The Social Life of a Fuel Shock

Companion to impact_analysis.md (impact analysis) and research_log.md (source log). Drafted 21 September 2026; P1–P5 recent-context notes added 27 September 2026; P6 added 28 September 2026.

The economic analysis says what the Tk 20 hike does to prices. It can't say how people experience it, how they cope, whom they blame, or how the shock changes social relations. That is the sociological gap. This matters for three reasons:

1. Two policies landed 48 hours apart. The National Pay Scale 2026 was gazetted on 19 September and the fuel hike took effect on 21 September. This splits the population cleanly by occupation: ~33 lakh public employees got pay rises of up to +142%, while ~7.5 crore private and informal workers got nothing. This rarely happens in the real world and is unusually useful for research.
2. More is coming, which makes this a baseline moment. BPC cannot fund its letters of credit in October, so another hike or a duty cut is near-certain. After that, the Boro season (Dec–Mar) will pass through irrigation costs. Anything measured in the next 2–4 weeks can serve as a pre-shock baseline. That window closes fast.
3. Bangladesh's previous shock was not studied much. The 2022 hike (diesel +42.5%, Tk 80→114) produced protests, fare disputes, and a lot of commentary. Very little systematic household-level sociology came out of it. 2026 can be studied properly and compared with 2022.

Part A — Open research questions

Grouped by theme. ★ marks questions that are time-critical (the baseline must be collected before late October 2026).

A1. Coping, sacrifice and household bargaining

• ★ How do households at different income levels cut back after a fuel/transport shock: food quantity, food quality, children's schooling, healthcare, remittances home, or savings? In what order, and does the order differ by who controls the household budget?
• Who inside the household absorbs the cut? Do women and girls eat less, or give up mobility (unaccompanied travel, trips to work, school or clinics) first? (P6 measures this directly, from both the woman and the man in the same household, rather than from a single household-level respondent.)
• How much does borrowing increase (NGO microcredit, shopkeeper credit (baki), kin, mobile-money loans), and does it become debt that households can't escape?
• Does kerosene use increase or fall among off-grid and load-shed households, and what replaces it (candles, solar lanterns, darkness)? What does that do to children's study hours?

A2. Relative deprivation and the pay-scale wedge

• ★ Do private-sector and informal workers compare themselves to government employees now that the pay scale and the hike landed together? Does that comparison predict grievance, lower trust in the state, or willingness to protest better than their own absolute income loss does?
• Inside mixed households (one government earner, one private or informal earner), how are the gains shared? Do they cause friction?
• Does resentment get framed as class ("the rich"), sector ("government people"), party ("their people"), or geography ("Dhaka")?

A3. Informal transport: the e-rickshaw as the poor's shock absorber

• Who are the new battery-rickshaw drivers of 2026? What share are displaced farmers, ex-garment workers or returned Gulf migrants? What pushed them from where they were, and what pulled them into driving?
• ★ How do fare-setting norms form when the official gazette is ignored? Who holds power at the fare negotiation: passenger, driver, garage owner, or local enforcer?
• How is illegal charging governed? What is the informal economy of charging points: who sells the electricity, who protects them, how much do they pay in rents or bribes? How does that change during load shedding?
• If the state cracks down on e-rickshaws, where do those livelihoods go?

A4. Farmers, irrigation and the Boro decision

• ★ How will farmers decide how much Boro to plant (Nov–Dec 2026) given diesel at Tk 135+, a frozen procurement price and an uncertain supply? Which farmers cut back: sharecroppers, smallholders or large farmers?
• How do irrigation-service markets (pump owners selling water to neighbours) re-price water? Do tenant farmers bear more of the cost than their share?
• Do farming households send members into the cities (to rickshaws or RMG) during this season, feeding back into A3?

A5. Garment workers, time and the city

• How do RMG workers change their commute (walking longer, moving closer to the factory, sharing rooms), and what does that do to their time, safety and health, especially for women on night shifts? (P6's RMG booster subsample targets this directly.)
• Does the combined pressure of energy shortage, transport costs and food prices change how workers organise at the factory level (wildcat stoppages, demands for transport allowances)?
• How much do the extra Tk 260–300/month in realised bus fares (~2–2.4% of the RMG minimum wage, README §4b), on top of food pass-through, feed into the demand for a new minimum-wage board?

A6. Legitimacy, blame and protest

• Whom do people blame: the government, BPC, the Hormuz war, India, "syndicates", or hoarders? How does blame vary by partisanship, sector and information source (TV, Facebook, WhatsApp, YouTube, the local mosque or bazaar)?
• Does the PM's public apology (5 Sep) and the "tolerable level" framing make the hike more or less acceptable?
• Moral economy test: do people accept higher prices when they understand them as justified by the global market, and reject them when they see them as unfair (duty kept in place, pay rises for government employees, rumours of smuggling)? Does the sense of being treated unfairly predict protest better than the size of the price rise?
• How does online sentiment about fuel on Bangla Facebook and YouTube move against real events (hikes, bus fare gazettes, load-shedding peaks)?

