Research analysis depends on our beliefs.
Among the factors we consider as a fundamental investor are evaluating a company’s strategy, products, supply chain, employees, financing, operating environment, competition, management, adaptability, and so on. Investment professionals conduct these assessments to enhance our assessment, yes, but to increase our confidence in the data and those whose activities measure the data. If we can’t trust the data and who created it, we won’t invest. In short, we must Faith Management
Our fraud and fraud detection methods are OK.
But by what repeatable method can we assess the credibility of companies and their people? Usually the answer is some combination of financial statement analysis and “trust your gut.” Here’s the problem with that:
1. Time and resource constraints
Companies exchange information through words rather than numbers. For example, from 2009 to 2019, the annual report of the component companies of the Die Jones Industrial Average equals 31.8 million words and numbers, according to AIM Consulting. The numbers are only 13.5% of the total.
Now, JP Morgan’s 2012 annual report is 237,894 words. Suppose an average reader can read and understand about 125 words per minute. At this rate, it would take a research analyst about 31 hours and 43 minutes to read the report thoroughly. The average mutual fund research analyst in the United States earns about $ 70,000 per year. WallStreetMojo. So the cost of a firm to evaluate a JP Morgan report is more than $ 1,100. If we had already invested in JPMorgan, we would have done a lot of this work to ensure our confidence in the company.
Moreover, quantitative data is always disclosed publicly at a significant time interval. Since the performance of a company is usually published quarterly and annually, the average time lag of such data is slightly less than 90 days. And once the data becomes public, the benefits it offers are quickly sold out. Most investment research teams lack resources to evaluate in real time near every company or portfolio in their universe, or just after the publication of quarterly or annual reports.
Conclusion: What is that old line? Oh, yes: time means.
2. Believing in our gut does not work.
In contrast to pan-cultural fiction, research shows that we cannot detect deception through body language or intestinal instincts. In fact, a meta-analysis of our fraud-identified capabilities has found the global success rate to be only 4% better than the chance. We can believe that as financial professionals we are exceptional. We will be wrong.
In 2017, we measured fraud detection skills among finance professionals. This was the first time our industry’s ability to detect lies was tested. In short: Alas! Our overall success rate is actually worse than the general population: we didn’t score 54%, we achieved a one-coin-toss worse than 49.4%.
But maybe our strength is in our own sector. Put us in a finance setting, say an earnings call, and we’ll do much better, right? No, not really. In investment settings, we can detect fraud only 51.8% of the time.
Here are two more news stories (sorry sorry): Finance professionals have a strong true bias. We rely more on other finance professionals than we do. Our research has shown that we lie only in .4..4% funding. So that the 51.8% accuracy rate is due to the tendency to trust our fellow finance professionals.
Another issue: when evaluating statements outside our domain, we have a strong 64.9% fraudulent bias. Again, this speaks to the innate sense of exceptionalism in our art. In previous studies, our researchers found that we believe we are told 2.14 lies every day Out of work Settings, and lies just 1.62 per day At work Settings. It again speaks of the bias of truth in meaning.
Finally, we believe that we can detect lies in money at a rate of 68% accuracy, the actual 51.8% has not been measured. Friends, this is the definition of overconfidence and it is an illusion in another name.
Conclusion: We cannot believe in our courage.
3. Strategy audit number of auditors.
But what about the auditors? Can they properly evaluate the company’s authenticity and save us both time and money? Yes, company reports are monitored. But auditors can only conduct their analysis through a small sample of transaction information. Worse, the strategies of auditors like us focus more on 13.5% of data that is too small to contain numerically. This excludes 86.5% of text-based content.
Further, because financial statement analysis – our industry fraud detection strategy – has moved one step away from what auditors see, it is rarely reliable. In fact, financial statement analyzes are just part of the table: we probably won’t be much different from our competitors. Looking at the same number as others is less likely to create a forgery or alpha.
And what about the private market? In recent years the investment research community has spent a lot of time looking for investment opportunities in that area. But when private market data is occasionally audited, their public market participants do not have the means to exert due diligence and additional application of trading activities. These can sometimes signal deception and deception.
