BD-INV-002 · Version 1.6 · March 2026

The
Impunity
Machine

Illustration representing Bangladesh's rape justice crisis — women seeking accountability in a system that produces impunity — Cover image · BD-INV-002 · Insight Gaps · March 2026

She is somewhere in Bangladesh right now.

She is in her twenties. She lives in Barishal Division — where 81.5 percent of women will experience violence in their lifetimes, the highest rate in the country. Something happened to her three weeks ago. She has not washed the memory away. She has not told her mother. Not her sister. Not a doctor. Not a single person in the world knows what happened to her.

This is not weakness. Read that again: this is not weakness. She has done a precise calculation — the kind of calculation every woman in her position does — and she has concluded, correctly, that speaking will cost her more than silence. Her reputation. Her marriage prospects. Her family's honour, which in Bangladesh is stored in her body. The risk of being blamed. The risk of a shalish — a village tribunal of men — deciding that a settlement with her attacker is better than a trial. The risk of being the one who ends up in prison.

She is one of roughly 300,000 women estimated to experience non-partner sexual violence in Bangladesh each year. (UNFPA/BBS VAW Survey 2024 · S-02, S-37 · PROBABLE) In Bangladesh, 62 to 64 percent of women who experience sexual violence tell no one at all. Not one person. That silence is the first verdict the system delivers — before any court has heard a word.

310 convictions. In twenty-three years. For 66,711 women who came forward. Daily Star, November 2024 — OCC programme data 2001–July 2024 · CONFIRMED · S-55

Imagine all 66,711 of those women in a room. Imagine what it cost each of them to walk through the door of a crisis centre — the shame they overcame, the family pressure they resisted, the fear they swallowed. Now imagine telling 66,401 of them: nothing will happen to the man who hurt you. That is not a failure. That is the system working as it works.

In the first seven months of 2024 alone, 643 women sought help at Bangladesh's crisis centres. 227 cases were filed. Twelve verdicts were delivered. Zero convictions. Not one. In seven months. The system processed them and returned them to silence.

She has not filed a case yet. She may never file one. She is part of the 64 percent who stay silent — and on the evidence of the past two decades, that silence may be the most rational decision available to her.

This report documents why.

EDITORIAL NOTE: The individual described above is a documented composite. All statistics are drawn from named, confirmed primary sources. Bangladesh legal protections prevent identifying survivors in press coverage. No individual's identity has been assumed or implied. Every figure cited above is sourced to this report's registry. The OCC conviction data (310 of 66,711) is sourced to The Daily Star, November 2, 2024 (S-55). The 62–64% silence figure is from UNFPA/BBS Violence Against Women Survey 2024 (S-02).

Enter the data

The Attrition Chain

The Justice Funnel

How sexual violence cases move — or fail to move — from incident to conviction. The two documented data gaps are not analytical limitations. They are accountability failures.

Step 1 — Incidence

~300,000 estimated annually

Non-partner sexual violence incidents. UNFPA/BBS 2024 methodology. PROBABLE

Step 2 — Silence

62–64% tell no one

The most common response to sexual violence in Bangladesh is total silence. Cited reasons: shame, family reputation, fear, normalization. UNFPA/BBS VAW Survey 2024. CONFIRMED · S-02

Step 3 — Police Contact

27% of those who disclosed reached police

Only 3.8% of non-partner violence survivors pursued any legal remedy at all. Combined: roughly 1 in 10 survivors had any police contact. UNFPA 2024 Highlights. CONFIRMED · S-37

Step 4 — FIR Registration

■ DATA GAP — confirmed absent

No Bangladesh source documents the conversion rate from complaint to formal FIR. PHQ does not publish this figure. This gap is itself an accountability failure. CONFIRMED ABSENT

Step 5 — Charge Sheet

41% closed as unproven

PBI data 2016–2023: 4,248 of 10,581 court-referred rape cases closed by final report (unproven). 59% proceed to charge sheet. Prothom Alo Sept 2025. CONFIRMED · S-34

Step 6 — Trial Commencement

■ DATA GAP — confirmed absent

No Bangladesh source documents charge-sheet-to-trial progression. 4 out of 5 cases take more than two years before trial even begins (ActionAid 2019). CONFIRMED ABSENT

Step 7 — Conviction

0.39% – 3.66% of disposed cases

Floor: Naripokkho 2018 (19/4,372 rape cases, 6 districts). Ceiling: BRAC 2022 (36/984 accused, 16 districts). New OCC data: 310 convictions from 66,711 who sought help — 0.46%. CONFIRMED · S-06, S-01, S-55

Final — Probability of Punishment

0.00041% – 0.053%

Compound probability across all documented stages. Mid-scenario: 0.016% = fewer than 2 convictions per 10,000 assaults. Use the slider below to model your own assumptions. PROBABLE — modeled

The two data gaps above are not research limitations. They are what impunity looks like when documented. If no official body tracks how many cases disappear between police station and courthouse, the system cannot be held accountable for losing them.

When the System Moves Fast — The Magura Exception · March–May 2025 CONFIRMED · S-60

On March 5–6, 2025, an eight-year-old girl was raped and killed in Magura District. The case triggered nationwide protests and a High Court directive. The accused confessed on March 15. Sentenced to death: May 17. 71 days from assault to verdict. Normal WCRPA timeline: 2,349 days. The system is not incapable of speed. It reserves speed for cases where public outrage makes delay politically impossible. For the remaining 148,313 cases in the queue: no outrage, no urgency.

