Police records show what enters the justice system. Interpreting them well requires attention to both the events being reported and the institutions that record them.
- Observation coverage
- 2020-01 to 2026-08
- Source build date
- Not recorded by the source; see coverage above
- Website refresh policy
- Twice monthly, on the 2nd and 17th
Published snapshot. A scheduled check does not imply that a new source observation is available.
Loading the published dataset…
Reading the evidence
01
Recorded cases and experienced crime are different measures
A case count depends on an event occurring, being reported, being registered and being classified. Changes in access to police, trust, enforcement priorities or registration practices can alter counts even if underlying victimisation is unchanged. A rise can reflect more reporting as well as more incidents; a fall can reflect barriers to reporting. Population surveys and administrative records answer complementary questions.
02
Compare like periods and like categories
Offence categories differ in how they are detected and reported. Combining them may allow common, enforcement-sensitive offences to dominate a total. Start with a particular category, inspect the monthly sequence and compare similar periods. A partial year should not be treated as a full year's outcome. Abrupt changes deserve a check of publication coverage, administrative disruption and classification before an explanation is assigned.
03
Population adjustment has geographic limits
Large reporting units usually record more cases because more people live or work there. Rates help put counts in context, but the denominator must match the population at risk. The optional rates here use fixed 2022 population proxies from the source extract. Police jurisdictions, commuting populations and census boundaries need not coincide. Railway Range has no comparable resident denominator, so its cases are excluded from combined rates and displayed separately in counts.
04
A reconciled transcript is a starting point
Printed row and column totals can identify many OCR mistakes. The source pipeline keeps a correction log and distinguishes months re-verified from original sheets from the earlier cleaned compilation. Those checks improve transcription reliability, but they cannot reveal unreported offences or validate the substantive classification of a case. The published correction table is part of the evidence, not a claim that the administrative record is complete.
Methods, sources and reuse
Monthly arrays retain offence and reporting-unit dimensions. Dacoity, robbery and burglary are combined only in the display; their source columns remain intact. Selected counts sum only available cells, while a missing cell makes the corresponding total unavailable rather than zero. The division map uses the eight Range jurisdictions; city markers show metropolitan forces separately, and Railway Range has no territorial or population geometry. Monthly and selected-period rates use fixed source-extract 2022 populations and are annualised to twelve months. The government-period comparison begins in 2021: August 2024 is attributed to the interim period except Murder, and February 2026 is split equally. New source data with unresolved transcription cells is rejected.
- 1. Bangladesh Police: Crime Statistics
Primary monthly and annual source sheets. Source availability can vary.
- 2. Bangladesh Police: May 2024 source sheet
Example of the published reporting-unit and offence-category table.
- 3. Transcription, reconciliation and geography pipeline
Source correction history, population proxies and unit mappings.
Download and cite this snapshot
Institute of Policy Dynamics, “Bangladesh Crime Statistics: Recorded Cases and Reporting”. Coverage: 2020-01 to 2026-08. Cite the original data providers alongside IPD.
Download the complete published dataset (JSON)Source fingerprint (SHA-256): 0638bd07be78a78db58e57ec4b69e736cc10fab77fa8ada07fed5bd1d7c15297
Chart CSV files contain the selected observations. Original data remain subject to their providers' terms; public access alone does not establish an open licence.