Assessing Merger Guideline Feedback With Machine Learning


This article analyzes the results of large language model processing to reveal several important patterns in the comments to the draft merger guidelines.

In December 2023, the U.S. Department of Justice (DOJ) and the Federal Trade Commission (FTC) released new merger guidelines, substantially updating their stated approach to mergers. The release of these guidelines concluded a process that began in the summer of 2022 with the release of a request for information (RFI) that asked for input on a wide range of questions and priorities related to merger enforcement, and continued through the release of draft merger guidelines (DMGs) in July 2023.

A total of 1,906 comments were submitted after the release of the 2023 DMGs, of which 1,689 could be classified as coming from members of the general public. These comments did not offer technical suggestions for the guidelines but instead pointed to general areas of concern, such as higher prices, or specific industries, such as telecoms.

Modern machine learning techniques, including large language models (LLMs), allow us to examine these comments to determine whether commenters favor greater enforcement and what industries or topics they believe need greater scrutiny. Cornerstone Research’s Data Science Center professionals quantized this model to run efficiently from entirely within our secure data center. A review using this LLM technique showed that more than 90% of comments supported greater enforcement, while only 2% could be classified as supporting less enforcement.

In this article, Andrew Sfekas, with help from the Data Science Center, applied this LLM technique to reveal several important patterns in how and where commenters wanted greater merger enforcement.

The article was originally published in Law360 in February 2024.

Assessing Merger Guideline Feedback With Machine Learning


  • Washington

Andrew Sfekas

Senior Manager