Warren Buffett advised never to invest in a business you don’t understand. But that didn’t stop many investors.
A new study from the McCombs School of Business at the University of Texas at Austin may help us better understand the complexities of portfolio companies. This study provides the most accurate and comprehensive tool to date for measuring business complexity.
The tool was devised by Sarah Toynbee, associate professor of accounting, and uses artificial intelligence to simplify measurement. It also turns out that in areas such as debt structuring, complexity can sometimes be a good thing.
She defines complexity from an observer’s perspective as how difficult it is to understand a company’s financial health and performance based on information in its financial reports.
The problem with measuring complexity is that it is very complex.
Traditional metrics such as company size and number of operating segments miss deeper levels of complexity in operating, risk and financial structures, she says.
“Business complexity is a very elusive concept because the sources of complexity can vary from company to company,” Toynbee says. “Our model captures that across many dimensions.”
She breaks down business complexity into 29 categories, including debt, equity, derivatives and hedging, income taxes, revenue, and compensation.
Classification of complexity
Together with Darren Bernard, Elizabeth Blankespoor, and Ties de Kok at the University of Washington, Toynbee trained a large-scale language model, a version of Meta’s Llama 3, on 200,000 sentences from the company’s financial footnotes.
The sentence contained an embedded iXBRL tag. This is a label that is invisible to human readers but visible to computers. They describe the “meaning” of numbers and allow models to be trained to predict what the numbers represent.
After the researchers trained the model, they had it classify more than 8 million personal numbers from more than 50,000 company reports from 2016 to 2024.
“This is like asking a sophisticated human being to read millions of sentences and explain in one phrase what a number means based on the surrounding text,” Toynbee explains.
The harder a number is to classify, or the less confident the model can calculate it correctly, the higher its complexity score.
Benefits of complexity
It turns out that complexity impacts business in both negative and positive ways.
- Stock price deceleration. Stock prices react more slowly to filings for companies with higher average complexity scores, suggesting that investors need more time to understand the information. The most complex report took 7.9% longer for prices to fully react than the least complex report.
- Debt stabilization. Complex debt structures are often seen as risky, but researchers have found that they can help manage risk. Debt with non-standard terms, such as the ability to convert into equity, is more complex. However, it allows companies to better manage risk and provide more predictable interest payments.
“People usually think complexity is a bad thing, but we’re showing that in some cases taking on complex debt can actually be beneficial for companies,” Toynbee says. “This provides more stability and sustainability for some businesses.”
Toynbee says the model could have a variety of uses. Investors can use this to focus on complex companies that require more in-depth analysis.
Standard setters and regulators could use this to identify categories of overly complex financial information, she says. “We may consider ways to simplify or strengthen reporting standards to help investors better understand a company’s financial health and performance.”
Finally, companies may use this tool to curb excessive complexity. “Some complexity can be beneficial, but managers may not realize that their area of business is more complex than others,” Toynbee says. “By highlighting these areas, you may find ways to simplify them.”
“ Using GPT to measure business complexity ” is published online in The Accounting Review.
/Open to the public. This material from the original organization/author may be of a contemporary nature and has been edited for clarity, style, and length. Mirage.News does not take any institutional position or position, and all views, positions, and conclusions expressed herein are those of the authors alone. Read the full text here.
