BLOOD ROUTES · BD-INV-001
v1.0
PUBLIC RELEASE v1.0
Published:
Last Updated: February 2026
DOI: PENDING
Corrections: 0
01 Data Gap 02 Operators 03 Map 04 Phoenix 05 Political Economy 06 Global Context 07 Solutions 08 Data Gaps

BLOOD ROUTES DHAKA

An open-source investigative dataset on bus-related fatalities in the Dhaka metropolitan area, the structural incentives that produce them, and the accountability failures that perpetuate them.

Bangladesh Road Safety Data Journalism Operator Accountability Citable Primary Record
REPORT METADATA
Report ID BD-INV-001
Version 1.0 — Public Release
Research Period 2018–2026
Sources 155+
Corrections 0
WHO Est. Deaths ~31,578
Evidence Tier Confirmed Probable

THE REAL DEATH TOLL

⚠ Headline Claim — Evidence Tier: CONFIRMED
Bangladesh road fatalities are systematically undercounted. The WHO estimates 31,578 deaths annually; official government figures report 5,480. The differential — a factor of approximately 5.8× — is documented by the WHO Global Status Report (2023), DGHS-CIPRB National Injury Survey (2023), and supported by hospital-level studies at BUET's Accident Research Institute. Sources: WHO Global Status Report on Road Safety 2023 · DGHS-CIPRB National Injury Survey 2023 · World Bank "Delivering Road Safety in Bangladesh" 2022 · BRTA Annual Statistics 2023
~31,578
WHO Estimated Annual Deaths (2021)
Most recent WHO estimate. Represents approximately 1.0% of all deaths in Bangladesh. WHO uses modelled estimates incorporating hospital surveys, police data, and adjustment factors for underreporting.
SOURCE: WHO Global Status Report on Road Safety 2023
Confirmed
8,543
NGO-Reported Deaths (2024)
Bangladesh Jatri Kalyan Samity (BJKS) methodology: media monitoring across 37 daily newspapers. Represents a 27.43% increase over 2023 (7,902). Not directly comparable to WHO estimates due to different methodology.
SOURCE: BJKS Annual Road Accident Report 2024
Confirmed
5,480
Official Government Figures (2023)
Based on police FIR records compiled by BRTA. Acknowledged undercount: many crashes settled out-of-court before police involvement, hospital deaths not linked to crash records.
SOURCE: BRTA Annual Report 2023
Confirmed
5.8×
WHO-to-Official Ratio
Calculated: WHO 31,578 ÷ BRTA 5,480. Note: ratios should not be presented as a precise figure — WHO models have confidence intervals not stated here. The underreporting direction is unambiguous; the magnitude is estimated.
Derived calculation — see Methodology §M-004
Probable — model-dependent
Bangladesh Road Deaths: NGO-Tracked Annual Trend (2018–2024)
Source: BJKS Annual Reports. Methodology: media scanning of 37 national daily newspapers. Limitations: media scanning undercounts rural incidents, night incidents, and cases resolved without press coverage. 2020–21 decline likely reflects COVID-19 mobility restrictions, not genuine safety improvement — no infrastructure change occurred.
SOURCES: BJKS Annual Road Accident Reports 2018–2024 · Dhaka Tribune (2024 report corroboration) · The Daily Star archive
CONFIDENCE: Confirmed — multiple years verified by cross-referencing BJKS press releases with major newspaper coverage.
REPORTING LIMITATION — Must Disclose
The 409 Dhaka-specific accidents / 219 deaths figure (RSF 2025) covers Dhaka city only and uses the same media-scanning methodology as BJKS. It is not a primary police or hospital record. The national 5.8× underreporting ratio should not be uncritically applied to derive a Dhaka-specific "true" figure — Dhaka has different reporting density (more press coverage) than rural Bangladesh, potentially meaning Dhaka-specific underreporting is lower. Inference limitation — see Methodology §M-009 (Known Blind Spots)
Who Dies? Victim Demographics — Dhaka City 2025
RSF 2025: 219 deaths in 409 accidents. Note: "children" defined as under 18 by RSF methodology — not verified against independent record.
SOURCE: Road Safety Foundation (RSF) 2025 Annual Report for Dhaka City · Confirmed (single primary source — no independent verification available)
Vehicle-Type Share of Fatal Accidents — Dhaka City
Buses represent 24.87% of fatal accidents despite being a minority of registered vehicles. Comparison to vehicle census not yet conducted in this dataset — noted as data gap.
SOURCE: RSF 2025 · ARI-BUET supporting studies · Confirmed