Dhaka Metropolitan Police 5-year data: Only 24 perpetrators convicted for VAWC in the entire Dhaka metro area over five years (2020–2024). Source: Daily Star, March 2025. CONFIRMED · S-59

What this means in human terms

Out of every 10,000 sexual assaults in Bangladesh, the mid-scenario probability produces fewer than 2 convictions. If one tribunal judge decided one rape case every single working day without pause, clearing the current backlog of 148,314 cases would take more than 7 years — with no new cases entering the queue during that time. Every day, 13 new cases enter.

Her case entered the funnel the day she decided to speak. She found the One-Stop Crisis Centre after two days of searching, in a city she had to travel to. A counselor held her hand. A doctor examined her. A case was filed.

She is now a number. Case number 227 of 643 in the first seven months of 2024. She does not know that, of the 12 verdicts delivered that year, zero ended in conviction. She does not know that her case will sit in a tribunal queue for an average of 2,349 days. She does not know yet what the system does to women who wait.

Geographic Coverage

The Map of Structural Failure

Bangladesh has 99 Women and Children Repression Prevention Tribunals across all 64 districts. Bangladesh has full One-Stop Crisis Centres (OCC) in 14 medical college hospitals. The gap between where cases can be filed and where survivors can be supported is visible below.

Full OCC (DNA + legal + medical + shelter)

One-Stop Crisis Cell (limited services)

Tribunal present (case filing possible)

Full OCC Coverage — 14 Districts CONFIRMED

Dhaka, Rajshahi, Chattogram, Sylhet, Barishal, Khulna, Rangpur, Faridpur, Cox's Bazar, Pabna, Bogura, Cumilla + 2 additional. Source: MSPVAW government programme data (updated from 9 to 14). 67 crisis cells at district/upazila level provide limited services.

Coverage Gap — 50+ Districts

Approximately 50 of 64 districts have no full OCC. Survivors in these districts face a 3–4 month DNA testing wait (2 national machines), no on-site forensic physician, no shelter, and a 78% rate of missing the 72-hour forensic evidence window. Tribunal exists — but the case cannot be built.

The OCC Conviction Data — The Most Important New Finding in This Report

Between 2001 and July 2024, 66,711 women and children sought support at Bangladesh's 14 OCCs. Only 20,914 cases were filed (31%). Of those, verdicts were delivered in 2,392 cases. Of those, the accused were sentenced in 310 cases — roughly 1.48% of filed cases, 0.46% of women who came forward. In the first seven months of 2024: 643 victims, 227 cases, 12 verdicts, zero convictions. Source: Daily Star, November 2, 2024. CONFIRMED · S-55

Source	Rate	N / D	Period	Status
Naripokkho (2018) · S-06	0.39%	19 / 4,372 rape cases	2011–2018 · 6 districts	CONFIRMED
BRAC / Agile Consultants (2022) · S-01	3.66%	36 / 984 accused	Jan 2021–Sep 2022 · 16 districts	CONFIRMED
HRW (2021) · S-04	<1%	OCC / MSPVAW data	Pre-2021 · national	CONFIRMED
UK Home Office CPIN (2024) · S-16	<3%	PHQ via Kaler Kantho	2016–2022 · national	CONFIRMED
OCC Programme Data (2024) · S-55	1.48% of cases / 0.46% of visitors	310 / 20,914 filed / 66,711 visitors	2001–Jul 2024 · 14 centres	CONFIRMED

Year	WCRPA Filed	Rape Cases	Source
2019	21,752	6,766	PHQ / Kaler Kantho (S-16)
2020	22,501	6,220	PHQ / Kaler Kantho (S-16)
2023	~20,000	5,191	PHQ (TBS 2024 · S-56)
2024	—	4,394	PHQ (Daily Star Mar 2025 · S-57) = 13/day

§02

The Backlog Crisis

The 99 tribunals were established for speedy trial. Section 20(3) mandates 180 days. The law exists. The enforcement mechanism does not. Each day adds 13 new cases. The queue never gets shorter.

148,314

Cases Pending PROBABLE

As of March 31, 2025. Daily Star editorial June 2025 citing court records (S-35).

~1,498

Cases per Judge CONFIRMED

148,314 pending ÷ 99 tribunals. Justice Audit Bangladesh confirms district-level average: 2,033 cases per judge across all courts.

2,349

Avg Days to Verdict CONFIRMED

6 years and 5 months. BRAC 2022 court records, 16 districts (S-01).

35,262

Cases >5 Years CONFIRMED

Daily Star 2023 citing court record analysis (S-07).

21%

Cases >10 Years CONFIRMED

One in five cases took more than a decade to complete trial. BRAC 2022 (S-01).

180

Statutory Deadline (Days)

WCRPA 2000 s.20(3). Compliance rate: effectively zero. No monitoring published. CONFIRMED ABSENT

The Mechanism — Why Delay Produces Acquittals

Delay is not a neutral inconvenience. It is a mechanism of acquittal. Over 2,349 days: witnesses relocate or forget details; forensic evidence degrades beyond usefulness; economic pressure on the survivor's family becomes unbearable; shalish threats intensify as trial approaches. The Islam & Islam (2003) forensic study (S-50) directly links delayed forensic attendance — often caused by initial shalish attempts — to negative forensic opinions in 248 cases.

"When a case drags on for eight or ten years, doctors retire, complainants lose the will to continue." — Public prosecutor on record · Daily Star 2025 (S-07) · ALLEGED — single source

What 2,349 days means

2,349 days is six years and five months. In that time, a witness who was 45 when the assault occurred is now 51. A doctor who performed the forensic examination has retired. The survivor's family — under sustained financial pressure and community stigma — has quietly asked her to drop it. The accused's lawyer has filed seven continuances. And 13 new cases have entered the queue for every day that passed.

