Sexual and gender-based violence (SGBV) is the most under- reported category in the BPO record. Several reasons converge: survivor reluctance to file a police complaint, family suppression, social stigma, weak prosecutorial channels, and — at the data layer — fewer press reports relative to the underlying incidence. The desk codes 23,238 incidents under the SGBV cross-cutting theme since 1991. Every SGBV researcher consulted on the BPO methodology has flagged that count as a fraction of the true incidence, probably a small fraction.
This story is about what the geographic distribution tells you when you read it as a reporting map, not an incidence map.
What the district-level pattern looks like
SGBV incidents in the record cluster in Dhaka division (about a third of the total), with secondary concentration in Chittagong, Rajshahi, and Khulna divisions. Per-capita rates across smaller districts look strikingly low — which would be implausible if interpreted as incidence rates, but is consistent with media-access rates: districts with stronger local press correspondents produce more SGBV stories per underlying incident.
The pattern is also tilted toward urban upazillas within each district. Rural sub-districts show extremely low counts. Researchers with field experience in these areas universally report this is a reporting artifact, not a real distribution.
Three changes since 2018
Three things have shifted in the post-2018 SGBV record:
1. A general increase in coding — partly reflecting the social-media-mediated visibility of cases the press wouldn't have covered before.
2. A specific spike around the 2018 anti-rape protests and the 2020 women's rights movement, both of which drove coverage of cases that would otherwise have stayed local.
3. An uptick in arrest counts coded against SGBV incidents, reflecting both new legal instruments (the Women and Children Repression Prevention Act amendments) and the public-pressure cost of inaction in high-profile cases. Arrest counts are not conviction counts, and the BPO record does not currently follow cases through the courts.
How to read the SGBV record
Three rules of thumb:
One. Don't compare SGBV counts across districts as if they were incidence rates. They aren't.
Two. Year-over-year changes within the same district are more meaningful than absolute counts — the reporting bias is more stable per-district than across districts.
Three. Read the SGBV record as a partial, biased sample of a much larger underlying phenomenon, useful for trend analysis and case-finding but not for incidence estimation. The honest methodological footnote is that for prevalence estimates, the BPO record is the wrong tool — survey data is the right tool, and several surveys exist.