OPERATOR DATASET

# Operator Status Documented Deaths Key Incidents BRTA Action Evidence Tier Sources
OP-001
Suprobhat Paribahan
সুপ্রভাত পরিবহন
Route: Sadarghat–Gazipur
BANNED
March 2019
3+
Single incident documented
19 March 2019: Bus Reg. Dhaka Metro-Ba-11-4135 killed BUP student Abrar Ahmed Choudhury at Jamuna Future Park zebra crossing. Same bus had struck student Cynthia Islam Mukta at Shahzadpur minutes earlier. Driver fled. Owner (named in charges) + 5 others charged under Road Transport Act.
Student protests triggered route permit cancellation
Route permit cancelled (March 2019). Owner charged. Confirmed Bangla Tribune (2019)
The Daily Star (2019)
Dhaka Tribune (2019)
Court charge records cited in press
OP-002
Victor Classic
ভিক্টর ক্লাসিক
Evidence suggests: successor to Suprobhat (same route)
ACTIVE / Disputed
5+
2019–2025
2019: Music director Parvez Rob killed in Uttara — investigating reports established the driver was an unlicensed helper (no licence on record).
2024–25: Multiple incidents on Progati Sarani corridor reported. Buses torched by crowds on at least 2 occasions per media reports.
On BRTA watchlist. No confirmed cancellation as of v1.0. Confirmed (incidents)
Probable (Suprobhat link)
Daily Sun (rebrand report 2019)
The Daily Star tags
Dhaka Tribune (2019)
OP-003
Akash Paribahan
আকাশ পরিবহন
Evidence suggests: successor to Victor Classic
ACTIVE
2+
Documented Oct 2024
October 2024 (exact date to be confirmed): Two Akash Paribahan buses racing on Progati Sarani, Madhya Badda. Tasnim Jahan Airin killed at the scene. Sister Nusrat Jahan Jerin critically injured. Eyewitness testimony (TBS News, Dhaka Tribune) confirms racing. Victim family explicitly stated belief in Suprobhat/Victor Classic ownership continuity — this is eyewitness testimony, not independently verified corporate record. Drivers arrested. Owner identity not yet confirmed in public record as of v1.0. Confirmed (incident)
Alleged (ownership link)
TBS News (Oct 2024)
Dhaka Tribune (Oct 2024)
Prothom Alo (Oct 2024)
OP-004
Jabal-e-Noor Paribahan
জাবালে নূর পরিবহন
PARTIALLY BANNED
4+
2018 key incident
29 July 2018: Two Jabal-e-Noor buses racing near Kurmitola General Hospital, Airport Road. Shaheed Ramiz Uddin Cantonment College students Diya Khanam Mim and Abdul Karim Rajib killed. This incident directly triggered the 2018 Safe Road Movement. 4 drivers and helpers given 7-day remand. Two bus registrations cancelled by BRTA (confirmed by court records). 2 bus registrations cancelled (confirmed). 3 staff convicted (life sentences commuted — see Financial Express 2020). Confirmed Dhaka Tribune (2018)
The Daily Star
BRTA registration records cited in press
Financial Express (2020 — sentencing)
OP-005
Imad Paribahan
ইমাদ পরিবহন
Long-haul operator (Expressway)
PERMIT REVOKED
19
Single incident · confirmed
Bus plunged off Padma Bridge Expressway, Shimana, Kutubpur. ARI-BUET formal investigation concluded the bus's fitness certificate AND route permit were both suspended at time of operation. Operator permitted to run on revoked credentials — a systemic failure of BRTA enforcement documented in the official investigation report. Permit revoked post-incident. ARI-BUET investigation published. Confirmed ARI-BUET Investigation Report (official, published)
Financial Express
The Daily Star
OP-006
Raida Paribahan
রাইদা পরিবহন
ACTIVE
3+
Pattern-documented
Multiple reports of speeding on Kuril Flyover. Pedestrian fatalities in Progati Sarani corridor attributed in media reports. Named in BRTA 2021 blacklist of 25 companies. Specific incident dates and victim names not independently confirmed in this dataset — flagged as data gap. BRTA blacklisted Dec 2021. No confirmed permit cancellation as of v1.0. Probable — pattern only TBS News (BRTA list 2021)
bdnews24 (BRTA list 2021)
DATASET CONSTRUCTION NOTE
No central "operator death registry" exists in Bangladesh. This dataset is constructed from: newspaper archives, court documents cited in media, BRTA blacklist (2021), and police FIR information cited in press. It does not represent a complete census of operator-linked fatalities. Operator rows marked Probable or Alleged must not be cited as factual without independent verification. See: Data Tab for full dataset schema and download.

