The Material Contracts Corpus

TL;DR

The paper addresses the need for large, public contract datasets to study contract design, language, and AI tooling. It constructs the MCC from SEC EDGAR filings (2000–2023), using automated agreement-type classification with a fine-tuned LLaMA-2 model employing LoRA, and extracts contract parties via RoBERTa-based NER with a robust alias-linking pipeline. Key findings include the dominance of employment and security agreements, Delaware as the predominant state of incorporation, and increasing contract length and readability over time, with IPO-driven spikes in shareholder agreements. The MCC provides a scalable, open resource for empirical contract research and for developing and evaluating AI-based legal applications.

Abstract

This paper introduces the Material Contracts Corpus (MCC), a publicly available dataset comprising over one million contracts filed by public companies with the U.S. Securities and Exchange Commission (SEC) between 2000 and 2023. The MCC facilitates empirical research on contract design and legal language, and supports the development of AI-based legal tools. Contracts in the corpus are categorized by agreement type and linked to specific parties using machine learning and natural language processing techniques, including a fine-tuned LLaMA-2 model for contract classification. The MCC further provides metadata such as filing form, document format, and amendment status. We document trends in contractual language, length, and complexity over time, and highlight the dominance of employment and security agreements in SEC filings. This resource is available for bulk download and online access at https://mcc.law.stanford.edu.

Figures (5)

• Figure 1: Count of Contracts by Year: This figure shows the number of contracts filed on EDGAR. The drop in 2023 is due to corpus collection terminating in the first quarter of that year.
• Figure 2: Count of Agreement Types by Year: This figure shows the number of agreements by agreement type by year. The dip in 2023 reflects the termination of the collection. The MCC's largest categories are employment and security agreements, reflecting the relative frequency of these agreements for public companies. The spike in shareholder agreements in 2021 coincides with a large increase in IPO activity.
• Figure 3: Median agreement length and readability by year: These figures show median agreement word count and Flesch-Kincaid grade level over time. Contracts have tended to increase in both length and reading complexity.
• Figure 4: Log-Log Histogram of Party Appearances by Agreement Type: This figure shows a histogram of party appearance by agreement type. The x axis represents the number of times an individual party appears in a contract for a given agreement type while the y axis shows the frequency of that appearance count. To adjust for skewness, the results are displayed on a log-log plot. All agreements types are highly skewed, as shown by the large number of parties featured in one contract.
• Figure A.1: Combined figure showing the median word count, Flesch-Kincaid grade level, and type-token ratio over time. M&A agreements demonstrate the longest average length of agreement type while services and supply agreements are generally the shortest (excluding filings identified as non-contracts).