A7. Compensation and the state

• ★ If the government introduces targeted compensation through the Family Card, who is excluded, and why (documentation, mobile-money access, local political gatekeeping)?
• Is cash compensation seen as a right or as charity, and does that change how legitimate the hike seems?
• Do people prefer universal cheap fuel or targeted cash? Does the answer change when the trade-off is explained?

A8. Comparative and historical

• How do coping and protest in 2026 compare with 2022, the last major hike, among the same kinds of households?
• Why did Nigeria's 2023 subsidy removal, Kenya's 2023 fuel VAT increase, Sri Lanka's 2022 crisis and Bangladesh's 2026 hike produce such different levels of protest?

Part B — Research proposals

Six proposals, ordered by urgency and by feasibility for a small team. Each can stand alone, though P1 is the backbone that P5 and P6 attach to.

P1. The Fuel Shock Panel — a rapid-response household phone panel (flagship)

Question. How do coping, well-being and trust in the state change over the fuel shock cycle, and how does that path differ between public-sector, formal-private and informal households?

Why it's important. The 19/21 September timing splits households by sector, and a second hike is likely in Oct–Dec. A panel that starts now captures the baseline, the October "cliff", Boro and LDC graduation in one dataset. Nothing like it exists for Bangladesh's 2022 hike.

Design.

• Sample: 1,500–2,500 households across Dhaka, Chattogram, Gazipur/Narayanganj (RMG belt), Rangpur and Rajshahi (Boro, diesel irrigation) and one coastal district. The sample is stratified by the main earner's sector: government / formal private (incl. RMG) / informal non-farm (incl. transport) / farm. It is recruited from an existing sampling frame if a partner lends one (BIGD, SANEM or a survey firm). Otherwise use random-digit dialling with quotas.
• Waves: W0 now (baseline, by mid-Oct) → W1 Nov → W2 Jan → W3 Mar → W4 Jun 2027 (after the Boro harvest). Each is a 15–20 min CATI (phone) interview.
• Core modules:
  • spending cuts, measured with a list of 12 items, each ranked 1st/2nd/3rd
  • food insecurity (FIES, 8 items)
  • transport mode, time and cost for the main commute
  • kerosene/lighting and study hours
  • borrowing
  • intra-household allocation (who ate less)
  • relative deprivation (comparison to government employees)
  • blame
  • trust in institutions
  • protest willingness and participation
  • Family Card receipt
• Embedded experiment in W0: randomise the explanation shown before asking whether the hike is acceptable: (a) nothing, (b) "global oil prices doubled", (c) "Tk 32/L of the price is government duty", (d) "government employees just received a pay rise". This measures how framing and perceived fairness shift acceptance, which is the moral-economy test.

Analysis. Panel fixed effects. Compare trajectories between sectors (difference-in-differences with sector × wave). Heterogeneity by gender of respondent and income tercile. Treatment effects for the framing experiment.

Outputs. A public dashboard updated after each wave, policy briefs timed before the Boro and budget decisions, and 2–3 papers (coping, relative deprivation, moral economy).

Budget (indicative). ~$35–60k for 5 waves at ~$4–6 per completed CATI interview, plus the analyst, the survey firm's management fee and incentives (Tk 50–100 mobile top-up per wave).

Risk. Phone surveys under-represent the poorest and women. Mitigate with quotas, asking for women respondents specifically, and a small in-person top-up.

Recent context (27 Sept). The National Pay Scale (+142% for ~33 lakh public employees) landed 48 hours before the Tk 20/L hike, insulating public employees while ~85% of employment (informal) faces the shock unmitigated. The government's stated mitigation, the Family Card (a digital safety-net database, ~1 crore households, launched March 2026), is tracked wave by wave in P1 for coverage gaps — documentation barriers, missing mobile-money access, and Union Parishad-style political gatekeeping are the likely exclusion channels.

P2. Charging the City — an ethnography of the e-rickshaw economy under load shedding

Question. How do battery-rickshaw drivers, garage owners, charging-point operators and local power brokers govern an illegal energy economy? How does the fuel hike redistribute risk and income across that chain?

Design.