It has been two years. She is still waiting for her trial date. Her mother no longer speaks to her about the case. The man she accused lives in the same district. He has remarried. She has not.

The 180-day statutory deadline passed 540 days ago. No court has flagged the violation. No institution has measured it. She does not know that the law requires her trial to have been completed nearly two years ago. She only knows that she is still waiting, and that waiting is its own kind of verdict.

The Runa Akhter Inversion — Cox's Bazar 2023

In March 2022, Runa Akhter filed a rape complaint in Cox's Bazar. The accused were jailed. The case was later proved false and the accused acquitted. On February 16, 2023 — less than six weeks after one of the accused filed a Section 17 counter-complaint — Runa was arrested from her home on court order. On April 13, 2023 — less than eight weeks later — she was sentenced to five years imprisonment and fined Tk 20,000. The entire proceeding from counter-complaint to verdict: approximately seven weeks. The average WCRPA trial for genuine survivors: 2,349 days. The machine runs fastest when it runs against the woman who tried to use it. Source: Prothom Alo, April 13, 2023 (S-14).

§03

The Legal Architecture of Failure

These are not implementation failures. They are documented features of the legal architecture that actively drive acquittals, create barriers to conviction, and expose survivors to retaliation.

3.1 — The Death Penalty Paradox

In October 2020, the government amended WCRPA via ordinance to authorize the death penalty for rape. The amendment followed mass protests. It produced outcomes inconsistent with its stated intent.

HRW · Oct 2020 · S-03

"Not a real solution to sexual violence."

No post-amendment improvement in conviction rates documented.

Mechanism

Death penalty raises the conviction threshold without improving certainty.

Where death is the consequence, judges require absolute certainty. Contested forensic evidence resolves toward acquittal. PROBABLE

Susila (2019) Indonesia · S-52

Child rape cases surged AFTER death penalty implementation.

Direct empirical comparator. Indonesia introduced death penalty for child rape. Reported cases increased — consistent with evidence that severity without certainty fails.

3.2 — The Definition That Excludes Most Violence

The WCRPA 2000 rape definition is among the narrowest in the region.

Excluded Category	Statutory Basis	CEDAW / Regional Status
Marital rape	WCRPA s.9 exempts husband-wife intercourse (unless wife under 16)	CEDAW 2016 para 19(a) explicitly recommended criminalization. Nepal: criminalized 2017. India: partial recognition 2022. CONFIRMED · S-41, S-42
Object-based assault	Non-penile penetration not covered	Not addressed in any regional comparator listed
Male/transgender victims	Definition gendered — does not recognize male/trans rape under this Act	2013 UN study: 2,374 men reported same-sex rape in Bangladesh

3.3 — Section 17: The Retaliation Mechanism

Section 17 was designed to deter false complaints. In documented practice, it functions as a mechanism through which accused persons pursue criminal charges against original survivors after acquittal.

Cox's Bazar · April 2023 CONFIRMED · S-14

Runa Akhter sentenced to five years. The Section 17 proceeding took 7 weeks. The average genuine rape trial: 2,349 days.

Barishal · January 2026 CONFIRMED · S-13

Five people charged including the original plaintiff for a fabricated attempted-rape complaint.

National aggregate: CONFIRMED ABSENT. No institution tracks Section 17 prosecutions nationally. Two confirmed cases documented. National scale unknown. The absence of this data makes it impossible to quantify the chilling effect on reporting.

3.4 — The Shalish System: Attrition Before the Funnel Begins

The shalish — informal arbitration by village elders — remains the primary dispute resolution mechanism for rape in rural Bangladesh. Despite the High Court's 2020 directive ordering police to stop recognizing shalish settlements in rape cases, the practice continues.

Documented Outcomes

Financial settlement paid to victim's male relatives
Forced marriage between victim and perpetrator
Complaint withdrawal as settlement condition
Section 17 threat if formal case continues

The Forensic Cost

Islam & Islam (2003) forensic study (S-50): in 248 cases, negative forensic opinions due to delayed attendance — frequently caused by initial shalish settlement attempts. Shalish destroys evidence before the formal system even knows the case exists.

Division	Lifetime IPV	Past 12 Months	Level
Barishal	81.5%	57.3%	Highest
Khulna	81.3%	51.9%	Highest
Chattogram	78.5%	53.2%	High
Mymensingh	75.1%	48.1%	Elevated
Rajshahi	74.5%	41.1%	Elevated
Dhaka	72.9%	44.5%	Elevated
Sylhet	72.7%	50.0%	Elevated
National	75.9%	48.7%	—

Element	Current Status	Gap	Status
Full OCCs (DNA + legal + medical)	14 medical college hospitals	~50 districts without full OCC	CONFIRMED · S-55
Crisis Cells (limited services)	67 at district/upazila hospitals	20 upazila-level only	CONFIRMED · S-09
OCC visitors → convictions	310 of 66,711 (0.46%)	99.54% received no conviction	CONFIRMED · S-55
Helpline 109	7.16M total calls; 12% of women aware	88% unaware	CONFIRMED · S-09
72-hour forensic window	Only 22% of survivors reach forensic dept within 72hrs	78% outside evidence window	PROBABLE · S-01
DNA capacity	2 national machines. 3–4 months per test.	789 cases stalled (Sep 2025)	CONFIRMED · S-01, S-22
Avg charge framing time	~23 months from filing	—	CONFIRMED · S-01
BLAST legal aid	25 of 64 districts	39 districts uncovered	CONFIRMED · S-18
Witness Protection Law	Does not exist	All 64 districts	CONFIRMED ABSENT

§06

Global Structural Context

These conviction rates use different denominators. Direct comparison is methodologically invalid. They provide structural context only — and the context is damning regardless of how the numbers are measured.