ACCIDENT HOTSPOT MAP

MAP METHODOLOGY NOTICE — v1.0 Limitations
This map is a schematic visualization, not a GIS-verified spatial dataset. Hotspot locations are derived from media reports and ARI-BUET published research. They represent qualitative high-risk zones, not statistically derived accident density calculations. Coordinates are approximate. This map does not yet incorporate BRTA accident registry data (not publicly available — see Data Gaps section). Inferences about specific routes or operators at specific locations are marked with evidence tier. Treating this map as a primary forensic instrument requires the data upgrades specified in the Map Specification Document (Data tab). See: Data Tab → Map Specification v1.0
Risk Level
Operators
Key Incident
Evidence
BURIGANGA AIRPORT ROAD MIRPUR ROAD PROGATI SARANI KURIL FLYOVER KURIL BADDA KURMITOLA ABRAR / JFP MIRPUR 10 JATRABARI GABTOLI DHAKA METRO · ACCIDENT HOTSPOT MAP · v1.0 SCHEMATIC — NOT GIS-VERIFIED · See Methodology
Fatality Risk
Extreme
High
Medium
HOVER DOTS FOR DETAILS
+ EVIDENCE TIERS
⚠ MAP LIMITATION: Dot placement is approximate. This map requires upgrade to a GIS-verified dataset before use as primary forensic evidence. See Data Tab → Map Specification v1.0 for required datasets and upgrade path.

THE PHOENIX PHENOMENON

EVIDENTIARY WARNING — High Impact, High Risk Claim
The claim that Suprobhat, Victor Classic, and Akash Paribahan represent the same beneficial ownership is alleged but not fully confirmed by corporate registry. The incident connections are confirmed. The ownership continuity is probable (route continuity confirmed; vehicle continuity probable; beneficial ownership alleged via eyewitness testimony and press investigation). This section must be read with the Phoenix Evidence Rulebook (Phoenix Tab). Do not cite the ownership chain as "confirmed" without noting the evidence tier.
~2010s – March 2019
SUPROBHAT
সুপ্রভাত পরিবহন
BANNED · March 2019
Abrar Ahmed Choudhury killed 19 March 2019. Owner charged. Route permit cancelled post-student protests.
Rebrand
~Weeks
2019–2023
VICTOR CLASSIC
ভিক্টর ক্লাসিক
Probable Successor
Parvez Rob killed 2019. Racing incidents 2024–25. Buses torched by crowds. Same Sadarghat–Gazipur corridor.
Rebrand
~2023
2023–Present
AKASH
আকাশ পরিবহন
Alleged Successor
Badda sisters killed Oct 2024. Victim family alleges same ownership. Corporate record not independently verified.
Next?
Unknown
Future
???
Unknown entity
Predicted pattern
Without chassis-linked digital registry, rebrand will be undetectable until next fatality.
WHAT CAN BE CLAIMED vs WHAT CANNOT — Phoenix Chain
✓ CAN CLAIM:
— Suprobhat Paribahan's route permit was cancelled in March 2019 (confirmed).
— Victor Classic subsequently operated the same route (confirmed — media and BRTA records).
— Akash Paribahan buses were involved in the October 2024 Badda fatality (confirmed — police + media).
— Victim family and eyewitnesses stated they believed ownership was continuous (confirmed — as testimony, not fact).
— Daily Sun reported "Banned Suprobhat Paribahan tricks to operate by disguising its buses" in 2019 (confirmed — with attribution to that specific report).

✗ CANNOT YET CLAIM:
— That the same individual legally owns all three companies (not confirmed by corporate registry).
— That specific vehicles moved from one company to another (chassis records not verified).
— That the rebranding was conducted with intent to evade prosecution (intent not established).