• Sites: 3 neighbourhoods in Dhaka (one core, one peri-urban, one where a crackdown or relocation is active) and 1 district town.
• Methods:
  • 4–6 months of participant observation at garages and charging points
  • 60–80 life-history interviews with drivers, sampled for ex-farmers, ex-RMG workers, returned migrants and women in the business (very rare, so an important case)
  • driver daily diaries (earnings, fares, charging costs, rent, "costs to police or fixers", hours) over 8 weeks, collected via WhatsApp voice notes, with a small stipend
  • mapping of charging points to cross-check the reported 48,136 illegal vs 3,300 authorised
• Mixed-method add-on: a structured survey of 400 drivers at charging points to estimate the share who moved into rickshaws from each prior occupation.

Outputs. An ethnographic paper on informal energy governance, a policy note on legalisation and metering (lever #5 in the README), and a photo-essay (with consent).

Budget. ~$15–30k (researcher stipend, RA, diaries, transcription, translation).

Recent context (27 Sept). The Tk 20/L hike widens the petrol-vs-battery-rickshaw operating-cost wedge from ~15:1 to ~17:1 (petrol ~Tk 11/km vs. ~Tk 0.4–0.8/km on illegal charging), pushing more of the shock onto the informal charging economy this study observes directly.

P3. Boro Decisions — farmers, water sellers and the irrigation cost shock

Question. How do diesel costs, supply uncertainty and a frozen procurement price change farmers' decisions about how much Boro to plant? Who in the agrarian hierarchy absorbs the cost?

Design.

• Sites: 6 villages in 2–3 upazilas, in Rangpur, Rajshahi and Mymensingh (diesel shallow-tube-well heavy), with 1 grid/solar-irrigation village as a contrast.
• Data collection:
  • a survey before planting (Nov) and at harvest (May), n≈600 plots or farmers: tenure, planned vs actual area, irrigation arrangement and price, borrowing, household migration
  • interviews with pump owners and water sellers about how they set prices
  • focus groups with women in farming households on food and labour trade-offs
• Administrative data: Department of Agricultural Extension (DAE) Boro area targets vs actuals, and BADC diesel/pump data if accessible.

Why it's important. This is the channel most likely to turn into a 2027 food-price and forex problem. Almost no one is watching it yet (see §4c of the README).

Budget. ~$20–35k.

Recent context (27 Sept). The Tk 20/L hike alone adds an estimated Tk 1,070 crore to the coming Boro crop's irrigation cost, against a frozen procurement price — the margin-compression mechanism P3 traces from farmer decision to acreage, right as BPC's October financing crunch makes a further hike likely before planting.

P4. Who Gets the Blame? — computational analysis of public discourse on the fuel shock

Question. How are responsibility and fairness framed in Bangla-language public discourse about fuel between 2022 and 2026? Which frames predict offline protest?

Design.

• Corpus:
  • Bangla and English news (Prothom Alo, Daily Star, TBS, Jugantor, Samakal, Kaler Kantho)
  • public Facebook pages and groups, YouTube comments on news clips, with platform terms and data-access rules followed
  • parliamentary debates
  • covering the 2022 hike, the April, June and September 2026 hikes, and the upcoming October–December 2026 events
• Methods:
  • Bangla NLP: topic models, a frame classifier (global market / government failure / corruption-syndicate / India / fairness-pay scale / apology-empathy), sentiment
  • event data on protests (coded from news, ACLED where available)
  • time-series linking frame intensity to protest events and hike dates
• Validation: human-coded sample (inter-coder reliability ≥0.7), plus a check against the P1 survey blame questions (the same framing categories).

Budget. Low: ~$5–15k (RA coders, compute). The most feasible solo or remote project, and a strong companion paper to P1.

Recent context (27 Sept). The 48-hour pay-scale/fuel-hike collision is itself a discourse event worth its own row in the corpus — a real-world test of the fairness-pay-scale frame, cross-checkable against whether P1 respondents cite the pay-scale contrast unprompted in their blame module.

P5. Cash or Cheap Fuel? — a survey experiment on preferences for compensation

Question. Do Bangladeshis prefer universal cheap fuel or targeted cash compensation? How do trust in the state, perceived corruption and past experience with safety-net programmes shape that preference?

Design.

• A conjoint experiment embedded in P1's W1 or run on its own (n≈2,000).
• Respondents choose between policy packages that vary: fuel price (Tk 135/150/165), compensation (none / Tk 500 / Tk 1,000 per month), targeting (universal / Family Card / employer-based), delivery (bKash/Nagad / union parishad / NGO), and who pays (duty cut / budget / BPC borrowing).
• Moderators: sector, prior safety-net receipt, trust, partisanship.

Why it's important. It directly informs the choice the government faces in Oct–Dec. It adds evidence from Bangladesh to the international literature (Iran 2010, Indonesia, Nigeria 2023) on why fuel subsidy reform so often fails politically.

Budget. ~$8–15k standalone, and small if embedded in P1.