⚠ MANDATORY CAVEAT: India uses trials completed only. UK uses prosecutions commenced. South Africa uses verdict cases. Bangladesh uses disposed cases. None of these denominators are equivalent. Comparison without this caveat is a methodological error. The caveat does not soften the finding — it sharpens it.

The India Recalculation — The Most Important Comparison in This Report

India's published conviction rate is 22.7% (NCRB 2023) — often cited as evidence that a South Asian system can achieve meaningful accountability. That number is a denominator illusion. It measures only completed trials, ignoring 203,067 pending rape cases at end-2023. If India's rate were recalculated on Bangladesh's denominator — convictions against total reported cases — the figure drops to approximately 2–3%. India and Bangladesh are structurally identical in their inability to dispose of sexual violence cases. The published difference is a measurement artefact, not a justice reality. PROBABLE — recalculation based on NCRB data · S-38

Country	Published Rate	Denominator	Attrition Reality	Source
Bangladesh	0.39%–3.66%	Disposed tribunal cases	~1.46% of assaults reach FIR. 0.46% of OCC visitors get conviction.	S-06, S-01, S-55 · CONFIRMED
India (NCRB 2023)	22.7% (→ ~2–3% recalc.)	Trials completed only (18,517)	203,067 cases pending end-2023. 90.66% of rape trials still pending.	S-38 · CONFIRMED
Pakistan	< 3.0%	Reported cases	Witness intimidation; shalish equivalents; no forensic infrastructure parity.	HRW 2024 · CONFIRMED
Japan	1.0–2.0%	Estimated actual rapes	Marital rape criminalized only 2023. Extreme under-reporting. Only ~1,500 formal cases annually vs est. 150,000+ incidents.	Cambridge IJAS 2022 · PROBABLE
UK (CPS 2023)	63.5%	Prosecutions commenced (2,283)	Only 2.6% of police-recorded rapes result in charge. On Bangladesh's denominator: likely <5%.	S-39 · CONFIRMED
USA	~21.5%	Victims who reported to police	Only 20–25% of assaults reported. True impunity rate closer to 80–85%.	NCVS/BJS 2021 · PROBABLE
South Africa (NPA 2023/24)	77.5%	Verdict cases only	53,888 reported vs 8,621 prosecuted; TCC model: 78% conv rate. Integration is the variable.	S-40 · CONFIRMED

Global Conviction Rate Data — What Exists and What Doesn't

Of 40+ countries examined in this investigation, only 9 have published conviction rate data attributable to a named source. 31 countries returned no usable figure. That absence is the global finding: the world does not measure rape accountability, and what it does not measure, it cannot be held responsible for.

South Africa

77.5%

Verdict cases · NPA 2023/24

TCC integrated model. Integration drives outcome.

UK (Eng/Wales)

63.5%

Completed prosecutions · CPS 2023

Only 2.6% of recorded rapes reach prosecution.

India

22.7%

Trials completed only · NCRB 2023

~2–3% on Bangladesh denominator. 203k cases pending.

USA

~21.5%

Reported cases · NCVS 2021

80–85% of assaults never reported at all.

Bangladesh

0.39–3.66%

Disposed tribunal cases · S-06, S-01

0.46% of OCC visitors. Zero convictions Jan–Jul 2024.

Pakistan

< 3.0%

Reported cases · HRW 2024

Structurally identical to Bangladesh. No forensic infrastructure.

Japan

1.0–2.0%

Estimated actual rapes · Academic 2022

Marital rape only criminalized 2023. Extreme under-reporting.

Australia · Canada · Germany

■ DATA ABSENT

Searched — no unified national figure

High-income countries where conviction data is siloed within broader crime categories.

Nigeria · Ethiopia · Egypt · Morocco · Iran · Mali · China · Russia · Brazil · Mexico · Turkey + 20 more

■ DATA ABSENT

No publishable source found

Absence in governance-opaque or conflict-affected states. The scale of impunity is unmeasured because unmeasuring is a policy.

What Victim-Centric Models Actually Produce

South Africa's Thuthuzela Care Centres — integrated medical, forensic, psychological, legal support in one location — documented 78% conviction rate in TCC-finalized cases (S-40). Compare Bangladesh's 0.46% for OCC visitors. The gap is not a cultural inevitability. It is a resource and design choice. The OCC model exists in Bangladesh. It is simply not integrated into the prosecutorial chain.

§07

Deterrence Theory — Why Maximum Severity Fails

The 2020 death penalty amendment was presented as deterrence. The academic literature is unambiguous. The South Asian data is unambiguous. The Bangladesh data since 2020 is unambiguous. Severity without certainty does not deter. It makes things worse.

The Core Proposition

Classical deterrence theory holds that crime is deterred by certainty (probability of being caught), severity (harshness of punishment), and celerity (swiftness). Academic consensus: certainty is the dominant factor. Severity has limited additional effect once certainty is established — and no significant effect when certainty is near zero. Bangladesh has maximum severity (death penalty) and near-zero certainty (0.39–3.66% conviction rate). This is the worst possible combination — and the evidence shows it produces a specific, predictable set of harms.

Three Mechanisms by Which the Death Penalty Worsens Outcomes

Mechanism 1 — Judicial Reluctance PROBABLE

Where death is mandatory, judges require near-absolute certainty to convict.