STRUCTURAL INCENTIVES

✓ Core Structural Finding — CONFIRMED
The waybill/joma system creates a documented economic incentive for bus racing. The TIB (2024) report on the Private Bus Transport Sector — a primary source — documents the mechanics of this system, the Tk 1,059 crore annual bribery figure, and the 92% political connectivity figure. These findings are attributed to TIB; they have not been independently verified by this investigation beyond confirming the TIB report exists and was published. Primary Source: Transparency International Bangladesh (TIB) — "Integrity in Private Bus Transport Business" — Executive Summary, 2024. Available: ti-bangladesh.org
🏢
Bus Owner
Tk 4,000–6,000/day
Fixed daily "joma" (জমা) charged to driver before any earnings. Owner's income is guaranteed; all operational risk transfers to the driver-helper team.
📋
Waybill Checker
Checkpoint system
Owner-employed "checkers" at route intervals count passengers and record waybills. Creates race between co-route buses: the first bus at each stop takes all passengers.
🚌
Driver / Helper
Net ~Tk 300–500/day
Driver pays joma + fuel first. Only surplus = wage. Racing and overloading are rational survival strategies under this structure — not recklessness but rational response to incentives.
💀
Racing
Pedestrian Risk
Two same-route buses racing = mortal hazard at bus stops and pedestrian crossings. The Badda October 2024 fatality is a documented output of this incentive structure.
🏛️
Political Shield
92% politically linked*
*TIB figure — attributed, not independently verified. Enforcement failure creates impunity loop: driver flees, owner settles, company rebrands.
Tk 1,059cr
Annual Bribes — Bus Sector (TIB 2024)
Transparency International Bangladesh (TIB) primary research. Methodology: structured interviews, field surveys. Attribution required when citing.
TIB "Integrity in Private Bus Transport Business" 2024
Confirmed (as TIB finding — not independently replicated)
92%
Operators with Political Connections (TIB)
TIB figure — requires attribution. Independent verification of this percentage not conducted in this dataset. Should be cited as "TIB estimates 92%" not stated as fact.
TIB 2024 · Bangladesh Centre for Development Research
Probable — single primary source
24%
Buses Without Fitness Certificate (TIB)
TIB survey finding. Dhaka-specific. Methodological note: sample size and survey geography from TIB report — not reproduced in this dataset.
TIB 2024 study on Private Bus Transport
Confirmed (as TIB finding)
5% GDP
Economic Cost of Road Crashes
Range cited: 2.5–5.1% of GDP. World Bank and ATO Bangladesh 2025 profile both provide estimates. Upper figure from ATO; lower from World Bank. Use range, not point estimate.
World Bank 2022 · Asian Transport Observatory 2025
Confirmed — range, not single figure
Annual Bribery Flows — TIB 2024 (Tk Crore)
Figures from TIB primary research. Cited as TIB estimates — not independently verified by this investigation.
SOURCE: TIB 2024 · Confirmed (as attributed TIB figure)
Fatal Crash Time Distribution — Dhaka City 2025
41.56% of fatal crashes occur at night. RSF 2025 — single primary source. Limitations: time-of-day definitions should be verified against RSF methodology.
SOURCE: RSF 2025 · Confirmed (RSF single source)

INTERNATIONAL BENCHMARK

✓ Headline Finding — CONFIRMED with Caveat
Bangladesh's WHO-estimated road fatality rate of approximately 19.3 per 100,000 population places it among the worst-performing nations globally and in the worst tier in Asia-Pacific. The figure 102 deaths per 10,000 registered vehicles (South Asia comparison) is from World Bank / ATO sources and should be cited as approximately that figure, noting that vehicle registration data has its own reliability issues in Bangladesh. WHO Global Status Report 2023 · Asian Transport Observatory Bangladesh Road Safety Profile 2025 · World Bank 2022
Road Fatality Rate Comparison — Estimated Deaths per 100,000 Population (WHO 2023)
All figures from WHO Global Status Report 2023 or ATO 2025 (Bangladesh). Nigeria figure is national rate, not city-specific. Comparisons across countries with different data quality should be read with caution. The direction of Bangladesh's relative position is robust; specific rankings may shift as better data emerges.
SOURCES: WHO Global Status Report on Road Safety 2023 · ATO Bangladesh Road Safety Profile 2025 · Confirmed for all listed figures
City/Country Bus Fleet Model Driver Pay Model Public Crash Dashboard Rate /100k Key Structural Difference Evidence
Dhaka, Bangladesh 5,000+ individual owners under 200+ competing "banners." No franchise. Trip-based joma. Driver pays owner first. Zero salary. None ~19.3 Joma incentive creates racing. No chassis registry. 92% political connectivity (TIB). Confirmed
Delhi, India DTC (public) + franchised private operators Salary-based for DTC. Some contractor variation. Partial ~16.0 Franchise model reduces racing incentive on regulated routes. Confirmed
Jakarta, Indonesia TransJakarta BRT (public) + reformed private sector since 2014 Salary-based post-2014 TransJakarta reform Yes — published ~12.2 BRT reform eliminated racing incentive on key corridors. Template applicable to Dhaka. Confirmed
Singapore SBS Transit + SMRT. Fully franchised, state-licensed. Full salary + performance bonus. Zero revenue-based target. Full real-time ~2.8 Zero racing incentive. Complete operator accountability. Route-level performance data published. Confirmed

