Investors and analysts often compare companies operating within the same sector, assuming that businesses in similar industries share important financial characteristics. However, a study examining all 500 companies in the S&P 500 suggests that conventional sector classifications capture only part of a company’s financial profile.

Researchers in Spain analysed fiscal year 2022 financial statements to investigate how closely companies’ financial structures corresponded with their assigned sectors. The analysis used accounting ratios covering several important dimensions of corporate performance, including profitability, leverage, liquidity, operational efficiency, and cash generation.

The researchers first examined how strongly these financial ratios differed across sectors. They then used seven machine-learning models to determine whether companies’ sectors could be predicted solely from their accounting information. If sector membership closely reflected financial structure, the models would be expected to classify companies with relatively high accuracy.

The best-performing model, K-nearest neighbours, achieved a validation accuracy of 49.3%. This was substantially higher than the 14.8% majority-class baseline, showing that accounting characteristics do contain meaningful information about sector membership. However, the accuracy remained too low for sector classifications to be considered a complete representation of companies’ underlying financial structures.

The researchers therefore turned to unsupervised machine learning, grouping companies according to similarities in their financial characteristics rather than their existing sector labels. “Hence, we used unsupervised learning to group firms by financial similarity rather than by their existing labels,” said corresponding author Ricardo Reier Forradellas of the Catholic University of Ávila. The approach identified nine economically interpretable groups of companies.

Each of the nine clusters contained companies drawn from more than one conventional sector, highlighting financial similarities that crossed traditional industry boundaries. The clusters also generally displayed less internal variation than standard sectors across most of the accounting ratios examined. Nevertheless, some familiar sector-specific financial patterns remained visible, with utilities, real estate, and financial companies more readily identifiable than firms belonging to several other sectors.

“Our findings do not mean that sector classifications are obsolete,” explained Forradellas. “They show that sectors tell only part of the story. When the aim is to compare companies by financial structure, accounting-based peer groups can provide a useful additional perspective.” Such groupings could therefore help investors and analysts identify financially comparable companies that may otherwise be separated by conventional sector classifications.

The researchers also compared company cluster assignments across subsequent annual reporting periods and found only moderate persistence over time, suggesting that financial peer groups can change as companies’ circumstances evolve. “This indicates that these peer groups should be updated rather than treated as fixed categories,” Forradellas added. The researchers conclude that accounting-based clustering can complement existing sector taxonomies, providing an additional tool for financial benchmarking, company peer comparisons, and broader financial analysis.

More information: Ricardo Reier Forradellas et al, Characterization of S&P 500 companies by sector using artificial intelligence: Statistical evidence and machine learning application, The Journal of Finance and Data Science. DOI: 10.1016/j.jfds.2026.100193

Journal information: The Journal of Finance and Data Science Provided by KeAi Communications Co., Ltd.