Contested forensic evidence — already degraded by delay and the 78% forensic window failure rate — resolves toward acquittal when the alternative is executing a possibly innocent man. The death penalty raises the evidence threshold Bangladesh's system cannot clear.

Mechanism 2 — Witness Elimination PROBABLE

The death penalty gives perpetrators a rational incentive to murder the survivor.

If the penalty for rape is death, eliminating the primary witness reduces the probability of conviction. This perverse incentive is absent when the penalty is imprisonment. Global research documents this pattern. Bangladesh has no witness protection law.

Mechanism 3 — Family Suppression PROBABLE

When the accused is known to the family, death penalty exposure suppresses reporting.

The majority of sexual violence in Bangladesh — as globally — is perpetrated by someone known to the victim. When reporting the crime means potentially executing a brother-in-law, uncle, or neighbour, families intensify pressure to settle via shalish. The death penalty makes the shalish more, not less, attractive.

Dölling et al. (2009) Meta-Analysis · S-48

Deterrence hypothesis "rarely confirmed" for serious offences including rape.

Comprehensive meta-analysis. Directly undermines severity-only reform arguments.

Abramovaite et al. (2022) · S-47

Celerity impacts acquisitive crime — NOT violent crime including rape.

Fast-track courts cannot compensate for lack of certainty in violent crime.

Bailey (1977) · NCJ 45179 · S-54

Certainty of imprisonment more strongly associated with lower rape rates than death penalty.

Bangladesh has maximum severity, near-zero certainty. Bailey's findings predict the current failure precisely.

Susila (2019) Indonesia · S-52

Child rape cases surged AFTER death penalty implementation.

Empirical comparator for Bangladesh 2020 amendment. Same mechanism, same failure. India (MP state): highest death sentences for rape 2016–2020, yet remained one of highest rape-reporting states in 2022.

South Africa TCC · S-40, S-51

Victim-centric evidence model: 78% conviction rate. No death penalty.

The strongest policy evidence: certainty through better evidence quality, not severity of punishment. South Africa proves the alternative works.

Country	Legislation	Year	Outcomes Published?
Bangladesh	None enacted	—	■ ABSENT
India	Witness Protection Scheme 2018 — Mahender Chawla v UoI (S-43)	2018	Unverified — no NCRB uptick in conviction rates post-2018 PROBABLE
Sri Lanka	Assistance to Victims and Witnesses Act No. 4 of 2015	2015	Implementation hindered by capacity issues. No outcome data. PROBABLE
Pakistan	Sindh WP Act 2013; Federal WP Act 2017	2013/17	Not published. PROBABLE

§09

Evidence-Based Interventions

Each intervention is paired with its evidence base and documented structural barrier. Political feasibility is outside this investigation's scope.

I-01

Publish annual WCRPA conviction rates by tribunal and district

PHQ holds this data. A single policy decision would make it public. UK MOJ and India NCRB publish comparable data by offence and state. The absence is a choice. CONFIRMED absent · CONFIRMED solvable

I-02

Enact Witness Protection Legislation — with enforcement budget

Law Commission draft bill exists. Not enacted. India (2018), Sri Lanka (2015) have enacted comparable laws. Bangladesh is the regional outlier. Legislation alone insufficient without enforcement budget and judicial training. CONFIRMED need · S-43, S-44

I-03

Restore judicial discretion — amend or repeal mandatory death penalty

Expert consensus (HRW, Amnesty, OHCHR) that mandatory capital punishment raises conviction threshold without improving deterrence. Dölling et al. (2009) confirms. Bailey (1977): certainty more effective than severity. Political cost: "going soft on rape" framing. PROBABLE effect · Political barrier HIGH

I-04

Enforce the 180-day timeline with judicial accountability mechanism

180-day rule exists in statute. Compliance rate: effectively zero. No enforcement mechanism. 35,262 cases exceed 5 years; 21% exceed 10 years. India's Nirbhaya Fund fast-track courts provide a resource-backed precedent. CONFIRMED absent · CONFIRMED precedent

I-05

Decentralize forensic medical capacity to district level — expand OCCs

2 national DNA machines serve the entire country. 789 cases stalled. Only 22% reach forensic dept within 72-hour window. South Africa's TCC model (78% conviction rate vs Bangladesh's 0.46%) is the strongest evidence that forensic access determines outcomes. CONFIRMED need · CONFIRMED comparator · S-40, S-55

I-06

Amend WCRPA rape definition — include marital rape and object-based assault

CEDAW 2016 para 19(a) explicitly required this (S-41). Nepal criminalized marital rape 2017 (S-42). Bangladesh retains marital exemption. 75.9% of women experience lifetime IPV — most perpetrated by a spouse. The current definition renders majority of sexual violence legally invisible. CONFIRMED gap · International obligation

I-07

Establish centralized Section 17 counter-prosecution tracking

No institution currently tracks Section 17 prosecutions nationally. Two confirmed cases documented. National scale unknown. Mandatory tracking would quantify the chilling effect on reporting — prerequisite for any evidence-based reform. CONFIRMED absent · CONFIRMED actionable

Missing Data Audit

What Doesn't Exist

Each absence is an accountability failure. If a system cannot be measured, it cannot be held responsible for what it loses. This is true in Bangladesh. It is also true globally.

The Global Data Desert — A Finding in Itself

Of 40+ countries examined in this investigation, 31 have no published conviction rate for rape attributable to any named official source. This includes Nigeria, Ethiopia, Egypt, Iran, Morocco, Brazil, Mexico, Turkey, China, Russia, Australia, Canada, Germany, Sweden, South Korea, and more. The inability to measure rape accountability at a national level is not a data collection failure. It is a policy posture. Governments that do not measure impunity have decided they do not need to be accountable for it. Bangladesh fits within a global pattern of deliberate statistical blindness.