WHAT WORKS — DOCUMENTED INTERVENTIONS

SCOPE OF THIS SECTION
This section documents interventions with evidence of impact in comparable contexts. It does not predict outcomes for Bangladesh. Each recommendation is paired with its evidence base and a note on transferability limitations. Political feasibility assessment is outside this investigation's scope.
# Intervention Evidence Base Comparator Local Barrier Priority
S-01 Abolish Joma/Waybill System
Convert city bus drivers to monthly salary contracts. Eliminate trip-based revenue target that directly incentivizes racing.
TransJakarta reform (2014) demonstrated racing reduction post-salary conversion. Delhi DTC model. Dhaka Nagar Paribahan pilot (partial). Jakarta, Delhi Requires buy-in from owner class (politically connected per TIB). Franchise model needs enabling legislation. P0 — IMMEDIATE
S-02 Chassis-Linked Digital Vehicle Registry
Permanent public record linking chassis/VIN to owner history, accident history, fitness certificates, and route permits. Makes rebrand-and-continue impossible.
Singapore LTA model. TransJakarta fleet management system. UK DVLA accident linkage system. Singapore, UK BRTA already holds registration data — public access is a policy decision, not a technical barrier. Requires political will. P0 — STRUCTURAL
S-03 Owner Criminal Liability
When an unlicensed driver hired by an owner causes death, the owner faces manslaughter charge. Breaks impunity where all risk falls on fleeing driver.
UK corporate manslaughter legislation. Australia's heavy vehicle chain of responsibility laws. Road Transport Act 2018 (Bangladesh) — provision exists, enforcement absent. UK, Australia Legal framework partially exists. Prosecution requires political independence of judiciary and enforcement. P1 — ENFORCEMENT
S-04 Independent Accident Investigation Body
Separate from BRTA and police. Authority to compel evidence, access CCTV, publish findings. Modelled on aviation/rail accident investigation boards.
UK RAIB model. Malaysia AAIB road equivalent. ARI-BUET Expressway investigation (Imad case) demonstrates what independent investigation can produce even with current capacity. UK, Malaysia ARI-BUET has investigative capacity. Requires legislative mandate and operational independence. P1 — SYSTEMIC

MISSING DATA AUDIT

Dataset Who Should Hold It Current Status Why It's Missing Investigative Impact Upgrade Path
BRTA Bus Crash Registry
Crash history by vehicle chassis
BRTA Closed / Not Public Classified. Enabling bribery ecosystem: operators pay to suppress negative records. Political sensitivity of permit holder registry. Would confirm or deny Phoenix ownership chains definitively. Would enable operator-level death rate calculation for the first time. RTI application under Information Rights Act 2009. Litigation if denied. International pressure via WHO/World Bank reporting.
Hospital Trauma Records
Injury cause + vehicle type
DGHS / DNCC hospitals Not digitized / Not linked No standardized injury cause coding in most hospitals. Paper records not aggregated. Hospital death ≠ road accident death in official statistics. Hospital records are the only ground truth for the 5.8× underreporting claim. Without them, the WHO estimate cannot be validated at Bangladesh hospital level. DGHS-CIPRB National Injury Survey (2023) is a partial proxy. Full hospital linkage requires health ministry mandate.
Police FIR Operator Name Field
FIRs linked to company, not just plate
Bangladesh Police Not collected systematically FIRs record registration numbers, not operator brand. This is a recording practice decision — the information exists (BRTA has it) but is not linked at point of FIR. Without this, operator-level accountability is impossible. Every fatality record is anonymous at the company level. BRTA–Police data sharing protocol. Single policy change that requires no new infrastructure.
Route Permit Beneficial Owner Registry
Actual individual behind each permit
BRTA Classified Many permits held by front companies or politicians' associates. Publication would expose political ownership of the sector and undermine the protection racket. Would confirm or deny Phoenix ownership chains. Would establish whether enforcement failures are random or targeted (i.e., connected operators never lose permits). RTI application. OCCRP/ICIJ corporate database cross-referencing. Company registry search under Companies Act 1994.

