When Do Contract Modifications Become Early Warning Signs? U.S. Federal Highway Construction Award Data, Analysis Code, and Replication Materials

Overview This repository provides publicly sourced U.S. federal contract transaction data, Python analysis code, computational results, and reproducibility materials supporting the research manuscript “When do contract modifications become early warning signs? Landmark evidence from U.S. federal highway construction awards.” The study investigates when post-award contract modification information becomes useful for identifying federal highway construction contracts with subsequent material increases in net federal financial obligations. It compares information available at original contract award with modification-related information observed at 90-day and 180-day landmarks, using chronological validation and fixed-capacity monitoring evaluations. Datasets The repository contains: Original dataset: 8,683 deduplicated U.S. Department of Transportation prime-award transactions obtained from USAspending.gov. Updated dataset: 9,822 deduplicated transactions, prepared on November 2, 2026, for the FY2024 holdout evaluation. Supporting documentation: Data dictionary, source-download provenance records, metadata, and SHA-256 file-integrity checksums. The primary analysis examines definitive highway construction contracts, including 521 contracts in the main modeling sample. The outcome is material positive net federal obligation growth during the 365 days following each observation landmark. Research Methods The computational framework includes: Contract transaction reconstruction and sample selection. Landmark-based analysis at 90 and 180 days after contract award. Comparison of original-award characteristics against post-award modification information. Chronological model validation and FY2024 holdout evaluation. Bootstrap uncertainty estimation and monitoring-capacity comparisons. Exploratory sensitivity analysis across additional observation landmarks. Code and Reproducibility The repository includes Python scripts, analytical outputs, and documentation to support independent reproduction of the reported research results. The package contains six main figures, one supplementary figure, and a graphical abstract, together with their associated output files. Detailed instructions, software dependencies, data provenance, and file descriptions are available in README.md. Interpretation and Limitations Positive federal obligation growth represents changes in recorded government financial commitments. It does not necessarily indicate verified construction cost overruns, construction delays, or project failure. The study population is concentrated in Federal Highway Administration and firm-fixed-price highway construction contracts. Model-generated monitoring rankings do not establish the effectiveness of actual managerial interventions. Findings from the relatively small FY2024 holdout should be interpreted with appropriate uncertainty. Data Source and Licensing The original transaction records were obtained from the publicly accessible USAspending.gov database. Original author-developed analysis code is distributed under the MIT License. Author-generated derived research outputs are provided under Creative Commons Attribution 4.0 International (CC BY 4.0), as documented in the repository's licensing information. Government-origin source data retain their applicable public-data status.