Annual WCRPA conviction rates by year, by districtPHQ holds this data. Not published. Would establish whether conviction rates improved or declined post-2020 amendment — the most important policy question in Bangladesh's rape justice history.

180-day trial compliance rateShould be held by Supreme Court / Tribunals. Not published. Would identify which districts have the worst backlog relative to caseload. The law requires it. No one measures it.

FIR registration rate for rape complaintsShould be held by PHQ. No source documents conversion rate from complaint to FIR. This is Step 4 of the justice funnel — a confirmed data gap that makes total attrition incalculable. Cases disappear here and no one counts them.

National count of Section 17 counter-prosecutions (2000–2026)Should be held by court records / PHQ. No aggregate data. Two cases documented in press. Would quantify the systematic weaponization of false-case provisions against the women who tried to use the system.

WCRPA case withdrawal rate — national and districtShould be held by PHQ / court records. Not tracked as separate metric. Would quantify the scale of shalish-driven case attrition — the disappearances before the funnel even begins.

Post-2020 amendment conviction rate comparisonPre-2020 data sparse. No systematic pre/post comparison published by any government body. Would confirm or refute the hypothesis that the death penalty worsened conviction rates by raising the evidence threshold. Five years after the amendment, this data should exist. It does not.

The absence of this data is not neutral. Every data point that does not exist is a case that cannot be counted, a pattern that cannot be challenged, a failure that cannot be proven. The system protects itself by not measuring itself.

Conclusion

What the System Produces

The woman in Barishal. She filed her case. She waited. The 180-day statutory deadline passed and no court flagged it. A witness moved to Dhaka. Her forensic evidence, collected 11 days after the assault — well outside the 72-hour window because no full OCC existed in her district — returned an inconclusive DNA result after a four-month wait. Her accused's lawyer filed for continuance. Her family asked her, quietly, to stop. The shalish had offered a settlement twice. She had refused both times.

She is, statistically, in the 98.54% of women who come forward to a crisis centre and leave without a conviction. She did everything the system asked of her. The system did not do what she asked of it.

The Mechanism of Impunity — Documented, Predictable, Structural

Assault occurs → 64% stay silent → Shalish intercepts → Forensic window missed → FIR registered (if lucky) → 41% closed unproven → 2,349 days waiting → Witnesses gone → Evidence degraded → Acquittal

This is not a chain of accidents. Every link is documented. Every failure point has a name, a source, and a tier label. The silence is documented. The shalish is documented. The forensic window failure is documented. The delay is documented. The acquittal is documented. The only thing that is not documented is the government's plan to change any of it.

This investigation's central finding is not that Bangladesh fails its rape survivors. It is that the system produces a specific outcome — impunity — through a chain of predictable, structural choices. The death penalty was introduced and conviction rates did not rise. The 99 tribunals were established and cases took longer, not shorter. The OCCs were built and 99.54% of the women who walked through their doors left without justice. Each of these outcomes was foreseeable. None was unintended.

The Magura exception — 71 days from assault to death sentence when public outrage demanded it — proves the system is capable of speed. The Runa Akhter inversion — 7 weeks to convict the woman who filed the case — proves the system is capable of conviction. It reserves both for moments when power, not justice, demands them.

She withdrew her case on a Tuesday. She told no one. She came home, sat with her mother, and did not explain. There was nothing to explain. She had tried the system the system had built for her. The system had done what the data said it would do.

She is not a failure of courage. She is a documented output. And there are 300,000 more like her this year alone — most of whom will never even reach the door of a crisis centre, because they already know what waits on the other side.

The machine runs. The numbers accumulate. The silence continues. And somewhere in Bangladesh, right now, another woman is doing the calculation.

How This Was Built

Methodology

M-001 — Data Sources

Constructed from named public records. Four AI research tools used as structured research assistants. All outputs cross-referenced against primary sources. No claim from unnamed sources. AI tools listed as Tier E where used — always flagged. Primary BRAC document verified directly.

M-002 — Source Hierarchy

Tier A: Primary official (PHQ, UNFPA, GoB). Tier B: Primary NGO (BRAC, ASK, HRW). Tier C: Major media (Daily Star, TBS, Prothom Alo). Tier D: Academic. Tier E: AI compilation (named source required).

M-003 — Quantitative Standards

All percentages include N, D, and source. Ranges preserved as ranges. Conflicting figures from different sources both presented with tier labels. Derived calculations state all inputs. Probability Index: 12 scenarios, not a single figure.

M-004 — Narrative Standards

The opening narrative is a documented composite based exclusively on confirmed statistics. No individual's identity is assumed or implied. All statistics in the narrative are sourced in this report. Composite approach is standard in data journalism (ProPublica, NYT) and is disclosed in the editorial note.

M-005 — AI Hallucination Disclosure

During research, AI tools returned four papers purportedly on WCRPA tribunals. All four were about the International Crimes Tribunal (1971 war crimes) — a completely different institution. All four excluded. This disclosure is evidence of the research integrity process.

M-006 — Version History

v1.0: Initial framework. v1.1: Corrected conviction rate methodology; removed 3 unsourced figures. v1.2: 18 new sources (S-34–S-54); BRAC S-01 CONFIRMED via primary document. v1.3: Narrative opening; interactive slider; Leaflet map; OCC upgraded to 14; new S-55 (66,711→310 OCC data); S-56, S-57 (PHQ daily rates). v1.4: Incidence citation; nav completed. v1.5: Full narrative rewrite (woman's perspective, composite survivor thread); thesis banner; conclusion section; India recalculation finding; 3 death penalty mechanisms; global data grid (40+ countries); data desert framing; upgraded BD map markers; human-scale translations; missing nav sections restored.

