Research Summary: Making Local Government Contracts Legible: A Computational Pipeline for Classifying and Mapping Intergovernmental Service Agreements

Research details a new computational pipeline designed to classify intergovernmental service agreements and extract critical institutional and financial data at scale. Applied to a substantial dataset of local government contracts in Iowa, this method leverages advanced Large Language Models (LLMs) to overcome prior challenges in systematically analyzing such agreements, which are fundamental to public service delivery.

Why it matters

This development is strategically important as it addresses a long-standing challenge in understanding intergovernmental collaboration and resource allocation. By enabling systematic analysis of service agreements, it provides unprecedented insights into the operational and financial mechanics of public service delivery across jurisdictions, fostering improved governance and efficiency.

Key insights

• Interlocal agreements, crucial for formalizing public service collaboration, have historically been difficult to analyze systematically due to their inaccessible institutional and financial content.
• A novel end-to-end computational pipeline has been developed to classify intergovernmental agreements based on institutional form and extract financial relationships between contracting parties.
• The pipeline integrates LLM-based summarization and classification, utilizing models such as LLaMA 3.1, GPT 5.2 Pro, and Gemini 3 Pro.
• This methodology was successfully applied to Iowa's 28E archive, which comprises 21,629 interlocal agreements, demonstrating its capability for large-scale data processing.
• The pipeline performs a four-class classification task on these agreements, indicating its ability to categorize different types of intergovernmental contracts.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2609.19225