How Much of Peru Official Crime Trend Survives Endogenous Underreporting? A Partial-Identification Evaluation of the 2024-2026 States of Emergency

Working paper (v1 preprint). Peru declared recurrent states of emergency against organized crime in 2024-2026. Because a policing surge changes how crime is recorded, standard evaluations of police complaints conflate a change in true crime with a change in reporting. Using a district-month panel of police complaints (SIDPOL) and a verified inventory of emergency decrees, the effect on recorded extortion is about +15 percent (Poisson pseudo-maximum-likelihood, clean pre-trends), but the effect on true crime is not point-identified. We develop a partial-identification framework (Manski bounds with a clean lethal-violence anchor calibration, to be composed with honest difference-in-differences) and report the identified set: under the assumption that the emergency did not reduce reporting, the true increase is at most 15 percent and a genuine decline is not ruled out. The paper characterizes what can and cannot be learned given Peru crime data, and releases an open, reproducible, CPU-only pipeline.