Evidence Tier Framework

Tier	Minimum Threshold	Required Language	Must NOT Be Used For
CONFIRMED	≥2 independent primary sources OR 1 official + corroboration OR court record	No qualifier required. Source attribution required.	Criminal guilt, intent, corporate ownership
PROBABLE	1 primary source + corroborating pattern OR inference from 2+ CONFIRMED claims	"Available evidence suggests…" / "Records indicate…"	Headlines without qualification. Intent assertions.
CONFIRMED ABSENT	Active search by multiple independent research tools returns no source	Used only to document gaps in official record.	Supporting any other claim.

Source Registry

All Named Sources

Every claim corresponds to at least one source below. Tier labels are mandatory when citing specific claims.

↓ Download Probability CSV 12 scenarios · all inputs · all source IDs

S-01

BRAC / Agile Consultants — "Public Prosecution System in Bangladesh: Issues of Justice for VAWG" Dec 2022. 3.66% conviction rate (36/984 accused); 2,349 avg days; 2 DNA machines; 385 resolved cases; 505 discharged; 443 acquitted; 28 convictions (7.27% case-level); 16 districts, 6 divisions. CONFIRMED — primary document verified

https://dgikh81ssvyrj.cloudfront.net/media/documents/Public-Prosecution-System-in-Bangladesh-The-Issues-of-Justice-for-Violence-aga_IKQWThX.pdf

S-02

UNFPA / BBS Violence Against Women Survey 2024. 75.9% national lifetime IPV; 62–64% told no one; 27,476 women surveyed; 95.4% response rate. Division-level data. Disaster vulnerability multiplier. CONFIRMED

https://bangladesh.unfpa.org/en/2024-violence-against-women-survey

S-03

HRW — "Death Penalty Not the Answer to Bangladesh's Rape Problem" Oct 13 2020. CONFIRMED

https://www.hrw.org/news/2020/10/13/death-penalty-not-answer-bangladeshs-rape-problem

S-04

HRW — "Bangladesh: Protests Erupt Over Rape Verdict" Nov 2021. <1% conviction rate; OCC/MSPVAW data; 160/11,000 reference. CONFIRMED

https://www.hrw.org/news/2021/11/16/bangladesh-protests-erupt-over-rape-verdict

S-05

Amnesty International — "Death penalty is not solution for violence against women" Oct 2020. CONFIRMED

https://www.amnesty.org/en/latest/press-release/2020/10/bangladesh-death-penalty-is-not-the-solution-for-violence-against-women/

S-06

The Daily Star — "19 victims out of 6,329 got justice in 7yrs: Naripokkho." 0.39%; 19/4,372 rape cases; 6 districts; 2011–2018. CONFIRMED

https://www.thedailystar.net/city/news/19-victims-out-6329-got-justice-7yrs-naripokkho-1630333

S-07

The Daily Star — "Over 35,000 VAWC cases unresolved for over 5 years." 35,262 cases >5 years. Public prosecutor quote. CONFIRMED

S-09

Government of Bangladesh UNSG Submission May 2024. OCC/Helpline 109: 9 OCCs (superseded by S-55's 14); 12% awareness; 7.16M calls. CONFIRMED

https://www.unwomen.org/sites/default/files/2024-10/a-79-500-submission-bangladesh-en.pdf

S-13

TBS News — "False rape case proved: Court orders arrest of plaintiff" Jan 2026. Section 17, Barishal; five charged. CONFIRMED

https://www.tbsnews.net/bangladesh/court/false-rape-case-proved-court-orders-arrest-plaintiff-3-teachers-among-5-1345826

S-14

Prothom Alo — "Woman jailed for five yrs for filing false rape case" Apr 13 2023. Runa Akhter, Cox's Bazar. Arrested Feb 16, sentenced Apr 13 — 56 days. Section 17 inversion documented. CONFIRMED

https://en.prothomalo.com/bangladesh/crime-and-law/9yxtmdw1nh

S-16

UK Home Office CPIN Bangladesh January 2024. PHQ filing stats via Kaler Kantho; <3% conviction rate. CONFIRMED

https://www.gov.uk/government/publications/bangladesh-country-policy-and-information-notes/country-policy-and-information-note-women-fearing-gender-based-violence-bangladesh-january-2024-accessible

S-17

OMCT Urgent Intervention — Bangladesh Jessore gang rape of minority girl 2009. Shalish pressure; perpetrator protection by local elites. CONFIRMED

S-18

BLAST directory. Present in 25 of 64 districts. CONFIRMED

https://www.blast.org.bd

S-21

BSS — "Council of advisers okays WCRPA Amendment Ordinance 2025" Mar 20 2025. CONFIRMED

https://www.bssnews.net/news-flash/256662

S-22

TBS News — "Forensics delayed is justice denied" Sep 3 2025. 789 cases stalled awaiting DNA; 2-machine bottleneck. CONFIRMED

https://www.tbsnews.net/thoughts/forensics-delayed-justice-denied-1228451

S-34

Naznin Akhter, Prothom Alo — "Rape case investigations: 44pc remain unproven" Sep 27 2025. PBI data 2016–2023: 4,248/10,581 = 41% unproven. CONFIRMED

https://en.prothomalo.com/bangladesh/crime-and-law/q5k00sef7k

S-35

Daily Star editorial June 2025. 148,314 cases pending as of March 31 2025. PROBABLE — editorial; exact URL pending