Recent context (27 Sept). Current mitigation discourse centres on the Family Card, but overlooks the kerosene burden of off-grid households; P5's targeting/delivery attributes will show whether the actual losers from this hike trust that mechanism or default to Union Parishad/NGO channels on a fundamental trust deficit.

P6. Whose Mobility? — a gender module paired with the P1 panel

Question. How does the fuel shock reshape gender inequality inside the household — who cuts back first, whose mobility is curtailed, what obstacles women specifically face in coping — and how much of the gender gap in coping is a gap in awareness of the shock itself, not just in exposure to it?

Why it's important. P1's own Module 5 asks one respondent (usually whoever answers the phone, more often the male household head) who ate less and who gave up mobility. That is a single person's account of the whole household. P6 asks the same underlying question of both the female and male adult in the same household, separately, which turns a self-report into a measurable intra-household discordance — plus an objective knowledge test, not just self-rated concern, to see whether women are actually less informed about the hike or simply report caring about it differently.

Design.

• Sample: a paired-respondent subsample of ~600–800 households drawn from P1's existing sample (adds the opposite-gender second respondent to each), plus a standalone booster of ~300–400 women-headed households and RMG women night-shift workers — populations P1's single-respondent design and 40% female-respondent quota still undercount.
• Waves: shares P1's calendar (W0 → W1 → W2 → W3 → W4), so gender gaps are tracked across the same shock cycle P1 tracks generally.
• Interviewer requirement: female respondents interviewed only by female interviewers — a real staffing cost this module carries that P1 does not.
• Core modules: an awareness/knowledge test (objective, not just self-rated); a 7-day mobility diary (trips made, shortened, cancelled or reassigned, by purpose, plus any safety incident on a cheaper/crowded mode); gendered coping and structural obstacles (unpaid-labour substitution, forgone reproductive/antenatal care, girls' school attendance, and whether the Family Card's NID/mobile-money access route sits in the woman's own name or the male head's); a commute/safety module for the RMG night-shift booster; bargaining/decision-making asked of both respondents identically; and an embedded vignette measuring the fare increase at which a market or clinic trip gets curtailed, by gender.

Analysis. Within-household paired comparison (female vs. male respondent, same household, same wave) on awareness, reported mobility loss, and cost thresholds. Panel fixed effects across waves, as in P1.

Budget. ~$20–35k incremental if piggybacked on P1's sample, waves and survey-firm contract (recommended); ~$30–45k if fielded standalone.

Risk. Reaching a named second respondent from a household that shares one phone (usually registered to the male head) without triggering suspicion or a proxy answer; mitigated with a strict privacy-check-and-reschedule protocol (see p6_instrument.md §0) rather than proceeding with whoever picks up.

Recent context (28 Sept). RMG employment — widely reported as majority-women, though not independently verified by this project — carries a minimum wage frozen since December 2023 (README §2), so the realised bus-fare increase A5 identifies lands on a wage that hasn't moved. The National Pay Scale that insulated public employees from the hike is overwhelmingly a formal, urban, disproportionately male benefit. And the Family Card's NID/mobile-money access route is frequently registered in the male household head's name — meaning the shock's own mitigation tool may be structurally harder for women to reach even inside a qualifying household. P6 is built to put numbers on all three gaps, wave by wave.

Part C — Data you can use without collecting anything

Part D — Sequencing (what to do this week)

1. This week: send out outreach emails (see outreach/research_collaborators.md). Ask each partner for a sampling frame, affiliation or host, and quick feedback on P1 W0.
2. By ~5 October: finalise the P1 W0 questionnaire (bilingual, piloted with 30 respondents) and the ethics plan. Get IRB approval through a host institution (BIGD, a DU department or a partner university), or a local ethics board.
3. By ~20 October: complete W0 fieldwork. It has to be before the October price or LC events. If W0 slips into November, P1 loses its baseline. P6's paired-respondent recruitment (and hiring female interviewers) should be arranged alongside P1's fieldwork prep, not after, so the second respondent can be reached in the same W0 window rather than as a late add-on.
4. In parallel, starting now: P4 needs no fieldwork, so begin scraping news and building the corpus immediately.
5. November: P3 survey before Boro planting. Scope P2 with 2–3 pilot sites.

Ethics notes

• Informed consent in Bangla. For phone surveys, record verbal consent.
• The e-rickshaw and charging research touches illegal activity. Never collect names alongside information about illegal activity, store diaries encrypted, and do not share site-level locations of charging points publicly.
• Protest and blame questions can be politically sensitive. Do not record partisanship alongside identifiers, and follow a data-minimisation plan.
• Pay respondents fairly for their time (top-up), especially in diary studies.