And as it turns out, both companies went on to face financial collapse soon after.
Greensill, a financial services company, famously imploded in March 2021 dragging down investors such as Credit Suisse with them (read our article on the Credit Suisse debacle for more details). The oil trading company, Hin Leong, collapsed late last year. Singaporean courts recently tacked its founder, O.K. Lim, with over 20 charges of fraud. To make matters worse, PWC is suing the company for close to $4 billion.
For both companies, the writing was on the sentiment-analysis wall before the collapse happened.
The flip side of sentiment analysis
As impressive as the results can be, there is a flip side to sentiment analysis. If done incorrectly, it can produce inaccurate results. Mined media is often ambivalent in nature. In addition, the English language (wondrous as it is) has words that can mean contrasting things in different circumstances. For example, the word ‘tears’ usually has a sad connotation to it. However, when used in the term ‘tears of joy', it means unbridled happiness.
Similarly, the term ‘Covid-19’ understandably has an overwhelmingly negative undertone to it. However, not all media releases using that term are negative. For instance, a report titled ‘Krispy Kreme offers free donuts to customers with Covid-19 vaccination certificates’ is positive news. These dimorphisms, while easy for our human brains to understand and differentiate between, can be tricky for computer systems. As a result, non-augmented analytical programs can produce a lot of false-positive and negative results.
A humorous example of this is the ‘Hathaway Effect’. Between 2008 and 2011, stock prices of the holding company Berkshire Hathaway soared by around 3% every time the actress Anne Hathaway made the news. This despite the fact that the actress is in no way connected to the company. Experts blamed this ‘phenomenon’ on faulty sentiment-analyzing trading programs that picked up reports on ‘Hathaway’ and incorrectly applied them to the stock market.
The key to effective sentiment analysis
The key here then is to have an application that can understand the context and see the intent behind the media rather than just picking up on word connotations. This is where TRaiCE stands head and shoulders above the rest. The system uses regularly augmented, advanced ML algorithms to pick up on the intent and entity sentiment behind the data it sifts through. What’s more, TRaiCE monitors public information 24/7/365. It then combines this with other relevant data to set up alerts that can warn credit risk managers of potential risks.
Conclusion – Is sentiment analysis hype or hope?
So, should an asset manager pay attention to sentiment analysis? If it is with a system like TRaiCE that can understand the business sentiment and the corporate impact of the data, then yes. If not, they would be well advised to take the analysis with a pinch of salt. It should also be remembered that sentiment analysis is just one piece of the asset monitoring puzzle and not the whole. As such, it is an important metric that can act as a forecaster of company health and should not be ignored.
Disclaimer
The information described in this report is provided to you “as is” and “as available” without warranties of any kind, expressed, implied , or otherwise, including but not limited to all warranties of merchantability, fitness for a particular purpose, or non-infringement. The content of this mailer is intended for general information purposes only and is prepared based on publicly available information that has not been independently verified by Menerva. Menerva shall not be liable for any loss whatsoever arising from any use of the information in this mailer. The information contained in this report is under no circumstances intended to be used or considered as financial or investment advice, a recommendation or an offer to sell, or a solicitation of any offer to buy any securities or other form of financial assets. Readers are encouraged to not use the information contained herein in making any financial or investment decision and must obtain independent financial advice from a professional in connection with or independently research and verify, any information that is found in this mailer and relied upon, whether for purpose of making an investment decision or otherwise.
Comments