METHODOLOGY DOCUMENTv1.0

Purpose
This methodology is published so that any external journalist, researcher, or auditor can: (a) understand how every claim in this report was constructed; (b) identify its limitations; (c) replicate or extend the research; (d) cite it accurately. It is versioned — changes are logged in the errata section. This document is a living record.
M-001
Primary Data Sources
This investigation drew on the following source categories, in order of evidential weight:
  • Government & Regulatory Records: BRTA Annual Reports, ARI-BUET published investigation reports, DGHS statistical bulletins, BRTA route permit blacklist (Dec 2021). All treated as primary unless contradicted by multiple independent sources.
  • Supranational Bodies: WHO Global Status Report on Road Safety 2023; World Bank "Delivering Road Safety in Bangladesh" (2022); Asian Transport Observatory Bangladesh Road Safety Profile (2025). Treated as primary — note that WHO country estimates are modelled, not enumerated.
  • NGO Research (Primary): Transparency International Bangladesh (TIB) "Integrity in Private Bus Transport Business" (2024 Executive Summary and full report); Bangladesh Jatri Kalyan Samity (BJKS) Annual Reports 2018–2024; Road Safety Foundation (RSF) Dhaka 2025 Annual Report.
  • Court Records (as cited in media): Charge sheets, conviction records, remand orders. These are treated as primary when the media report includes docket/case numbers or direct quote from court documents.
  • Major Media Archives: The Daily Star, Prothom Alo, Dhaka Tribune, TBS News (The Business Standard), Financial Express, Daily Sun. Used as secondary unless they are the only source for a specific incident — in which case labeled SINGLE SOURCE in confidence tier.
  • Academic Research: BUET-ARI crash severity studies, MIST conference papers. Treated as primary when peer-reviewed or formally published.
M-002
Search Strategy
  • Languages: English and Bengali (Bangla). Bengali-language sources searched using operator names in Bangla script.
  • Date Range: July 2018 (Jabal-e-Noor incident / Safe Road Movement origin) to January 2026. Incidents before 2018 referenced where directly relevant to Phoenix chain.
  • Keyword Sets — English: [operator name] + [accident OR crash OR killed OR fatality]; "road accident Bangladesh"; "bus crash Dhaka"; "BRTA blacklist"; "waybill system"; "Suprobhat" / "Victor Classic" / "Akash Paribahan" / "Jabal-e-Noor"; "safe road movement 2018"
  • Keyword Sets — Bangla: [সুপ্রভাত পরিবহন], [আকাশ পরিবহন], [জাবালে নূর], [সড়ক দুর্ঘটনা ঢাকা], [বিআরটিএ কালো তালিকা]
  • Search Platforms: Google News, Prothom Alo archive, Daily Star archive, TBS News archive, Bangla Tribune, bdnews24, Google Scholar, BUET-ARI repository.
  • Archive Access: For articles unavailable via direct access, Wayback Machine (archive.org) used where available. Accessed links logged with date of access.
M-003
Inclusion / Exclusion Rules
  • INCLUDE: Any incident in the Dhaka Metropolitan Area (DMA) involving a city bus (not intercity/long-haul, unless the incident occurred within DMA boundaries) resulting in at least one fatality, where: operator name is mentioned in at least one source, or bus registration number allows operator identification via BRTA records (where accessible).
  • INCLUDE: Any BRTA enforcement action (permit suspension, cancellation, blacklisting) regardless of resulting fatality.
  • EXCLUDE: Incidents where operator is unidentified and cannot be traced via registration number. Excluded with notation.
  • EXCLUDE: Incidents where fatality occurred due to bus being struck by another vehicle (bus not the primary causal factor). Assessed case-by-case; decision logged.
  • EXCLUDE: Intercity/long-haul incidents occurring outside Dhaka DMA — with the exception of the Imad Paribahan Expressway case, included because it demonstrates BRTA fitness certificate enforcement failure as a systemic claim.
  • BORDERLINE CASES: Reviewed by at least two sources before inclusion. Decision and rationale noted in dataset notes column.
M-004
Deduplication Logic
Multiple media reports of the same incident are treated as a single data point. Deduplication applied by: (1) incident date, (2) location, (3) victim name where stated. Where victim names conflict across sources, the source closest to the official record (police/court) is treated as authoritative. Where victim names are absent, deduplication uses location + date + reported death count. Deduplication decisions logged in dataset notes field.
