Anomaly Rates After Reform: Evidence from Kenyan Public Procurement

Governments adopt digital procurement systems to curb misconduct, and anomaly detection offers a cheap way to check whether they work. A falling anomaly rate, however, can reflect better record-keeping as much as better behavior. This paper asks what moved when the share of flagged contracts in Kenya fell after two reforms in 2024, a digitization upgrade of the procurement portal and a capacity-building levy. I use contract-level data from Kenya’s Public Procurement Regulatory Authority for January 2024 to June 2025 and estimate the changes with interrupted time-series regression, since no untreated comparison group exists. The flagged share dropped from about 46% to about 9%. One flag accounts for most of the fall: contracts with no supplier record went from about 43% of contracts to under 1%. Flags unrelated to supplier records did not fall. On this evidence I read the drop as an improvement in record quality and find no evidence that other irregularities declined. The paper also offers a reproducible template for auditing audit metrics where data are scarce.