A newly released computational study has turned 21,629 Iowa intergovernmental agreements into a searchable picture of how cities, counties and special districts share services and move public money. The paper, submitted Sept. 16, applies document scanning, large language models and network analysis to the state’s unusually complete archive of so-called 28E agreements.
Those agreements exist because Iowa law lets public agencies exercise powers jointly, contract for cooperative action and create separate entities for shared work. The statute says agreements must spell out their purpose, duration, financing and termination method, and it requires filing with the secretary of state before an agreement takes effect. The state’s public database can be searched by participant, service type, filing date, organization type, county and region.
The research team said the archive covers filings from 1993 through 2020 across 33 service areas, including policing, fire response, roads, emergency management, water systems, education and economic development. Because many older contracts are scanned PDFs rather than machine-readable documents, the researchers first used optical character recognition, then classified each agreement as a service contract, resource-sharing arrangement, joint operation or new joint entity.
After testing several approaches against 1,128 hand-coded agreements, the researchers selected a classifier that reached 82% overall accuracy and a 0.92 F1 score for service contracts. They separately checked 100 extracted service contracts against the original documents and reported that 96% of the identified payers, providers and dollar amounts were correct. Those validation rates are important because the authors also note that automated legal analysis can misread noisy scans or complex contract language.
The resulting map shows a system with 1,642 standardized agency nodes. Cities were predominantly buyers of services, purchasing from counties in 1,577 of the city-originated contracts analyzed. Police protection, roads, public works and health were the most common services in those city-to-county arrangements. Counties occupied more varied roles, buying from both cities and other counties while also serving as major providers.
The network was concentrated as well as local. One group of 83 agencies received about $538.7 million in contract inflows, more than any other group identified by the model. At the same time, agencies were more likely to contract within the same county and service area than would be expected in 1,000 randomized versions of the network.
For Iowa officials and taxpayers, the practical value is not that an algorithm replaces review of the underlying contracts. It is that the archive can now be screened for concentrated provider dependence, unusual payment relationships and areas where local governments repeatedly buy the same services. Municipal guidance already emphasizes that every agreement must be filed electronically. The new work offers a way to examine that record at statewide scale while preserving the original filings as the controlling public documents.