M-005
Handling of Rebrands / Aliases
  • A "rebrand" is recorded only when at least one of the following conditions is met: (a) same route operated post-ban; (b) media report explicitly links old and new name; (c) eyewitness or victim family testimony states ownership continuity; (d) same registration numbers appear under new name.
  • The confidence tier of the rebrand link is assessed independently of the confidence tier of associated incidents. A confirmed incident under a new brand name does not by itself confirm ownership continuity.
  • All rebrand links are labeled with the specific evidence type supporting the link, per the Phoenix Evidence Rulebook (Phoenix tab).
  • Rebrand links supported only by eyewitness testimony are labeled ALLEGED, not CONFIRMED or PROBABLE.
M-006
Temporal Boundaries
  • Primary research period: 29 July 2018 (Jabal-e-Noor incident) to 31 January 2026.
  • Statistical data: Uses most recent available year from each source. BJKS uses 2024 data; RSF uses 2025 data (Dhaka city); WHO uses 2021 modelled estimate (most recent as of this writing).
  • Cross-temporal comparisons must acknowledge that different datasets cover different years. The BJKS 2024 figure (8,543) is national; the RSF 2025 figure (219) is Dhaka-only. These are not directly comparable and are not presented as such in this report.
M-007
Quantitative Claims — Derivation Standards
  • Point estimates (e.g., "5.8× underreporting") are derived from simple arithmetic on two sourced figures. Both source figures and the calculation are stated.
  • Range estimates (e.g., "2.5–5.1% of GDP") are preserved as ranges. Point estimates are not extracted from ranges without explicit justification.
  • Percentages derived from this dataset are labeled as such and include the denominator. E.g., "47% of victims were pedestrians (103 of 219, RSF 2025)" — not "47% of victims."
  • Extrapolations (applying a national rate to Dhaka) are explicitly marked as inferences and their limitations stated.
M-008
Update Policy
  • This report will be updated when: new primary sources are published; an error is identified; an operator provides a formal response; court proceedings conclude on referenced cases; new BRTA enforcement action is taken.
  • Updates will increment the version number (v1.1, v1.2…). Major structural revisions increment the major version (v2.0).
  • All updates are logged in the Corrections & Errata register (Legal tab). No retroactive editing of text without a versioned correction notice.
  • Dataset downloads (Data tab) are versioned separately and dated on each file.
M-009
Known Blind Spots & Structural Bias
  • Media coverage bias: BJKS and RSF methodologies rely on newspaper coverage. High-profile incidents in Dhaka receive disproportionate coverage vs. peripheral areas and night incidents witnessed by few.
  • Language bias: This investigation has stronger coverage of English-language sources than Bangla-language sources due to researcher language access. Bangla-only coverage may contain incidents not captured here.
  • No official BRTA data: The absence of a public BRTA crash registry means this dataset is constructed from secondary sources. Operator-level figures are patterns, not enumerations.
  • Right of reply not yet received: As of v1.0, no operator named in this report has provided a formal response. Claims against named operators may be incomplete or contain inaccuracies. See Legal tab.
  • Dhaka-city vs. DMA vs. national: Different sources use different geographic boundaries. This is noted in each chart but remains a structural ambiguity.
M-010
Corrections & Errata Policy
  • Factual errors: corrected in the body text with visible markup ("CORRECTED [date]: original text → corrected text") and logged in the Legal tab Errata Register.
  • Evidence tier upgrades/downgrades: change logged with rationale; previous tier preserved in history.
  • Corrections requested by operators: reviewed against source evidence; accepted if documentation supports the correction; disputed corrections logged with both positions.
  • Anonymous corrections not acted upon without supporting documentation.