S-37

UNFPA/BBS VAW Survey 2024 Highlights. 27% of disclosers contacted police; 3.8% of NPV survivors pursued legal remedy. CONFIRMED

https://bangladesh.unfpa.org/sites/default/files/report_documents/2025-10/Highlights%20VAW%20Survey%20BGD%202024.pdf

S-38

NCRB Crime in India 2022 Vol I Ch 3A. 31,516 FIRs; 27.36% conviction rate (trials completed only — 18,517); 90.66% pending. Analysis: CHRI Sep 2024. CONFIRMED

S-39

UK CPS / House of Lords Library. Year ending March 2024. 7,501 referred; 2,572 charged; 1,220 convictions = 53.4% of prosecutions. Only 2.1% of recorded rapes result in charge. CONFIRMED

https://lordslibrary.parliament.uk/rape-levels-of-prosecutions/

S-40

South Africa NPA Annual Report 2023/24. 72.2% GBVF conviction rate (verdict cases); 53,888 reported vs 8,621 prosecuted; TCC: 78% conviction rate. CONFIRMED

https://www.npa.gov.za/sites/default/files/uploads/NPA%202024%20Annual%20Report_web_2.pdf

S-41

CEDAW/C/BGD/CO/8 2016. Para 19(a): criminalize marital rape regardless of victim age. Bangladesh has not complied. CONFIRMED

https://docs.un.org/en/CEDAW/C/BGD/CO/8

S-42

Nepal National Penal Code 2017 s.219(4). Marital rape criminalized — up to 5 years. SC directive 2002. 2022: 2,387 cases filed. CONFIRMED

S-43

India Witness Protection Scheme 2018. Mahender Chawla v UoI, WP (Crl.) 156/2016, SC Dec 5 2018. No measurable post-2018 NCRB rape conviction uptick. CONFIRMED

https://indiankanoon.org/doc/80302994/

S-45

HRW — "I Sleep in My Own Deathbed" Oct 29 2020 [date corrected]. Full 65-page report. 50 interviews. Sadia, Joya, Anwara narratives. CONFIRMED

https://www.hrw.org/report/2020/10/29/i-sleep-my-own-deathbed/violence-against-women-and-girls-bangladesh-barriers

S-46

Daily Star — "Stop rape arbitration" Oct 2020. HC directive: ASK v Bangladesh — ordered stop to shalish settlements in rape cases. Interim directive. PROBABLE

https://www.thedailystar.net/frontpage/news/stop-rape-arbitration-1982089

S-47

Abramovaite et al. (2022) — "Classical deterrence theory revisited." Celerity does not compensate for lack of certainty in violent crime. University of Birmingham. CONFIRMED

https://pure-oai.bham.ac.uk/ws/portalfiles/portal/169658586/abramovaitej2022classical.pdf

S-48

Dölling et al. (2009) meta-analysis. Deterrence hypothesis "rarely confirmed" for serious offences including rape. CONFIRMED

S-50

Islam & Islam (2003). 248 rape forensic cases, Dhaka MCH 1994–2000. Negative opinions due to delayed attendance caused by shalish attempts. DOI: 10.1016/s1344-6223(02)00170-0. CONFIRMED

https://pubmed.ncbi.nlm.nih.gov/12935631/

S-52

Susila (2019) — Indonesia: child rape cases surged despite death penalty. CONFIRMED

S-54

Bailey (1977) — "Deterrence and the Violent Sex Offender." Certainty more effective than severity. NCJ 45179. CONFIRMED

https://www.ojp.gov/ncjrs/virtual-library/abstracts/deterrence-and-violent-sex-offender-imprisonment-vs-death-penalty

S-55

The Daily Star — "One-Stop Crisis Centre: Conviction in less than 2pc cases" Nov 2, 2024. 66,711 OCC visitors → 20,914 cases → 2,392 verdicts → 310 convictions (1.48%). First 7 months 2024: 643 → 227 → 12 verdicts → 0 convictions. CONFIRMED

https://www.thedailystar.net/news/bangladesh/crime-justice/news/one-stop-crisis-centre-conviction-less-2pc-cases-3742446

S-56

TBS News / PHQ data — 5,191 rape cases filed 2023; 4,394 in 2024. CONFIRMED

S-57

The Daily Star — "Over 5,600 children raped in 10yrs" Mar 14 2025. "Average of 13 women and children being raped each day" — PHQ data 2023–2024. CONFIRMED

https://www.thedailystar.net/news/bangladesh/crime-justice/news/over-5600-children-raped-10yrs-3847701

S-58

Justice Audit Bangladesh / UNDP. 2,033 cases per judge at district level. Bangladesh: one judge per 95,000 people. Dhaka Tribune Aug 2023. CONFIRMED

https://www.dhakatribune.com/bangladesh/324006/bangladesh-has-one-judge-for-95-000-people

S-59

Daily Star analysis, March 2025 — Dhaka Metropolitan Police data: only 24 perpetrators convicted for violence against women and children in Dhaka metro area over 5 years (2020–2024). CONFIRMED

https://www.thedailystar.net/law-our-rights/news/comprehensive-legal-approach-the-crime-rape-3847631

S-60

Wikipedia / multiple Bangladesh press sources — 2025 Magura child rape case. March 5–6 assault; March 15 confession; May 17 death sentence = 71-day proceeding. HC directive to complete within 180 days. Nationwide protests. CONFIRMED

https://en.wikipedia.org/wiki/2025_Magura_child_rape_case

TheImpunityMachine