Conclusion: We need to have another tool to help us fight fraud.
Analysis to retrieve scientifically based text
James W. Computers extract language features from text, such as word frequency, psychological description, or negative financial terms, in fact, dust for language fingerprints. How do these automated strategies work? Their success rate ranges from 64% to 80%.
In private conversations, as we mentioned, people can detect about 54% of lies. But their performance is even worse when assessing the authenticity of the text. Studies published in 2021 show that people have a 50% chance of detecting text fraud or currency-reversal. A computer-based algorithm, however, had a 69% chance.
But of course adding people to the mix improves accuracy? Not at all. Our overconfidence as investors undermines our ability to catch fraud even in human-machine hybrid models. The same researchers discovered how human subjects evaluated computer fraud judgments that they could then reject or change. When humans can surpass, computer accuracy drops to just 51%. When human factors can change computer judgment in a narrow range around the evaluation of algorithms, the hybrid success rate drops to 67%.
Computer companies can offer investors a huge advantage in evaluating the authenticity of communications, but not all fraud detection methods are suitable for one size fits all.
In a computer-driven text-based analysis published in 2011, 10-Ks had the ability to predict stock prices negatively for companies that included a high percentage of negative words. By scanning documents for words and phrases associated with the tone of financial communication, this method has detected elements that could indicate fraud, deception, or poor financial performance in the future.
Of course, businesses whose share prices were hurt by this strategy. They have completely removed offensive words from their communication. Some executives have even hired speech instructors so they never speak. So word-list analysis has lost some of their brilliance.
Where do we stay from here?
It can be tempting to dismiss all text-based analysis. But that would be wrong. After all, we don’t throw away financial statement analysis, do we? No, instead we should search and apply text-based analytics that work. This means that the methods are not easily deceived, which evaluates How language is used – Its structure, for example – no No language is used.
With these issues in mind, we have created fraud and truth analysis (data) with Orbit Financial. Based on 10 years of research into deceptive technology that works inside and outside the sample ই hint: not reading body language D DATA examines the fingerprints of more than 30 languages on five separate scientifically proven algorithms to determine how these speech elements and language fingerprints interact. Each other.
The process is similar to the standard stock screener. That screener identifies our performance fingerprints and then applies these quantitative fingerprints to screens across the entire universe and makes a list on which we can publish our financial analysis. Data works the same way.
Fingerprint in a native language is the use of articles such as a, an, and so on. One of these is associated with deception rather than an extra truthful statement. But the frequency of the article is only one factor: how the articles are used is really important. And since articles are directly linked to nouns, it is difficult to transcend data. A potential divider needs to change the way they communicate, how they will use their nouns, and how often they will use them. This is not an easy task and even if successful, only a single data language will resist fingerprinting.
Other key results from recent data tests include:
- Save time and resources: Data evaluates 0.400,000 words per second, or the equivalent of a 26-page book. It saves 99.997% time on people and saves more than 90% cost.
- Fraud accuracy: Each of the five algorithms is measured at the rate of fraud detection accuracy far above what people can achieve in text-based analysis. Moreover, the combination of five-algorithms makes the data difficult to work with. We estimate its accuracy to exceed 70%.
- Fraud prevention: DATA can detect the 10 biggest corporate scandals of all time – suppose Satyam, Enron – with an average time of more than six years.
- Outstanding Performance: In a data test, we measured the fraud of each component of the Dow Jones Industrial Average each year. The following year, we bought everything except the five most fraudulent Dow companies. From 2009 to 2019, we repeated the exercise at the beginning of each year. Despite an occasional nine-month delay in implementing the strategy, the strategy yields an average annual return of 1.04%.
The writing is on the wall. Text-based analysis that uses computer technology to detect fraud and deception and saves significant time and resources. Future articles in this series will further detail the results of data testing and win the basic analysis that makes such technology possible.
If you liked this post, be sure to subscribe Entrepreneurial investors.
All posts are the author’s opinion. As such, they should not be construed as investment advice, or the opinions expressed do not necessarily reflect the views of the CFA Institute or the author’s employer.
Image Credit: Getty Images / Broadcaster