EVIDENCE TIER DEFINITIONSv1.0

Purpose & Scope
This framework defines the four evidence tiers used throughout this investigation. Every factual claim, every table cell, and every map element carries one of these tiers. These definitions are public and fixed for v1.0 — they cannot be retroactively changed. If a tier definition needs updating, a new version must be issued with a correction log entry.
TIER 1
CONFIRMED
Minimum Threshold:
  • The claim is supported by ≥2 independent primary or official sources, OR
  • The claim is supported by 1 official/government primary source + 1 corroborating media record, OR
  • The claim is supported by a published court record, official investigation report, or government statistical publication with no credible contradiction in secondary sources.
Usage Rules:
  • ONLY Confirmed claims may be used in headlines, section titles, or pull quotes without qualification.
  • Confirmed claims should still be attributed: "According to WHO..." or "TIB documents..." — not stated as self-evident facts.
  • A "Confirmed" incident does not automatically confirm operator identity, ownership, or intent. Each element is assessed separately.
Examples in this report:
  • Abrar Ahmed Choudhury killed on 19 March 2019 by Suprobhat bus — CONFIRMED (court charge records + 3 major papers)
  • BRTA recommended cancellation of 25 operators' permits (Dec 2021) — CONFIRMED (TBS + bdnews24 + primary BRTA announcement)
  • WHO estimates ~31,578 annual road deaths in Bangladesh — CONFIRMED (as WHO estimate — note: modelled, not enumerated)
TIER 2
PROBABLE
Minimum Threshold:
  • The claim is supported by 1 primary source (official, NGO, or verifiable investigative report) where no contradicting source exists, AND
  • The claim is consistent with corroborating documented patterns (e.g., same operator involved in multiple incidents across different sources), OR
  • The claim is a direct inference from two or more Confirmed claims following sound logical reasoning.
Usage Rules:
  • Must be qualified in body text: "evidence suggests," "available records indicate," "is consistent with..."
  • Cannot be used in isolation in headlines without qualification.
  • If cited by others, they must reproduce the "Probable" tier label.
Examples in this report:
  • Victor Classic operated same route as banned Suprobhat post-2019 — PROBABLE (route continuity confirmed; ownership continuity from Daily Sun investigation — single source)
  • TIB: 92% of operators politically connected — PROBABLE (single TIB source — not independently replicated)
TIER 3
ALLEGED / UNVERIFIED
When This Label Is Mandatory:
  • The claim originates from eyewitness testimony, victim family statement, or anonymous source without documentary corroboration.
  • The claim is made in media reports but the media source had access only to secondary information.
  • The claim involves intent, motive, or conspiracy (e.g., "deliberately rebranded to evade prosecution") — intent claims require direct evidence and cannot be inferred.
  • The claim is sourced from a single non-official source without corroboration.
  • Right of reply has been sought from the named party and not yet received.
Usage Rules:
  • Must use explicit language: "it is alleged that," "according to [named source], who claims," "evidence suggests but has not been independently confirmed."
  • The source of the allegation must always be named. Anonymous allegations cannot be published under this tier.
  • Legal action risk is highest at this tier. Review by legal counsel before publication is strongly recommended.
Examples in this report:
  • Akash Paribahan is owned by same individual as Victor Classic / Suprobhat — ALLEGED (victim family + eyewitness testimony, not corporate record)
TIER 4
UNVERIFIABLE
Definition:
  • The claim cannot be verified through currently available open sources because: the underlying data is classified/not public; sources are contradictory without resolution; the required dataset does not exist (see Data Gaps).
Usage Rules:
  • Must be labeled as Unverifiable in both the text and any associated data tables.
  • Unverifiable claims may be included to document the existence of a gap — not as evidence of the claim.
  • Never cited in support of other claims.
Examples in this report:
  • Specific annual fatality count attributable to each individual operator — UNVERIFIABLE (no public BRTA crash registry; media data insufficient for census)
CROSS-CUTTING RULE
Headline and Title Standard
  • ONLY Tier 1 (CONFIRMED) claims may appear in report headlines, section titles, or introductory statements as unqualified factual assertions.
  • Tier 2 (PROBABLE) claims may appear in headlines only with explicit qualification in the same sentence ("evidence suggests," "records indicate").
  • Tier 3 and 4 claims must not appear in headlines. They may appear in body text with mandatory qualification language.
  • Visual elements (charts, maps, tables) must display the tier label for each claim they represent.